Where I come from

I am the fifth generation of my family to live and work on the same piece of land in the prairies of Saskatchewan, Canada. For over a century, my family has read the sky, worked the soil, and depended on the land staying productive for the next generation. On the farm, I learned a simple rule: you don’t take from tomorrow to pay for today.

That rule is harder to follow than it sounds. A modern farm runs on diesel, fertilizer made from natural gas, and grid power. In the last few years, fuel and fertilizer prices have swung so wildly that if you bought at the wrong time, your profit for the year evaporated. For my grandfather, fuel and fertilizer were a fraction of the total cost. Now they are among the largest.

Looking for the right problem

I left the farm to find an answer to one question: how do we power a modern farming operation, and everything else, without drawing down resources that won’t come back?

I’ve spent most of my career chasing pieces of that answer. Right out of school, I joined Areva (now Orano) at McClean Lake in northern Saskatchewan to work on SABRE, a way of mining uranium by drilling from the surface and cutting ore out with a high-pressure water jet. I ran a shift during the 2014 campaign, the first time the program pulled a significant amount of ore with equipment built for the method, and that season locked in most of the techniques the program still uses. Then the project slowed into a long design phase, and after two years I left to find work where I could build things faster. The first commercial SABRE mining started in 2025, more than a decade after our campaign.

Alongside all of that, I was building a company. A few friends and I founded Innocorps Research in university, starting with equipment built in my garage, and I worked other jobs to keep it going while we patented a method for recovering clean water from industrial waste and black water. That work took us into the Department of National Defence’s IDEaS Pop Up City contest, a four-round competition to rethink how the Canadian Armed Forces supply temporary camps with water, power, and waste handling. Our Circular City team won the $2 million grand prize at Canadian Forces Base Suffield in 2023, cutting solid waste tenfold and recycling the camp’s wastewater into drinking water that met Health Canada guidelines. I drank it straight from the output before the lab results came back, and I’m still here. But a prize isn’t a procurement contract, and a small company can’t wait years for one. Eventually I couldn’t justify paying myself anymore, and I moved on.

Then my brother and I went back to the farm and built off-grid homes, with the goal of producing more than they used. We designed for thermal efficiency first: thermal mass walls to store winter sun, and roof overhangs sized to let the low winter sun in and shade the glass from the high summer sun. Solar panels and batteries supplied the power, water and wastewater were treated on site, and our second building was coupled to a four-season greenhouse that grew food year round. The first one overshot and ran too hot to live in, so it became our workshop. They hold up from -40°C to +40°C. A house can collect enough sunlight to run itself, but a farm, let alone the plants that make its fuel and fertilizer, needs far more energy than that.

Each of those projects solved a piece of the problem but none got to the core. SABRE showed me how long good technology can take to reach the field. Innocorps showed me that technology that works still loses if it can’t win on cost and timing. The houses showed me that sunlight can run a household, but not an industry. Back on the farm, we still needed diesel and fertilizer to turn a profit, and fossil fuels keep winning because they are cheap, dense, and available on demand. Anything that replaces them has to beat them on cost.

Finding Terraform

I was researching ways to turn solar power into fuel and fertilizer, the two biggest bills on the farm. Fertilizer interested me most: the nitrogen is free in the air, and the hard part is making the hydrogen and finding the energy to break the nitrogen bond. But I was still looking at fuel too, and that research led me to episode 36 of the S3 YouTube channel, which covered Terraform Industries. They had just run their first full-loop Terraformer, making methane from air, water and electricity, with the goal of producing it cheaper than gas pulled out of the ground. It was almost exactly what I wanted, because farm machinery can run on methane. The only problem was that there was nothing I could buy yet. So I decided to go help build it.

By then, I was mostly a seasonal hand on the farm, and my dad, uncle and brother could handle the work without me. I wanted a challenge, so I applied. A day or two later I had a screening call with Casey Handmer, Terraform’s founder and CEO. The way he thought about the problem convinced me the project was feasible. After a deep dive with one of the DAC team members, they invited me for an onsite and gave me a week to prepare, along with a homework problem: design a flash calciner with cyclones that could calcine carbonate material. Flash calciners with cyclone separation are a standard industrial method, and at the time it was the path Terraform’s calciner was on. It later proved infeasible for our process, but that homework was my first look at the problem I’ve been working on ever since. I flew from Saskatchewan to Burbank, from farm life to the biggest metro area on the West Coast, did the day-long onsite, and got a job offer that evening. They wanted me to start in three weeks and they wanted me to lead the Direct Air Capture team, to pull carbon dioxide (CO2) from the air as cheaply as possible.

The problem

We had to build three major components that work together in a loop to pull CO₂ out of the air and deliver it as a pure stream suitable for a reactor. This is called the Lime cycle. The carbonator moves a large volume of air past the sorbent, which captures CO₂ and turns into carbonate. The calciner heats the carbonate to release pure CO₂ and leaves quicklime behind. The hydrator turns the quicklime back into fresh sorbent so the loop can start again. None of the three is as simple as it sounds, and that’s where doing it the hard way starts.

Pulling CO₂ out of the air

To make fuel from air, you first need a pure stream of CO₂. CO₂ in the atmosphere is about 425 to 430 parts per million, so you have to move an enormous amount of air to collect a useful amount of it. On a volume-to-volume basis, it’s roughly one million to one.

Our Direct Air Capture (DAC) system uses hydrated lime, Ca(OH)₂, which reacts with CO₂ in the air to form calcium carbonate, CaCO₃. The chemistry is old and well understood. The hard part is material handling. Lime wants to be a powder, and powder blows away. We have to form it into a structure that holds together and doesn’t turn to dust, leaves room for air to flow through, and still reacts fast enough to be worth the energy spent moving the air. That means hitting several narrow windows at once: humidity, temperature, and how the material is formed. One of my first successes at Terraform was finding a way to form flakes effectively in large amounts using off the shelf equipment.

The team

When I started at Terraform, the DAC team was small: one senior engineer and two technicians. Several approaches to calcium looping had already been tried, and none had worked yet. Casey was upfront with me: I might be taking on a problem that didn’t have a viable solution, but the company needed one to succeed. I’m not one to back down from a challenge, so we got to work.

Russ, the senior engineer, had been on the team the longest and had seen nearly every method we’d pursued. His focus was hydration, but he got me up to speed on where the system stood and what had already been tried. He had an encyclopedic knowledge of the tests available to us and was good at explaining complex processes, like the energy flows through a system, with diagrams. Our two technicians, Jo and Eghan, each brought their own set of skills that made the builds possible, spanning trades and art forms from welding to pottery.

We built out the rest of the team with interns, starting with Jake and AK, and later Aggy. They came straight out of school with a lot of drive and energy, and all of them moved up the skill tree quickly and took ownership of projects across the three main branches of calcium looping. Along the way, Katie joined as a summer intern and Brian as a consultant. That was the team for the first year, and this year we’ve started expanding again.

Getting the CO₂ back out: the calciner

Once the CO₂ is locked up as limestone, we have to get it back out. That step is calcination, one of the oldest chemical processes people use. Lime has been made by heating limestone since the Neolithic.

Heating limestone drives off CO₂ and leaves quicklime, CaO. Pure CaCO₃ loses about 44% of its mass as CO₂ when it calcines. The CaO goes back to the hydrator and the loop starts over.

Industry does this every day in large kilns, but the normal way breaks every one of our rules.

Why conventional calcination doesn’t work for us

A cement or lime kiln burns fuel to make heat, and the CO₂ from the fuel and from the limestone goes up the stack together. We can’t burn anything, and the whole point is to capture the CO₂ that comes off. Industrial kilns also take days to reach temperature and run continuously until they need a rebuild. Because our power comes from solar panels, the calciner has to go from cold to operating temperature in about an hour each morning, run through the day, and cool down overnight. That combination rules out almost every conventional design and a lot of unconventional ones.

What we tried

Before building anything new, we went back to how calcination is done elsewhere. Rotary and shaft kilns, and the flash calciners used for fine material, all burn fuel, so the CO₂ they release comes out diluted with combustion gases. Vacuum calcination drops the decomposition temperature from about 900°C under one atmosphere of CO₂ to roughly 750°C at a tenth of an atmosphere, but nobody has made it work as a continuous process. Plasma calciners deliver enormous power in a small space, but they overshoot the temperature by thousands of degrees and wear out their electrodes. Electrochemical routes run at room temperature, but they use more energy and make Ca(OH)₂ instead of CaO. The closest match to what we wanted was the electrically heated fluidized bed, where the CO₂ released by the reaction is itself the gas that fluidizes the bed.

Terraform had also built numerous calciners of its own before I arrived, and each one taught us something. One of the early ones was a tube furnace: a quartz tube inside a clamshell of Kanthal wire heaters, loaded with a batch of carbonate and sealed. It made CO₂, but it was a batch process, and hot CaO attacked quartz. Next came what the team called the giant toaster, an array of plates with embedded heating wire that flakes flowed down between. The material sintered, bridged the gaps, and locked up the whole machine with no easy way to clean it out. The rotary valves at the top and bottom couldn’t handle the powder, and steam clogged the inlet and corroded the heating wire.

The Thor heater was the flash calciner with cyclones that Terraform was building when I interviewed. An array of silicon carbide elements heated a stream of CO₂ that carried the carbonate through a cyclone, then a recuperator cooled the gas and a blower sent it around again. It was partly built when I joined. After some initial tests on the gas heater, I ran the numbers and showed it couldn’t work: the gas couldn’t be heated enough in a single pass, and the cyclone would lose heat faster than the heaters could put it in. We changed direction before the buildout was finished.

So, I pursued the drop tube calciner instead. Silicon carbide elements surrounded a ceramic tube inside a sealed vessel, and material dropped through the heated zone to flash calcine above 1100°C. It worked in the way that mattered most: It produced CO₂ and fully calcined material. However, it never reached our purity or production targets. As a continuous process, it depended on feeding material in and taking it out without interruption, and we couldn’t make the feed and takeoff systems run reliably. The tubes were prone to failure, the CaO attacked the insulation, and too many parts were running at their limits.

We stepped back and ran a fast, cheap prototyping sprint, building and testing candidate approaches in weeks instead of months. Most were ruled out, some on paper before we built anything, and we settled on a new design.

Then testing surprised us. One of the hardest parts of the design turned out to be unnecessary, and taking that part out deleted a large share of the system’s complexity, along with the grinding, the particle sizing, and the material loss that came with it. We reused what had worked in earlier systems and deleted everything else. The result is a simple, electrically heated batch calciner that produces CO₂ at 99.9% purity and meets our production target.

We scaled this architecture from a simple prototype to a ~30 kW unit in just two weeks, brought it out to the Muroc test site, and operated our first fully solar-powered calciner, generating pure CO₂ directly from off-grid DC electricity. It was our tenth ground-up redesign, and now I understand why many other teams have not managed to solve this problem – they gave up after the ninth attempt.

The hard part for me

The hardest part wasn’t the technical problems. It was managing the team. The DAC system threw a wide spectrum of problems at us at once, and trying to cover all of them pulled my focus away from overseeing the work and delegating it well. Our leadership saw it before I fully did, and brought in Stephanie to manage the team so I could focus on the engineering. That’s why I spent the final stretch on the calciner’s latest prototype, the one that hit our production and purity goals and proved our models for scaling. It’s where I did my best work. Knowing where you’re most useful is part of doing it the hard way too.

Why this matters

Every one of these problems is solvable, but none of them are solved by copying what industry already does. That is what doing it the hard way means here: going back to first principles on processes that have barely changed in centuries, because the old way only works if you can burn fuel and vent the CO₂. Now, we have a calciner that works, and the next hard problem is building them by the thousands.

Someday I want the sixth generation on our land to fill the tractor with fuel made from the same sunlight that grows the crops. No more taking from tomorrow to pay for today.

Join us

We’re hiring. If you want to work on problems like the calciner, where the textbook answer doesn’t work and you have to build your way to one that does, there’s a place for you here. Open roles right now include Direct Air Capture Mechanical Engineer and Direct Air Capture Technician, along with roles across our electrolyzer, reactor, methanol, ammonia and solar teams. The full list is at terraformindustries.com.

Send your resume and one page of evidence that you can execute, like a summary of hardware you’ve designed, built and tested, to hiring@terraformindustries.com. I showed up from a farm in Saskatchewan with a homework problem and a week to prepare. If you build things, we want to hear from you. Every role is on site at our castle in Burbank.

My foray into hardware began in 2024, when I showed up to a construction site as an ironworker two weeks after quitting my tech job. I was the only crew member who wasn’t a felon and still had all my teeth.

Terraform Industries has a way of collecting intriguing misfits and catapulting them into eyewatering amounts of responsibility. I was hired June 2025 as Casey’s Technical Chief of Staff, essentially a high level critical path expeditor role. From October 2025 through September of 2026, I had the privilege of leading the charge on Terraform’s R&D test site called Project Muroc. We speedran the land acquisition, permitting, and construction of an offgrid fuel synthesis plant in Southern California. This was a crucible project for both Terraform Industries and myself personally.

It started in October of 2025 when our CEO Casey Handmer looked over to me after gracefully relieving the former construction lead of his duties and asked “Will you do it?” I had been there for 3 months and heeded the call. Thankfully, my esteemed colleague Tiana Wong was there to support and had diligently researched available parcels of land along with knowledge of the regulatory frameworks we would be dealing with.

The first thing we did was drive to Bakersfield and develop a map of all of the relevant parties who could both help or block our project as it advanced through the process. Journeying through the permit landscape is bespoke for every project and region, particularly when you’re trying to build something that involves high voltage and hazardous materials in a configuration that has never existed before. Parties we met with included the planning department, environmental health personnel, the air quality district, building reviewers, and consultants. I grabbed every business card within reach at every waiting room and secretary desk.

Some conversations were more productive than others, and we kept showing up to understand how permitting and construction works with utility-scale solar. In total, we logged around 2,500 miles in permitting and construction visits. There was a tangible moment when the lead engineering manager of the building department realized that we were absolutely serious about doing this in a matter of months. He paused after I asked him about a non-conventional idea that would accelerate the timeline and began collaborating about how we could go about the permit submission process. I could feel his sentiment move to our side of the table.

At the end of the day, the crushing bureaucracy that is so easily vilified by accelerationists like myself is made up of normal people. While regulators at the Public Works department raised their eyebrows at our idea to convert sunlight and air into fuel, they did not actively want to harm our prospects at radically innovating the primary energy consumption layer of society. Despite what you may have heard about California, it is still possible to build here! Pleasant persistence worked as we consistently called to ask questions and check in on submittals, left voicemails, and followed up with emails and texts wherever the bottleneck existed. We brought cards and chocolates at Christmas time for every person we interacted with during the process.

The land that we selected was flat, secluded, fenced, near a town, and zoned for heavy industrial use. Once we selected the land, we made an offer and submitted our conditional use permit as soon as the tract went into escrow. This allowed us to know if the project was dead-on-arrival before the diligence period closed. It also gave us a head-start on breaking ground once the land was officially ours. Conditional use permits are approved on a monthly basis at the Kern County Planning Commission. Our permit was delayed by a month because the only employee that manages it was out and the paperwork wasn’t filed. Nuances like this can make huge monetary differences in a project’s execution.

In the Kern County halls of power, receiving our official approval for land use permitting in December 2025.

Once we knew the project location, we contracted a leading solar builder to design and install our 1.8 MW solar array. This was the longest lead time scope on the project and Planted Solar executed impressively during a time of intense competition from datacenter builders who needed similar utility-scale electronics.

We split the project into multiple permits that were submitted in parallel. This allowed us to move faster and not have an issue with one piece of the project holding up the rest from advancing. We engaged three structural engineers and an electrical engineer for the project after speaking with well over a dozen. As inexperienced construction managers, our first drafts were a far cry from the polished drawings that a larger, slower company would produce. Our earnestness came through as we rapidly iterated with new information and consistently followed up to ensure our counterparties had everything they needed from us to advance the work. We had to replace one of the consulting engineers for not moving fast enough.

Initial site improvements, adding a fire road and compaction March 2026

In the construction phase, we started with basic site improvements like adding a rolling gate, repairing a fence, compacting a fire road, and creating a gravel turnaround area for deliveries. Once this was complete, we drilled and poured concrete piers to mount our equipment skids and began the solar array buildout. Despite a small delay in material delivery and a few minor installation hiccups, the solar deployment was done on time.

Planted Solar rapidly executing the buildout of our solar array, May 2026

We bootstrapped industry knowledge by speaking with multiple contractors for all the trades (concrete, electrical, fabrication) to get an understanding of market pricing and estimation process. Most of our contractors were extremely collaborative and did great work. A mistake we made was being too collaborative with an electrical contractor on the design of our system. It turns out their ideas weren’t very good and their execution wasn’t there. We pivoted quickly to a new contractor that was much better. A lesson in always having a backup and taking the time to form an opinion before engaging a contractor on project architecture.

Our most difficult failure on the project was obtaining a permit for our battery system. One of our systems was not yet adapted for HVDC and needed 3-phase AC power, pushing us into a much larger class of battery than we originally intended. This larger battery triggered an extremely rigorous permitting process with the Fire Department who had a sordid past with large-scale battery projects in the area. Despite meeting the National Electric Code and California Electric Code with our product choice, the permit was denied because it is within the county’s purview to demand a higher standard than the CEC. This forced a pivot in our design to supply AC power and a large sunk cost of time and capital. Lesson: There is non zero risk in placing a large capital order before the permit is in hand.

Were the systems for which we were building finalized in their design? No. Was there in-house expertise that could guide the process? No. Did we have a clear idea of what it would take and how much it would cost? Definitely not.

The design changed constantly and we became adept at understanding when we needed to revise permitted plans and when the design was made accommodating enough to keep moving with the existing paperwork. We standardized around a shipping container footprint for the skids (8’ x 20’) and overbuilt the structural steel to handle changes in the actual equipment that would be mounted. Sometimes the best thing you can do for engineers in a design phase is to implement constraints for them to build around, rather than trying to build to a moving target.

The numbers:

• 58 days: Closing to permits and ground breaking
• 10 working days: Solar installation
• 50 days in the field
• 2500 miles driven
• 2984 panels installed
• 1.8 MW nameplate power
• $1.3m total cost (including land, site improvements)
• $60k costs attributable to errors, first timer n00b fees, and high speed preference
• $0.016/W: amortized power cost – best in North America. Not bad!

What are the takeaways? You must have urgency, conviction, and a willingness to figure it out in-flight. Pleasant persistence and comfort with looking foolish can be credited. We went into this fearing that maybe what the veterans told us was true. That any new build solar in the US would take four years and cost $1.40/W. This turned out not to be the case, but we could never have known if we didn’t do it ourselves.

The most rewarding piece of the project is seeing our engineering teams put the site to use. Since completing the infrastructure, Terraform Industries has notched a series of additional world firsts, producing offgrid CO2, hydrogen, and methanol.

A brief update on the state of synthetic fuel in late 2026, T+11 months since the Singularity. First, commentary on the global State of Energy. Second, what we’ve delivered so far in 2026. Third, commentary on regulatory and permitting reform.

The power to destroy a thing is the absolute control over it

My mapping tool shows energy infrastructure damaged in the last five years. Conflict in Iran, Yemen, Saudi Arabia, Russia, Ukraine, Iraq and Syria has resulted in the effective deletion of much of about 10% of the world’s oil production and delivery capacity.

It has been obvious enough to military planners since the 1970s that the global oil supply chain was vulnerable to interruption by unconventional forces.

I founded Terraform Industries in late 2021, and by early 2022, Russia seemed determined to put this notion to the test by invading Ukraine. In the years since, Ukraine and Russia have systematically destroyed each other’s energy and oil refining infrastructure, placing enormous stresses on Eurasian oil supply chains, especially since the “developed West” spent the last 30-40 years voluntarily (!) demolishing its own oil refining capacity.

In 2026, this kicked into high gear with Iran’s denial of the Straits of Hormuz in addition to a simmering conflict in Yemen around the Red Sea.

We hope for a speedy resumption of peaceful trade and conventional international relations, but we’re not putting that genie back in the bottle. In 1944 it took the might of the Nazi war machine to launch a few hundred inaccurate autonomous drones a few hundred miles at London and Antwerp, and the might of US airplane production to deliver tens of thousands heavy bombers to respond by destroying Nazi coal liquefaction plants and fuel delivery logistical capacity.

Today, any non-state actor, proxy, covert/clandestine operative, cartel, or sufficiently motivated middle schooler can extract losses measured in billions of dollars with commercially available hardware that costs thousands of dollars. Kinetic deterrence still exists but is ineffective if you can’t deliver it, your adversary is protected, believes they can hide, or is willing to die.

If you don’t control your own energy supply chain, you are one third party misunderstanding away from rapid regression to the pre-industrial age.

Terraform Progress

Against this cheery backdrop, Terraform has continued to accelerate its efforts to provide cheap synthetic fuel from sunlight and air.

Muroc Desert Test Site

The headline of our 2026 efforts is the development of our Muroc desert test site.

I write this on the 79th anniversary of the first supersonic flight by Chuck Yeager in the Bell X-1. Back then, the nascent X planes experimental program needed a lonely, unpopulated place with good weather, and they settled on an airstrip then known as Muroc, now known as Edwards Air Force Base.

With similar needs, we found a test site in the same town of Rosamond, California, and named it Muroc in honor of those who went before, and those who favor us with a sonic boom at 9:36 every morning!

We built the cheapest solar ever deployed in North America

We closed on the land in January. By the end of March, all the solar panels (nearly 3000!) were in. A lot of cable work and inspections followed. By early July, the word came through from the solar team: we’re ready to go live. The site team, led by Matt Weickert and Tiana Wong, had successfully built a large solar array in California for $0.70/W in less than six months, and this includes its share of first time and small site expenses. With amortization over the coming years, we’re producing power today for 1.6c/kWh, which I’m quite confident is the cheapest power ever produced in North America. Terraform was founded on the supposition that falling prices of solar modules would ultimately trickle down to applications, enabling gigascale energy arbitrage turning cheap solar electrons into valuable energy molecules, and as of 2026, we’ve done it in the real world.

We produced pure CO2 from sunlight and air

Within a week, the CO2 team successfully produced a high purity (>99%) stream of CO2 from our demo calciner, at our Muroc desert test site, in 110 F heat, using our own captive solar PV array, on the first try.

This represents several world firsts, and is an incredible testament to the team’s brilliance and grit, led by Stephanie Coronel.

I directed each of the three power consuming teams (DAC CO2, electrolyzer H2, and reactor) to fab and deploy a meaningful end-to-end “smoke test”. I knew the desert would test our faith and that our team needed to experience the dust, the heat, and the wind in order to streamline logistics for the full deployment later this year.

DAC was first off the mark. I was fortunate to find a corner of the tent where I was out of the way as we turned on the calciner for the first time, on site. Without drama the array pumped about 30 kW into the test fixture and the temperature rose smoothly to over 1000 C. Soon after that, the team proceeded to <redacted> and CO2 came out. At the end of the test, we had calcined >85% of the carbonate we started with, exceeding our requirements and validating our design, architecture, and processes.

The Terraform CO2 production system captures CO2 from the air (400 ppm) and concentrates it to >99% purity in a single step, feeding directly into a hydrocarbon synthesis reactor which produces methane or methanol.

The overarching challenge of Terraform’s approach is to produce CO2 cheaply enough that it becomes a profit center rather than a cost center. Our strategy of direct consumption of solar power from the array cuts capex and opex to the barest minimum, and our internal production cost, even at this relatively early stage, is already well below $200/T-CO2.

Other than plants and some cyanobacteria, I believe this is the first ever commercially useful direct air capture and production of CO2 at scale.

We produced hydrogen with our own electrolyzers

A few weeks later the hydrogen electrolyzer team at Terraform pulled off yet another first, the sustained production of >99.9% pure H2 from our proprietary, vertically integrated, California-manufactured electrolyzer stack while it was coupled directly to a solar array at our Muroc desert test site.

Let’s break this down.

Most commercial electrolyzers need carefully conditioned power from expensive battery-meditated backup systems. Ours runs directly off the sun. Clouds pass, the day turns to night, and we maintain purity from a stack whose Bill of Materials cost is well below $100/kW.

Terraform’s single-minded focus on capex reduction has allowed us to convert sunlight to hydrogen in an unprecedented demo at a cost below $2/kg. If that wasn’t enough, we have a crystal clear plan of steady execution to push that cost below $1/kg in the coming years.

Rather than linger on this point in response to experts-with-spreadsheets who said this was not only beyond my team, it was forbidden by known laws of physics, let me tell you a bit about how we actually pulled this off.

We started building this test site in March. Once the panels and electrics were in place the CO2 team were the first to demonstrate production on site. Close behind them the electrolyzer team planned the logistics necessary to project substantial operational ability into a hostile test site in the middle of nowhere. It’s no good to find you’re missing a wrench half way through the day!

Much of the test prep was completed before dawn, when the panels go live. The sun came up and the stack immediately started splitting water into hydrogen and oxygen. The team carefully monitored purity and flammability as the sun climbed through the sky. As designed, the stack warmed up and conducted even more power, maintaining solid production until late afternoon when the setting sun shaded the panels.

Terraform’s synthetic fuel system is uniquely designed to follow the sun and extract the maximum possible value from cheap solar panels.

Hydrogen is a pernicious molecule. It leaks through and embrittles metals, burns almost invisibly at a wide range of mixtures in air, burns hot and fast and can easily undergo a detonation transition, and has about half a dozen other spookily dangerous properties. My advice is to never work with it unless you absolutely have to. The Terraformer produces and consumes H2 in one compact discrete area with a minimum of complexity and fuss, and as expected this demo was completed in accordance with our rigorous safety standards and no unscheduled excitement!

This successful demonstration was also a profound milestone for the team after a testing anomaly last December compelled us to finally rip off the bandaid and move decisively towards the “future design,” which has half the parts but considerable complexity in assembly. No-one else makes electrolyzers this way and we, more than anyone, know exactly why. And also how to do it anyway, translating directly into a unique cost advantage.

A huge congratulation to Ken, Sherman, Chris, Nikhil, Abdullah, and Aaron for their successful test campaign.

Terraform’s hydrogen and CO2 are the chemical precursors for synthetic methane and methanol, which we make in our own synthetic fuel reactor.

Combined, we make oil and gas out of sunlight and air. We are breaking the geological and geographical monopolies on oil production.

Terraform will deploy these electrolyzers by the millions and they will all be plug-and-play with solar PV arrays.

The Muroc smoke test campaign was far from over.

We will win!

The reactor team produces methanol at Muroc

In September, the reactor team at Terraform demonstrated successful production of high purity (food grade) methanol at scale at our Muroc desert test site.

Let’s recap our progress in 2026. In January, we closed on our Muroc test site. By March, construction was underway. By May, we had 1.8 MW of solar installed. In July, we produced CO2 on site, and in August, hydrogen.

Terraform Industries has built an incredible team to revolutionize not just one or two key technical pieces, but the entire stack required to convert sunlight and air into fuel. Like plants, only 100 times higher productivity per acre and requiring no irrigation, fertilizer, or pesticide.

Two years ago, Terraform produced pipeline-grade natural gas for the first time. Since then, we’ve scaled up our methane reactor, increasing production by 250x and reducing contaminants by 10x. A chemically pure grade commands a 100x higher price, increasing revenue and margin.

But it was always clear to me that natural gas was only one piece of the puzzle. It is possible to convert natural gas into other hydrocarbon products like gasoline or jet fuel, but given our precursors are CO2 and hydrogen, we can do one better by directly synthesizing methanol as well.

Methanol is far easier to convert into longer chain hydrocarbon products, either via the MeOH->DME->ethylene pathway, MTG, or variations on Fischer-Tropsch.

Methanol also has significantly higher revenue per molecule than pipeline natural gas, which forms an essential part of our cost-reduction-through-profitable-scale-up strategy, explained in more detail here.

TL;DR: If our learning rate is above (a very conservative) 11%, our margins will strictly increase as we scale up into a market worth about $10t/year.

Amir Ahdnoparvin joined the team earlier this year and speed ran the methanol reactor development program, successfully demonstrating recirculation and heat recycling before integrating these new technologies in a production scale reactor, built from the ground up at Terraform’s facility in Burbank.

As soon as the hydrogen team was done, we shipped our latest reactor up to the desert, plugged it in, and began activating the catalyst. An extended test campaign through the heat of summer followed, punctuated by bunnies nibbling on hoses!

The hard work paid off, with Terraform’s production of significant batches of high purity methanol, independently certified as meeting food grade and shipped to a happy customer. A huge congratulation to the entire team, including Amir, Enric, Dominic, Hugo, and Lucie.

With the completion of this test, Terraform has now demonstrated TRL 9 operation of all of its subcomponents, each developed from scratch, in-house, to deliver the highest possible value to the end customer.

It is hard to describe the visceral feeling of staggering beneath the weight of pure wealth flossed from the air by human ingenuity, grit, and toil. Terraform will never rest until we have broken the geological and geographical monopoly on oil, unlocking unconditional energy abundance for all of humanity.

But wait, there’s more! Chemically pure methane distribution

Not to be outdone, the methane reactor team produced thousands of dollars of product at chemically pure grade, sold a batch to Astro Mechanica, and hammered out a distribution deal with WestAir.

Ian Brooke, the founder of Astro Mechanica, is designing his supersonic electrically optimized jet engines to use methane as a fuel, since it has a higher specific energy than any other hydrocarbon and its cryogenic nature as a liquid will be helpful for supersonic applications.

A token contribution

Earlier this year, we published DCDCs, a roadmap for powering large scale AI data centers with off grid solar+battery power. A few months later we had a large solar array up and running in the desert and I thought … why not?

So we built and demonstrated an H100-scale system running autonomously off our own solar array. This is not just a cute demo.

Dashboard showing 3 days of operations testing configs and various open weight models.

As we demonstrated in our post, an economically optimal solar inference rig achieves roughly 99.9% utilization on the GPU (the expensive silicon) with about 20% utilization of the solar array (the cheap silicon) and battery pack.

Implication: Every solar+battery offgrid datacenter will have copious amounts of free power available to any offtaker who can accept less than 99.9% utilization. This could be a local utility district looking for additional power during peak AC-load summer months. In our case, it’s our colocated modular intermittency-friendly synthetic fuel plant, which we just spent five years lovingly engineering to match an unfirmed solar array.

Further implication: We intend to host modular inference clusters at our Terraformers, wherever zoning allows. These clusters consist of their own compute plus battery, and they share the solar array along with the fuel plant, doubling its economic utility and further reducing power costs for fuel synthesis.

We no longer have to wait until 2031 for the solar array installation market to shake out in order to push our power costs down below one cent per kWh. We don’t have to dump millions of our equity-funded engineering effort into accelerating a soon-to-be-commoditized layer in solar installation. We will maintain our core focus on scaling production of the highest quality synthetic fuels at the best possible price!

What comes next?

Here’s an interactive website showing our deployment plan, indicating new products and steadily increasing margins.

Interactive: the live scale-up roadmap. Scroll inside it to move through the stages, or open it at roadmap.terraformindustries.com. Open the interactive on the blog.

In summary, we’re production limited and have positive unit economics. First, max out the Muroc test site production, revenue, and margin. I’m optimistic the site will pay for itself next year – an audacious goal for any first-of-a-kind full scale pilot operation, but we’re more than equal to the challenge. Then scale up production and deployment of Terraformers, build out the fleet, achieve profitability, continue to push production cost down and production volume up. Wash, rinse, repeat.

Career Opportunities

Terraform takes talent development very seriously. Do you want to fast track your up-leveling? Work on real world problems with real stakes? Spend time around conspicuously ambitious people? Job opportunities are listed at terraformindustries.com.

Addendum: Reform Opportunities

The US risks losing the energy race if we don’t get out of our own way, especially on land-use permitting. Yes, the irony that environmental laws slow solar development is not lost on me.

We’ve seen enormously positive developments on this front, particularly in reigniting innovation in the US atomic power industry. The following suggestions are easy wins that unlock enormous wealth potential for Americans and future generations. We have focused this analysis on three opportunities in Southern California but similar issues and opportunities exist all over the US. At the end, we follow up with some pointed commentary on the proposed BAAJA Act.

Broadly, the American West is full of undeveloped land. Why? Not enough water, the federal government was unable to give it away at scale during the land grant era, even with extraordinary generosity to the railroads! The result: In 2026, the US is losing the solar deployment race 10:1 to China because of zoning issues. This is existential.

Opportunity 1: Kern County farmland zoning

Proposal: A written determination from the Kern County Planning Director that solar-powered synthetic methane production falls in the same agricultural-industry category as an ethanol plant, and so is a conditional use on farmland, with conditions set by its own effects: no combustion, no feedstock trucks, and equipment that can be removed.

Context: In Kern’s zoning ordinance, Exclusive Agriculture (A) is a working farmland district. Solar is permitted by right when its accessory to a permitted or conditionally permitted use and generates no more than the site consumes (§19.12.020). The same district allows ethanol plants and biomass energy conversion by conditional use (§19.12.030.A.2). Also, the ordinance includes a ministerial process, the similar-use determination, where a “potential applicant” can ask and the Planning Director rules within 45 days, appealable to the Board of Supervisors (§19.08.030 to 19.08.080).

Problem: Because solar-to-methane isn’t a listed use, every site becomes its own negotiation. Landowners have no clear answer when developers ask to lease farmland for solar-powered fuel production.

This is like the old joke about the Air Force (where everything not explicitly permitted is forbidden) and the Navy (where everything not explicitly forbidden is permitted). Solar-to-methane/methanol is an emerging technology and is explicitly permitted when intermediated by relatively inefficient, high-intensity corn-to-ethanol, and de facto forbidden when performed using advanced, low-intensity, highly efficient Terraform technology.

Solution: We ask the County to determine that solar-powered synthetic methane/methanol production belongs in the same Agricultural Industries category as “Ethanol plant” and “Biomass energy conversion” under §19.12.030, with conditions based on its own effects rather than theirs, and that co-located photovoltaic (PV) serving only on-site demand is accessory under §19.12.020. If granted, any A-zoned parcel in the county could host a plant with one Conditional Use Permit (CUP), with no zone change. Growers get a lease option for hard-to-farm land; roughly half of California’s Central Valley is arid because of a structural lack of irrigation water. The County allows a use with no combustion, no feedstock trucks, and removable infrastructure. In addition, the County answers the question once instead of a dozen times, reducing its administrative load and associated costs.

If Kern County prefers an amendment to the Zoning Ordinance, the Board of Supervisors can start a text amendment (§19.112.020). In that case, we ask the Board to initiate an amendment to the A-district conditional use list at §19.12.030.A.2, adding “Renewable synthetic fuel production” between “Oil extraction, non-mineral” and “Saw or planning mill,” and to amend the definition of accessory use at §19.04.018 so that a multi-parcel array on contiguous lots under common ownership or control counts as accessory to the principal use. Because every project would still need a conditional use permit, the County retains site-by-site review.

§19.12.030.A.2 (Agricultural Industries), verbatim from the February 16, 2026 Kern County Zoning Ordinance:

── Flour mill

── Glucose processing

── Oil extraction, non-mineral

── Solar fuel production (making fuels and chemicals from air and water using on-site solar energy)

── Saw or planning mill

── Soil amendment, not involving liquid chemicals or organic materials, including blending and distribution

§19.04.018 (Accessory use), with the insertion after “same lot”:

“Accessory use” means a use of land or of a building that is customarily incidental and subordinate to the principal use of the land or building located on the same lot, or on contiguous lots under common ownership or control. An accessory use may be established only after the principal use is established.

Opportunity 2: California farmland permitting

Proposal: Solar and solar-powered fuel production becomes a permitted use on agriculture-zoned land in California.

Context: For small rooftop systems, the state already sets one rule for every city and county (Government Code §65850.5). Farmland is different. Zoning is local, so each county and city decides for itself, on its own schedule, whether farmland can host solar and the equipment that turns it into fuel (§65850). According to the California Energy Commission (CEC), California also imports nearly 90 percent of its natural gas. Land that can host an in-state producer gives owners new lease income and supplies the state with gas it would otherwise import.

Currently, the CA Department of Conservation Farmland Mapping and Monitoring Program has mapped about 31 million acres of agricultural land. Of the 31 million private acres, only about 8 million acres are irrigated and 17 million suitable for grazing, leaving a fifth of the state’s ag-zoned land as not necessarily ag-appropriate. This land is essentially locked out of participation in the economy, even while Texas out-builds California in solar by 10 to 1.

Problem: Two factors set the timeline for a solar-fuel plant on farmland, and neither is predictable. First, because zoning in California is local, each county and city runs its own discretionary process, and timelines can vary drastically from one to the next. Second, though the Permit Streamlining Act sets maximum time limits for public agencies to approve or disapprove development projects, it doesn’t close the gap: its decision clock doesn’t start until environmental review is finished. In 2025, the Legislature exempted advanced manufacturing facilities from CEQA, but only on sites zoned exclusively for industrial use, so farmland is left out. A permit with no firm deadline burdens both landowners and developers because landowners can’t tell producers when, or whether, the land can be used, and developers can’t plan or finance a plant around an answer with no date.

Solution: We ask an Assembly member to carry a bill adding the section below, which would sit in the zoning chapter of the Government Code (Title 7, Division 1, Chapter 4), next to Section 65850.5, the state’s solar permitting rule for rooftop systems. Landowners and developers get a date to plan around, the state gets in-state gas production, and counties keep their existing objective standards for agricultural structures. The bill follows the approach the Legislature took in 2025 (SB 131) for advanced manufacturing on industrial land, which defines a project type and removes open-ended review.

SECTION 1. Section [____] is added to the Government Code, to read:

[____]. (a) Solar energy and solar fuel production, including solar arrays, capture of carbon dioxide from air, electrolysis, synthesis, storage and loading of products, and pipeline interconnection, are permitted uses on any parcel zoned for agricultural use. “Solar fuel” includes chemicals made in the same process.

(b) A city or county shall approve an application for a use described in subdivision (a) ministerially, applying only the objective standards that apply to barns and other agricultural structures in the same zone**, and no standard that applies only to solar energy or solar fuel production**.

(c) A city or county shall approve or deny a complete application within 60 days. An application not acted on within 60 days is approved. A denial shall identify the objective standard the application fails to meet, and a corrected resubmission shall be decided within 30 days.

(d) An application is deemed complete 30 days after submission unless the city or county identifies in writing the specific items missing. Items not identified then may not be requested later.

(e) A use described in subdivision (a) is a compatible use under Section 51238.1.

(f) This section applies to applications pending on its effective date, at the applicant’s election. An interim ordinance under Section 65858 may not prohibit or condition a use described in subdivision (a).

SEC. 2. An approval under Section 1 is a ministerial approval for the purposes of the California Environmental Quality Act.

Opportunity 3: Fuel production on Air Force land at Edwards

Proposal: The Air Force offers land at Edwards Air Force Base for domestic fuel production under the same lease authority it is using for AI data centers, starting with any of the data-center sites that go unleased.

Context: The Air Force is offering seven sites totaling 2,115 acres at Edwards for AI data centers through an enhanced use lease, the military’s general authority to lease base land to private parties (10 U.S.C. §2667). Solicitation AFCEC-26-R-0002 asks bidders with on-site generation to plan for up to 35 MW of backup power for the base during a grid outage, and in its notice of intent the Air Force said it may favorably consider energy resilience. Our Muroc test site is next door in Rosamond.

Problem: The lease is written for data centers. Bidders had to show at least three completed AI data center projects over 100 MW, so a fuel producer can’t compete for the land. Meanwhile the base’s fuel, like everyone else’s, comes from refineries and supply lines it doesn’t control, the same ones the first half of this post showed can be cut with commercially available hardware. And off base, every new fuel plant in California waits on county zoning and state environmental review (Opportunities 1 and 2).

Solution: We ask the Air Force Civil Engineer Center (AFCEC), which runs the lease, to reoffer any Edwards data-center site that isn’t leased as an enhanced use lease for energy and fuel production, and to accept proposals for fuel production on other suitable land at the base. A site of about 500 acres would host Muroc 2.0: a Terraform plant scaling to 100 MW of solar and beyond, making methane and methanol inside the fence. That is fuel the base can’t be cut off from, made from sunlight and air, and methanol is the starting point for synthetic jet fuel. The Air Force gets lease income or in-kind consideration, as the lease authority allows, and a working model of domestic fuel production for other installations. As with the data-center leases, the lessee pays for the environmental review.

BAAJA: Where permitting reform could shoot itself in the foot

S. 5653 (BAAJA) is the most serious federal permitting reform in decades, and we support its core: hard deadlines, page limits, state-review substitution, short limitation periods and categorical exclusions for disturbed land. Skimming all 417 pages, though, we found places where the text could hurt the projects it is meant to speed up, or hand opponents the case for repealing it. Most fixes below are a sentence, and we think they improve the ability of the bill to deliver its intended outcome and accelerate prosperity for all Americans.

Clocks that bite the applicant

• The 150-day window to sue can close before you know you were denied (§2201(b)–(c)). It runs from the date a decision is required to be posted, not the date it is posted or the applicant is told. Fix: run it from the later of public posting or written notice to the applicant.
• Deadlines plus personal penalties reward a fast “no” (§1106(g)(3); §2213). Missing a deadline costs senior staff half their week and limits agency heads’ travel, and BLM may “grant or deny” within 30 days. Denial is the safest way to hit a date. Fix: require a reasoned denial that can be cured without restarting the clock.
• “Incomplete” can be repeated forever (§1109; §2213(b)(1)). The clocks start at completeness, and nothing limits how many times an agency can ask for more. Fix: one completeness review, and the application is deemed complete 30 days after resubmission.
• A casual yes to “a revised schedule” waives the remedy (§1106(g)(2); §1302). Fix: consent must be written, name a new date and be single-use, and the new date is enforceable.
• A two-week shutdown erases deadlines rather than pausing them (§1106(g)(4)(C)(vi); §1403(c)(5)). Fix: extend each deadline by the length of the lapse.
• No transition rule. It isn’t clear whether applications pending at enactment get the new clocks. Fix: apply the Act to pending applications at the applicant’s election.

Fast lanes that turn out slower

• The fastest projects have the longest litigation tail (§2201(a)(3)(B); new NEPA §111). Our reading: the 150-day bars cover environmental documents, so projects approved under a categorical exclusion, or excluded from NEPA entirely, may default to the general 6-year limit. Fix: extend the bars to categorical-exclusion and no-document determinations.
• One extra permit kills the loan-only exclusion (§1102; §2301). It applies only where a federal loan is the “only” federal involvement. A single wash-crossing verification or right-of-way pulls the whole project back into NEPA and Section 106. Fix: a loan doesn’t by itself make an action federal, and review is limited to the other permit.
• Definitions narrower than real projects. “Commercially available solar photovoltaic systems” (§2213(c)(2)(D)) can exclude a novel system built from standard modules. “Previously disturbed” is frozen to a 2025 regulation in §2213 but floats in §2214. The categorical-exclusion anti-stacking rule (§1105) blocks phased growth of a small pilot. Fix: “systems using commercially available modules”; one statutory definition of disturbed land that includes former farm, mining, storage, industrial and military land, as well as land subjected to offroading and other recreational disturbances; and one follow-on exclusion of equal size unless the first phase has been found to cause a significant effect.
• Challenges to the definition of waters of the United States (WOTUS) get a 120-day window and then disappear (§1201(b)). Our reading: review under CWA §509(b) bars raising the issue later in an enforcement case. Fix: keep the appellate venue, but preserve as-applied defenses.

Remedies and powers that backfire

• Uncapped damages from the Judgment Fund (§1402(d)(2); §1401(b)(4)(D)). Awards of 50–100% of construction cost, for any energy type, with no appropriation vote, invite permit-farming. This is the provision most likely to sink the bill or draw a constitutional challenge. The winner must also repay everything if the project changes (§1402(d)(3)), which no startup can promise. Fix: cap damages at documented costs, limit them to bona fide projects (site control and documented spending on the project), and make the injunction and a court-ordered schedule the main remedy.
• Protection lost over any violation of law (§1401(b)(1)(B)). “Materially” qualifies the breach of the permit but not “otherwise violated applicable law”, so a late report becomes grounds for revocation. Fix: “a material violation of applicable law directly related to the authorized activity”.
• Litigation moves to the laws that still allow injunctions (new NEPA §111(c)(5)–(6)). The bill makes remand without vacatur the default under NEPA: a flawed review gets fixed while the project keeps building. Opponents will move to the ESA, the Clean Water Act and Section 106, so litigation moves rather than shrinks. Fix: carry the same discipline to those laws. To halt construction, a challenger must prove with site-specific evidence that the project is likely, before the court can rule on the merits, to cause harm that can’t be undone afterwards and that meets an existing statutory bar: jeopardy to the continued existence of a listed species, or destruction or adverse modification of designated critical habitat (the ESA §7(a)(2) tests, which look at the species as a whole, not individual animals); destruction of the features that qualify a property for the National Register; or fill of more than half an acre of waters covered by an approved jurisdictional determination (the limit in the main nationwide permits). Each resource must have been designated before the application was filed. The applicant never has to prove a species, site or wetland is absent, and nothing designated after filing can stop the project.
• An unreviewable CEQ override of ESA and Section 106 consultation (NEPA §107(g)(1)(C)(iv)). Today it speeds projects up, but the power is symmetric: a future administration can use the same unreviewable override to stall any industry. Fix: limit the override to enforcing the bill’s deadlines. CEQ can force a consultation to finish on time, but can’t decide its substance either way.

Doors left open to the veto

• Old documents lock out new technology (§1106; §1107). Agencies may ignore research published after the notice of intent, and may rely on a programmatic document for 10 years. Our reading: a programmatic review written around 2026 technology could be used in 2036 to screen out designs that didn’t exist when it was written, or to push them into a full review. The research cutoff also cuts the other way: an agency can ignore newer data showing a design is cleaner or cheaper. Fix: define what a programmatic document covers by footprint and effects (land area, water use, emissions), not by technology, so any design inside the analyzed envelope qualifies. Newer information may be used to show a design falls within the envelope. Require agencies to refresh programmatic documents, and the technology assumptions behind loan and grant criteria, at least every 5 years, at their own expense and without new surveys from applicants.
• Commenting is still the ticket to sue (new NEPA §111(b); §1106). The bill lets only parties who filed a substantive comment and can show direct harm bring a claim, and caps scoping comment at 45–60 days. That is the right direction, but tying standing to a comment keeps the comment period as the gate to litigation. Every opponent learns to file a “substantive comment” on every project to keep the lawsuit option open, and agencies spend months answering them to protect the record. It also catches applicants, who rarely comment on their own draft environmental impact statement and so can lose the right to challenge their own permit. Our reading: NEPA’s own text requires that environmental documents be made public, not that the public be invited to comment or that objections be answered. Fix: replace comment periods with public notice. Agencies owe no response to unsolicited comments. Standing goes only to the applicant and to parties with a concrete, direct injury to their own property or contract rights from the specific action, and filing a comment neither creates nor preserves it.
• Tribes’ new permitting power can be taken back on an outsider’s petition (§2212). A right-of-way is the right to build across land: a road, pipeline, transmission line or solar array. Today every right-of-way on tribal trust land needs approval from the Bureau of Indian Affairs, which is slow. §2212 lets a tribe with an approved regulation grant rights-of-way on its own land “for any purpose” with no BIA approval, which makes willing tribes a real siting option. But on a petition from any “interested party” the Secretary may rescind the tribe’s regulation and take the job back, the bill doesn’t say whether grants already made survive, and the United States isn’t liable for the losses. Lenders will price that in. Fix: only the tribe, or a party with a direct injury to its own property, may petition, and only on the ground that the tribe isn’t following its own approved regulation. Rights-of-way granted before a rescission remain valid for their term.

Rules that are clear, bounded and hard to game survive changes in who holds power. Keep the procedural core, close the gaps that let a deadline, a definition or a petition reopen the veto, and trim the extreme remedies and unreviewable powers that make the bill a target.

How does new technology enter the market?

New technology costs more. Adoption at scale requires two strategies: A beachhead market willing to pay enough to make early production profitable, and economies of scale that drive down costs as adoption increases, with cost falling at least as fast as the beachhead is exhausted. 

Think of Tesla’s strategy around electric cars. First, the Roadster, an overpriced toy to mature the tech at a smaller and cheaper scale. Then the Model S and X, premium cars that addressed a much larger market. Finally, the 3 and Y for the mass market. This pattern even continues with the piloting of cybercab tech through the cybertruck, e.g. the 48 V bus and large aluminum castings.

What are economies of scale?

Bigger things are cheaper, everyone knows this! But hang on a second. Foxconn makes phones cheap by making millions of them. TSMC makes chips cheap by making millions of them. Hanwha Ocean makes container ships cheap by making them enormous. Wärtsilä makes marine Diesels cheap by making them enormous. 

Is it strictly correct to label both of these economies of scale? One gigantic thing vs trillions of tiny things? I think this is fine, provided one understands what is actually delivering improved value in this operation. The world does not need one gigantic chip or a trillion tiny container ships. What delivers value is the repetition and ultimately automation of numerous atomic processes. For ships and chips the square-cube law also comes into play – I will leave the unification of this observation as an exercise for the reader.

I will note that scale doesn’t automatically make things cheaper. There are also profligacies of scale, with examples too numerous and obvious to even mention. 

Why does repetition make things cheap? What is the learning rate? 

The learning rate, also called Wright’s Law, is a phenomenological description of a real world process. Manufacturing engineers optimizing US production of warplanes in WW2 noticed that cost dropped by a fixed percentage, typically about 15%, per cumulative doubling of production. In the context of WW2 plants churning out thousands of airplanes, this became a powerful driver in US over-production of war materiel, hastening the end of the war. 

This curve, for B-29 production, shows a late time acceleration of the learning rate, which has been observed in a few other industries, typically when scale of production finally breaks through whatever the previous limiting factor was.

For example, solar photovoltaic module production turned a corner in 2009 when scale finally justified dedicated silicon processes rather than using leftovers from chip manufacturing. Since then, the solar learning curve has continued to steepen, reaching 48% last year! Currently, global production doubles roughly every 24 months, so the price of modules falls by nearly a factor of two in that time. 

In contrast, the lithium ion battery learning curve is a more modest 23%, similar to Boeing’s B-29 plant. As a result the price halves roughly every three years, even though manufacturing is doubling at a faster pace than solar, doubling every year for most of the years since 2018. 

Sea water reverse osmosis learning rate (15%). 

Why is it that learning rates vary so much? 

To first order, the more complicated the product, the lower the learning rate. Solar modules contain no moving parts, relatively few materials, have a fungible supply chain, an efficient market for production tooling, and so there’s a broad attack surface for process engineers to suck out cost and increase production rate. SWRO, in contrast, involves lots of parts, lots of moving parts, corrosive, fouling chemistry, high pressures, and rather finicky membranes. Still, 15% is far better than zero, and the most advanced plants churn out fresh water for just 40c per cubic meter!

So what does Terraform’s learning rate need to be?

There’s an aspect of Terraform Industries’ market expansion that I always thought was extremely obvious, but recently an investor pushed back on it so I actually did the math. It turns out not to be super obvious, but after this post it will be, so strap in!

Question

Okay, fine, your beach head market is premium chemically pure methane with higher revenue, but what does your learning rate have to be to ensure that your cost of goods sold (COGS) drops fast enough that as you expand you can still sell profitably, even as your beachhead premiums go away? That is, we believe you can produce chemically pure methane and sell it profitably, but that market is worth only a few hundred million dollars per year globally, and that’s just not enough to justify venture returns. The global oil and gas industry is worth over eight trillion dollars annually, but your current cost is nowhere near competitive, so what’s the plan? Economies of scale?

Answer

The oil and gas market is big. You just won’t believe how vastly, hugely, mind-bogglingly big it is. I mean, you may think, say, Google, earns a lot of money, but that’s just peanuts to oil and gas.

You might spend 15 minutes reading this post. In that time, the oil and gas industry turned over $250m. 

The way to think about this is to consider the size of the incremental market expansion bought by a trivial cost improvement, say 1%. Short run oil demand is highly inelastic, but even so, a 1% reduction in price will increase newly addressable demand by about 0.08%, or $7b/year. $7b on top of Apple’s iPhone sales would be a 12% sales bump – enormous! In the oil industry, that’s Tuesday evening.

Short run elasticity isn’t the right measure, though. Below are two charts demonstrating methane’s long run addressable market elasticity is 2.6, while methanol, the oil precursor, is 5.0. That means that when Terraform decreases at-scale production cost by 1%, the methane market expands by 2.6% ($41b), while the methanol market expands by 5% ($115b). It is difficult to imagine a more inviting addressable market!

Because we have 24 doublings of production before saturation, even a 10% learning rate cuts cost faster than the beachhead premium decays. We never sell at a loss on the way down. Indeed, if we can compound our production skill with a learning rate of more than 12%, our margins will strictly increase as we scale out. 24 doublings is a lot of runway to push our production costs down by the factor of four required to unconditionally undercut drilling in any market, no questions asked.

That’s the intuition, let’s make it rigorous.

Here is our current estimate for the global methane market ladder. Currently we can produce methane for under $30/MCF – an achievement in itself! The red curve shows a 10% learning rate. 10% is nothing spectacular – it’s less than half the improvement seen in B-29 production, during war time, in the 1940s, on a plane whose size and complexity was so unprecedented that the flight test program took the lives of the test pilot Edmund T. Allen and 31 other people. 

This curve also assumes that the bulk Henry Hub market remains artificially depressed by natural gas co-production during fracking for oil products. Pure play gas fracking projects typically turn a profit at about $9/MCF, which would necessitate a learning rate of only about 6%, which is even more ludicrously unambitious. To be blunt, we have designed the Terraformer from Day 1 to be scalable in production and as simple as possible. We will exceed a 10% learning rate!

Natural gas is about a fifth of the oil and gas market. The remainder can be addressed with methanol and its derivatives, so we’ve accelerated our progress on methanol production and expect to begin direct-to-consumer sales soon! The methanol ladder is more complex, showing higher purity premium product tiers (green) and methanol-to-X downstream products such as gasoline, aviation fuel, plastics, etc (purple). A mere 6% learning rate easily clears this ladder retaining at least 25% margins. A 20% learning rate will result in accelerating margins and substantial additional induced demand and economic growth.

Of course, if we doubled production every six months for a decade it is possible that we could temporarily outrun the solar learning rate before becoming its primary driver, but this would be an excellent outcome for everyone. 

The Hydrogen Ladder. Every market Terraform methane or methanol can sell into, ranked by what it pays per MWh, against Terraform’s falling synthesis cost. Hover a step or a table row for detail; switch the horizontal axis between annual revenue and hydrogen deployed. Open the interactive on the blog.

Further implications of these facts?

While mass scale deployment of Terraformers is capital intensive, it is readily financeable with quantifiable risk and solid returns above 20% IRR, and at times far higher. You can explore various configurations at terraform-simulator.com. The Amazon fulfillment center build out might be a good parallel for this process. The business will expand as quickly as possible while retaining positive EBITDA less re-investment in growth. 

Once we get the ball rolling, expansion and strategy is straightforward. Just try not to be drowned by the nearly infinite money gusher.

What is the smallest possible catalytic spark? What is the key to unlocking growth? 

Terraform is in the process of deploying our first ever full scale Terraformer at our Muroc test site in Rosamond, California. We expect this development unit to have positive unit economics. The pilot will actually generate value, no $1b leap of faith required. This is critically important to avoiding the so-called “Valley of Death”. From there, we scale up production by bringing up new sites. Each site generates net revenue. We expect company level profitability to occur with between 15 and 25 Terraformers operational and will size our second development site well in excess of this. 

Join us!

Terraform is hiring. Check out open roles at terraformindustries.com.

Lucie Nurdin, PhD. Lucie is the Reactor Lead Engineer at Terraform Industries where she leads the team building solar-powered reactors that turn atmospheric carbon dioxide and green hydrogen into synthetic methane and methanol. She holds a PhD in inorganic chemistry from the University of Calgary and climbs alpine routes in her non-work time.

Terraform has opportunities for MechEs, EEs, ChemEs, technicians, and others!

It is 6:30 PM on a Friday in Burbank and my week is over. My inbox is empty, and next week is already planned. So is the week after that. So is three months from now, and six months from now. Same as every Friday.

For two decades, I’ve run my life on what I describe, without irony, as a game of chess against myself. I already know the endgame. I’m always two moves ahead of my opponent: the next meeting, the next pivot, the next hurdle.

We are living through a productivity crisis. The tools designed to organize us, Slack, Notion, Asana, Jira, have become the thing we need to be organized against, and the prevailing advice is to do less: quiet quitting, slow productivity, the four-day week. My approach is the opposite.

I lead the reactor team at Terraform Industries, a clean-energy startup building solar-powered Terraformers that capture atmospheric carbon dioxide and turn it into synthetic methane and methanol. I manage engineers across multiple projects, run hiring interviews, handle customer sales, and own logistics. I also do my own IC work: thermodynamics and kinetics of our reactors, design doc reviews, test campaign planning. Outside of work I climb in remote alpine places that require federal permits booked six months out and enough technical skill to stay alive. Ideally, a little more than the bare minimum!

I don’t feel behind. If everything is planned, it’s one less thing to worry about. Being on top of everything is what lets me juggle chemistry, engineering, project management, sales, hiring, and operations without any single one of them dropping. 

For me, organization isn’t “a professional skill” that I switch on at work but rather just how I operate, and I apply the same chess logic to everything outside work too.

The same system, applied to climbing

Most of the best outdoor climbing in California requires permits from recreation.gov, which open either six months out or two weeks out. Checkmate is a permit for the exact weekend I want. Two moves ahead means phone reminders set for the moment each permit window opens, and if the six-month window doesn’t hit, the two-week reminder is already queued. The prep work, which permits cover which area, which regions are open based on weather, how snowpack shifts the plan, what gear I need, runs in the background. I update the calendar every time an input changes.

Identify the goal, reverse-engineer the dependencies, automate the reminders, stay flexible when the inputs change. I use the same system at work.

Close everything on the spot

The single most important habit in my system is what I do in the last 5-10 minutes of every meeting. I don’t leave a meeting planning to deal with it later. I close everything before I move on.

“Who is doing what by when?” Admiral Rickover.

Action items get written down the moment they’re decided. Calendar reminders get set for anything date-specific. Stakeholders get emailed or tagged immediately. If it was a hiring meeting, notes go to the hiring committee right then. Action items get pasted into next week’s meeting notes so future-me already has the agenda.

I do not defer these tasks. Deferring is where organization goes to die. The ten minutes I spend closing out a meeting saves me hours of “wait, what were we supposed to do about that?” later in the week.

Plan the next meeting before you leave this one

Every recurring meeting (weekly, monthly, whatever the cadence) has its own Google Sheet with a template for the agenda, shared with the relevant stakeholder. Take my weekly 1:1s. I already know what I’ll talk to each direct report about, because I know what they were working on the week before and we ended the last meeting on the next steps. Those next steps are already pasted into Monday’s notes.

So at 6:30 PM on Friday, I know exactly what I’m telling my team on Monday morning. I know how it ladders up to my Q2 priorities, and I know what checkmate looks like: by the end of Q2, the methane reactor is built and shipped to the field test site, the methanol reactor is scaled up, I have a signed purchase agreement with another customer, and I’ve also successfully onboarded two new hires. 

Own your calendar

A lot of people hand their calendar over to Calendly or other platforms. I don’t. When I need to schedule with someone, I offer three specific time slots. In 95% of cases one works, and if none do, I ask for their availability and set it up myself. It takes an extra 30 seconds, but being in charge of my calendar is non-negotiable as the calendar is a critical part of the system.

I also know myself. I know how long things actually take, and I budget accordingly. A 30-minute meeting gets a 5-10 minute buffer after it for personal reset and note cleanup. A 1-hour meeting gets 15-20. People might look at the buffers and see wasted time, but they’re actually the whole reason my system works.

Triage everything, aka Eisenhower’s method

Every task that lands on my plate runs through the same filter. Important and urgent, I do now. Important but not urgent, I schedule. Not important but urgent, I delegate. Not important and not urgent, I delete.

Most people only use the first bucket and let everything else pile up. The delete bucket is the one people forget exists, and it’s the one that protects your bandwidth the most. I ask myself what’s actually important for me to do today, this week, this month, this year, and I reset hard when I catch myself slipping into task-mode with no purpose. Everything should ladder up to the big picture.

What a day actually looks like

I don’t do time-blocking or themed days. Because I’m organized at the root, I can context-switch without friction, and honestly I like it that way. Here’s a realistic day.

8:30 AM, 15-minute stand-up, then 5 minutes to write out the day’s key action items. 8:50, 40 minutes on big-ticket items: review a design doc, send five important emails to get things moving, maybe a call or two. 9:30, hiring interview. Template ready. I log in three minutes early, skim the 30-second candidate memo, run the 30-minute call. After: write notes, decide next step. If it’s a yes, send notes to whoever owns the next stage, set a reminder to follow up, schedule the email to the candidate. Close everything.

10:00 AM, block for reading a design review. 10:15, someone needs something right now. Fine. I jot a quick note on the design doc so I know where I stopped, drop my first round of comments, and go deal with it. If the interruption can’t be resolved in 25 minutes, it goes on the to-do list and gets processed properly: root cause, then is it a people, time, or money problem, then prioritize, then action plan, then action items.

11:00 AM, empty slot on the calendar. Back to the design doc. Close it out, write my key points, follow up with whoever owns it, make sure it moves forward. 12:00 PM, 1:1 with my direct report. Template ready, five minutes beforehand to add anything I’ve been thinking about. 12:30, meeting done, five minutes closing action items and prepping the next one. 

2:00 PM, check on the team: are we blocked, is there a decision I can make right now to let us move forward faster, safer, cheaper. 

5:00 PM, scan the rest of the week and tomorrow. Anything I need to prepare for? Everything is under control. 6:30 PM, I bike home feeling like the system is working. 

Automate your future self’s memory

Calendar reminders are mandatory. If a vendor says I’ll get the tracking number on Friday, I drop a reminder to check that Friday. Sometimes I add a second one a few days before: a nudge to email the vendor and confirm we’re still on track. If I don’t hear back, I pick up the phone. Often I pick up the phone first.

Sometimes I go further and draft the follow-up email in advance. Vendor said the PO would be ready Friday? I write the Wednesday-morning email now that sounds like this: “Hey, you mentioned the PO and tracking would be available by Friday. It’s Wednesday, can you confirm we’re still on track?” and schedule it to send itself. Never assume someone else will remember to do the thing. You’ll never lose by being too organized.

Decide fast, communicate clearly

I make fast, decisive calls. I can’t afford to dwell. If I don’t have the authority, I go straight to the source, usually our CEO, Casey Handmer. If I do, I make the call and communicate the reasoning behind it. I welcome feedback, but I rarely reverse a decision. I go with it.

I probably make 20 decisions a day. Some are huge: what price point do we sell methane at, which nearly decides the fate of the company. Some are small: do we need to order another gas analyzer for $1,000? Both get the same treatment. Decide, communicate, move. If you dwell on small decisions, you run out of bandwidth for the big ones.

Why this reduces stress instead of adding to it

People sometimes assume all this structure must be exhausting. It’s the opposite. Because I’m on top of everything, my brain is free to hold the things that actually matter like the big strategic questions, the engineering problems, the people dynamics. The admin, the note-wrapping, the scheduling, the reading queue, all of it is on autopilot, handled in small closeout windows, never piled up.

If I weren’t in charge of my time and genuinely ready to give 100% to whatever needs my attention, I couldn’t be this proactive. 

The cost

I’ll be honest about the one real cost. I’m not built for last-minute chaos. If something blows up with zero warning, like a truly last-second change of plans, I have to force myself to adapt. I can do it, but I don’t enjoy it. Nothing is worse, to me, than a plan that doesn’t get executed. A system this tight makes you phenomenal at anticipated variance and slightly worse at pure improv. I’ve decided it’s a good trade, but your mileage may vary.

For me, a few minutes after every meeting and a few seconds to schedule a reminder is worth it. Compare it to the alternative: hours trying to keep up, constantly feeling behind, reacting instead of acting. If you always feel like you’re reacting, you’re not playing chess. Or you are, but your opponent is the one two steps ahead.

My System and the takeaway

No Jira. No Notion. No fancy productivity stack. Google Calendar, Gmail with scheduled send, Google Meet with AI note-taking and transcripts, and Google Sheets. That’s it. The simplest system is usually the best one, because simplicity is what you actually maintain. 

Everything I’ve described breaks down into single, granular steps. That’s the point. Good organization isn’t about adding complexity. It’s about adding structure, clarity, and a strong sense of ownership to things you’re already doing. Close out every meeting. Own your calendar. Schedule the reminder. Triage the task. Make the decision. Move. Build the discipline to run this system.

I’ll update this in six months with changes I’ve made by then.

For the past half year, I have been building hardware to remove carbon dioxide from ambient air at the lowest cost ever. This post is the first part of a series that tells the story of how I got here, the fundamentals of direct air capture (DAC), and my approach to solving this problem.

But before we begin:

The Castle is where I reside. If you are interested in fighting humanity’s war against carbon dioxide or if you simply love building hardware, please apply at Terraform Industries or reach out directly.

My Origin Story

For a brief moment in middle school, I get infected by the “global warming is going to end us soon” mind virus. It comes from consuming the following cocktail of ideas in a science class: fossil fuels are finite, burning them is bad for nature, and disaster for mankind is coming soon. I am still fuzzy about what the end of the world would look like, but I am scared-straight after watching the movie “2012”. My ills are cured after observing my parents’ nonchalance about this impending catastrophe and reading the geoengineering section of my science textbook. We might just be fine.

Years later, a different virus descends upon us. COVID never catches me, but it traps me inside. I devote part of my screen time to a MOOC about carbon capture and post Reddit-style insights in our chemistry class’s group chat on interesting lessons. I am glazed for my active participation during a parent-teacher meeting, and this feels rewarding. It is difficult to pinpoint causality, but these experiences play some role in my decision to pursue engineering and focus on climate tech in university.

In college, I start by building electric racecars, which is the ultimate technical learning experience. I think making the future of batteries is my North Star. However, three things coalesce and discourage me from going deeper into climate tech: I take a soul-sucking public policy class on energy, an internship at Rivian convinces me that all the low-hanging fruit in the EV industry have been picked, and I get attracted to other shiny problems.

Fast forward to a month after graduation, where I read my onboarding document at Terraform Industries (TI) and realize that I am taking over the Carbonator’s design, a reactor for direct air capture of CO₂. This feels like a crude full-circle moment. For reasons I can’t fully explain, I picture the bubbles in a Coke bottle when I read about the Carbonator. The carbon dioxide in carbonated products turns out to be a potential product line, but before we get to that part of the story, we must go back to the basics.

Fundamentals of Direct Air Capture

A covalent double bond connects the carbon and oxygen atoms. The distance between their nuclei is short, and their similar size enables good electron orbital overlap. These factors combine to make a strong bond that is difficult to break. Life seems hard for the DAC engineer who has to find this fully oxidized and inert molecule that is not abundant in air. Luckily, there is a chink in the armor of carbon dioxide.

Oxygen is super-duper electronegative, and it pulls the electron cloud in the double bond toward itself. This gives the molecule some local polarity, as the carbon is slightly positive and the oxygen is slightly negative. The electron-poor carbon atom can be attacked with a base—a substance that likes to give away its electrons. This acid-base reaction is the most common way that carbon dioxide is captured from air. It happens even under ambient conditions because the products are stabler than the reactants, and it feels like a game about finding ways for molecules to climb down the energy landscape.

For all the talk about rising CO₂ levels, there is only ~0.04% of it in ambient air, which makes DAC akin to finding a needle in the haystack. Physically feasible solutions have been around for close to a century, but the current moonshot for a carbon capture device is going from $1000 / ton to $100 / ton of CO₂ captured.

TI’s plan to achieve this is based on “calcium looping,” and the sketch below shows the cycle that allows us to reuse the calcium-based sorbent to gobble up the CO₂. The hydration step converts quicklime (calcium oxide) to slaked lime (calcium hydroxide), and a flaking machine converts a slurry of this into flakes that remind you of Tony the Tiger. The carbonation step converts these flakes into limestone (calcium carbonate) by blasting a lot of humid air through them. Finally, a very-very-very hot kiln decomposes this carbonate to release CO₂ and calcium oxide, which is fed back into the input of the hydrator to complete the loop.

I have focused on the carbonation step, which is thermodynamically feasible at ambient conditions but kinetically infeasible without water. The water acts as the battlefront for the reaction, and it comes from three sources: the humidity in the air, the initial moisture of the flakes, and as a byproduct of the reaction. The initial moisture of the flakes is what holds them together for long enough to prevent disintegration into powder. It also provides the starting burst of carbonation that leads to a shell forming around the flake—think soft and moist flakes turning into crispy frosted flakes. The reaction produces heat, but large amounts of humid air keep the flakes cool and wet, allowing more of the flake to carbonate.

The water is critical because polar H₂O molecules partially break apart the ionic structure of calcium hydroxide, which dissociates into calcium and hydroxide ions. When the CO₂ molecule finds its way to the battlefront, it dissolves and gets attacked by the hydroxide ions, which break the double bond and form a bicarbonate ion, and another hydroxide ion strips the proton off this bicarbonate to form a carbonate ion. Now we have calcium ions and carbonate ions floating around in an aqueous solution that react almost instantaneously and precipitate into solid calcium carbonate.

The reaction itself is really fast; the hard part is getting the CO₂ to reach the aqueous solution of calcium ions. The flake is initially composed of calcium hydroxide and some water, but a shell of calcium carbonate grows on this porous flake until most of the hydroxide is trapped inside. As more carbonate forms, the unreacted calcium hydroxide core shrinks, making it harder for the CO₂ to find the calcium hydroxide, and thus carbonation slows down.

AK’s law of carbonation states that to get a lot of needles, one needs a lot of haystacks. So we use powerful fans that sound like little jet engines to blow 5,000 liters per second of air (you just breathed half a liter of air in the second it took to read the line). Despite getting blasted with all this air, it takes a day for half the flakes to turn into calcium carbonate. This is because the diffusion resistance scales with the amount of carbonate in the flake. Like the diminishing pleasure of guzzling each successive pint of cookies and cream, the returns from blowing more air on heavily carbonated flakes drop off sharply. When the rate has slowed below a cutoff, the Carbonator is ready to dump these flakes into the calcination kiln.

You can do all the chemical analysis in the world to figure out your carbonation conversion, but the Rick Rubin way to FEEL the carbonation progress is to release a handful of flakes and hear the sound of pennies falling into a jar.

The Design Philosophy

To make money and achieve the orders of magnitude reduction in cost, we must churn out reactors at the TERRA scale (technically mega scale, but I couldn’t resist). Instead of performance maxxing like a race car engineer, I now design to minimize the time and money spent building the reactor. “SIMPLIFY DELETE REPEAT” is the mantra of the day, the night, and the weekend, and now it has spread to other parts of my life. As if hearing this mantra in every stand-up was not enough, we were handed “Delete” stickers to put on parts that needed elimination.

An ambitious startup must move as fast as LeBron on a fast break. Prototypes must be churned out rapidly and tested against reality. There are many allotropes of these textbook “easier said than done” ideas, such as Elon’s algorithm or the Skunkwork’s manual and they are not new. However, reminders are useful, and their impact on generations of businesses, from Ford’s factory lines to contemporary SaaS startups, illustrates the power of these methods.

I guess the repetition works because, like calories in Bryan Johnson’s diet, all my parts must fight to survive. I have only ever heard tales of custom CNC-machined components. They never have a happy ending! This forces me to think real hard about making things the simplest they need to be to ensure the survival of my designs. These precautions are essential to stay safe from the hardware gremlins because each added component exposes more surface area for them to attack.

It has taken a few hardware failures to really digest some of these lessons, and I have a long way to go. The nexus of my current growth is determining the optimal number of tokens to devote to thinking about a problem before building prototypes. Only reps will tell if this is an intuition I can develop.

These “reps” are the next part of my story.

This post explains our current views on how humanity will achieve Kardashev Level 1 status by exploiting the full energy resources of an entire planet. More specifically, how pure solar+batteries will power AI scaleup beyond gas turbine manufacturing limits.

[Edit: Thank you for the nice write up, Ana!]

It is an extension to my earlier post of March 2024 on using solar to power AI datacenters, and a response of sorts to the Scale Microgrids paper that showed a mix of solar and gas could reduce emissions for the developers and operators of next gen AI datacenters. In that paper, Kyle Baranko, Duncan Campbell and co-authors showed that around 90% solar with local natural gas backup generators would be the fastest way to get power. In this work, we show that taking this trend to its obvious conclusion and deleting all the legacy fuel-based power components can be even faster and cheaper. We also include a discussion of space-based inference. 

Let’s examine this problem from first principles. What is silicon cognition?

You can call it a tensor core, a Blackwell, a GPU, but these are all versions of the same thing. A sliver of silicon with billions of transistors, through which cascade a torrent of electrons converting the entropy of a few volts to the entropy of information generation, and the entropy of waste heat. A GPU is a very complicated switch that regulates current flow, with some other side effects.

For the foreseeable future, the GPU will be the expensive part, currently valued at around $50,000/kW. All it needs to continue to operate is an infinite supply of moderately spicy electrons, that is, a DC power supply at a few volts. Given that making power is much simpler than thinking, the job of the power supply is to be uncomplicated and relatively cheap. In no universe should providing power be the hard part.

Solar and batteries are a natural match to this demand. A solar panel is a slice of silicon (without logic gates) that absorbs solar photons and drives electrons uphill. To a good approximation, a solar module is a constant current source that maxes out at about 40 V. A battery is a reversible chemical reaction that stores and releases electrons, and to a good approximation is a constant voltage source. Modern lithium chemistries hold at about 3.9 V across nearly their entire state-of-charge range. 

For logistical reasons related to the relative scarcity of copper in the crust of the Earth, it makes sense to operate solar cells, batteries, and GPUs in series so that the entire system runs at about 1000 V and each electron can be reused a few hundred times. 

Our radical claim is that, in the limit, Earth-based AI compute will look like this: 

By area, thousands of acres of solar panels. 

By cost, a pile of GPUs. 

In the limit, Earth-based AI compute will be a direct current (DC) solar array connected to a DC battery bank connected to a DC GPU rack. 

This approach brings numerous other advantages: 

• No grid connection. 
• No moving parts. 
• No turbines. 
• No gas connection. 
• No nuclear fuel. 
• No emissions. 
• No power conversion. 
• No transformers. 
• No inverters. 
• No power transmission. 

None of these parts make the AI smarter, and all of them can potentially intrude onto the critical path. 

Delete.

It sounds nice in theory, but how can this work?

The key metric to optimize is tokens per dollar. For example, take Scale Microgrids’ work on a 90-10 AI system, increase the size of the solar and battery farm enough to get to 99+% uptime, delete the gas power side, and compare overall economic productivity. A gas system that’s used only 1% or 0.1% of the time still costs time and money, and that’s the core reason why deleting it can end up reducing overall cost. 

The graph below shows the tokens per dollar landscape for two hypothetical solar powered AI systems, one with a gas powerplant (blue) and one without (orange). Both have a solar array (size given in nameplate multiples of the peak AI load) and battery size (given in hours of capacity at full load). 

The key insight is that there are two stable attractors. One with a pure gas energy supply, with solar and battery supplementation for vibes, CO2 reduction, or marginal capacity expansion. The other with pure solar and batteries, no gas. The pure gas system capex is minimized with no solar and batteries, as natural gas itself is relatively cheap given no preference for emissions reduction. But the two manifolds intersect along a frontier, and beyond that the solar array and battery are capable enough that it’s actually cheaper to delete the gas powerplant entirely.

This tradeoff does not come at zero cost. In exchange for deleting the cost, complexity, and schedule risk of a gas powerplant comes the sizable land demands of a solar array. To a rough approximation, 15 acres of solar are required per MW of DC AI load. For reference, the USA has about 150 million acres of unpopulated desert west of the Mississippi, enough for 10 TW of AI development. 10 TW is much more than total global electricity generation today. There is plenty.

On the other hand, while fracked gas is relatively abundant (for now) the turbines that convert it into power are hard to make, hard to ramp, and largely already spoken for. If AI seeks growth beyond the production ramp of turbines, it is clear which way the wind is blowing. 

Before we get to the methods section, I’ll give a rough heuristic for performance. Assuming an on-off binary state on the load, a 15 MW solar + 15 MWh battery can get to ~99% utilization anywhere in the US south west, but is that good enough? The short answer is yes – maximizing tokens per dollar spent, or ROI, justifies throttling demand on a few of the longest, coldest nights of the year. 

But it’s actually better than that. Remember that a GPU is a glorified silicon switch intermediating the flow of electrons downhill. Power consumption is proportional to clock frequency multiplied by the square of the voltage (P ~ f V²). GPU power consumption is not fundamental: Token production rate is. If we’ve deleted DC-DC converters then voltage is set by the state of the battery, and frequency is controlled by software. This means that a 3% reduction in token production rate can buy us a 9% reduction in power consumption. So the math changes from 99% utilization to more like 99.7%. This shifts the economics around solar and battery plant sizing considerably, given that GPU frequency modulation allows for a 3x discount in actual utilization and token production.

There is one other implication of these wildly capable and versatile solar+battery AI data centers. They have enough power to operate at full, or nearly full, capacity for the entire year. For 10 months of the year they are oversupplied, and can provide electricity and low grade heat (from their cooling systems) to neighboring customers essentially for the marginal cost of power transport. These could be seasonal or intermittently friendly loads such as the synthetic hydrocarbons and primary materials being pioneered at Terraform, and/or local communities. At Terraform, we believe that power should be as cheap as possible. 

Methods

Epistemology. How is it possible for this lightly evolved monkey to know these things?

You will need:

A year (at least) of real time solar data from a target location. This is data for an EW fixed tilt array in Texas that we generated by feeding fixed south tilt data into a slightly non-trivial geometry model.

A solar PV module IV curve model. This is based on the JAM72S30-540/MR/1500V but they’re all pretty similar.

A Li-ion discharge voltage curve. 

A frequency-power curve for a typical GPU.

Plug these all together in a model that charges the battery when the sun is up, provided the panel voltage is high enough. We initially simulated a system with no power electronics whatsoever, but found that battery charging efficiency was inhibited when the battery state of charge was low, because pulling the panels to a lower voltage actually decreased their efficiency. Given that MPPTs are not that expensive, we could put them back in.

Then, provided the battery can deliver power, the GPUs are powered and we count how many tokens are generated.

Throw in a basic “governor” that throttles the GPU when it predicts the battery will be exhausted before dawn. 

This graph shows performance over a ten day period in winter. Note how the governor throttles output early on the fourth day by rationing power until the following morning. The cubic power consumption of GPUs means that throttling a little bit early is much better for token production than running full blast into a wall and then dropping to zero production until the sun comes back up.

Now run thousands of simulations for every combination of battery size and array size, measuring overall utilization of the load. 

Interactive: a simulated year for every array and battery size; change the prices to see when gas stops paying. Open the interactive on the blog.

This chart shows yearly utilization of a GPU asset given solar and battery sizes, including our basic governor. Note that a “steepest” ascent starting at zero solar and zero batteries turns first on adequate solar, then adequate batteries, then marginal solar, then marginal batteries. This reflects the shape of the resource curve and the degree of exploitation required to get to the marginal nth 9 of reliability.

This chart shows curtailment reduction with adoption of the minimum viable governor vs some naive on/off operator, showing 2.3-2.6x improvement, which is close to the 3x implied by the GPU’s cubic power consumption. This governor is not very sophisticated, for example, it has no ability to take weather prediction into account. It merely assesses the time of day, the state of the battery and of solar generation and curtails GPU utilization accordingly. 

Interactive: every design’s solar utilization, GPU utilization and profit. Set the compute revenue and costs. Open the interactive on the blog.

Throwing in assumptions about capex, we can assess capital efficiency.

This chart shows token production per dollar (in arbitrary units), showing a rather broad peak with considerable flexibility. Adding too much solar or batteries degrades capital efficiency – the correct response is to add more GPUs in this case. 

Because the peak is so broad, there is freedom to choose for one additional preference. That is, we can alter the size of the array and the battery by 20-30% with respect to the load and still get much the same return on capital. Given that land is finite, we may want to maximize tokens per acre while holding development cost constant, which puts us towards the lower edge of the peak in the diagram above. Then, holding land use and tokens per dollar equal, adding more battery towards the bottom right of the peak increases absolute token production on a fixed GPU and solar array asset base. This mirrors actual operational optimization, which is to say, pave all available land with solar, then add GPUs and batteries until revenue peaks. 

At last, the machinery to perform a comparison with a gas or gas-solar hybrid system is in place. Plug in some assumptions around GPU cost, solar cost, battery cost, gas turbine cost, gas fuel cost, and amortization period, and you can produce this chart.

Here we assume that GPUs are $50,000/kW, batteries (including all ancillary power electronics) are $200/kWh, solar is $200/kW, gas turbines are $2500/kW, gas is $55/MWh, and we’re amortizing over 10 years. 

The chart suggests the possibility that under some set of assumptions, it’s actually cheaper to delete the gas power system entirely, so what are those assumptions? For any given cost, select the peak utility point for solar array and battery size and marginalize across these parameters.

The left side of this chart shows where a pure solar system is the best value. As a rough rule of thumb, this is where 1.3 x battery cost + solar cost < $500/kW(h). 

As a sanity check, in early 2026 we’re seeing large scale integrated battery storage systems ship from China for well below $150/kWh, while the cheapest industrial scale solar systems are going in for under $200/kW-DC-nameplate. This is beyond the critical cost threshold – delete the gas system.

Let’s recap.

It is possible and even optimal to run a datacenter on pure solar and battery. The optimum level of availability is between 99% and 99.9% utilization, with the balance taking the form, primarily, of throttled use rather than lights out. 

The pure solar+battery data center is cheaper than a gas-assisted or pure gas data center if solar and battery costs per kW are <~10% of the turbine/gas generator cost. For example, if a 1 GW gas turbine costs $2.5b and the solar array costs $250m/GW and the BESS costs $250m/GWh, then the utility is roughly at parity. 

We estimated the opportunity cost of a month of delay on a 1 GW gas turbine being delivered at about $20m, which probably isn’t high enough to justify deleting the component out of pure suspicion, provided that you are confident it will be delivered eventually. 

On the other hand, we found that it is possible to commission a pure solar+battery data center with high utility and then backfill additional GPU capacity if/when a gas turbine becomes available. 

There is broad latitude for design flexibility within the peak utility (tokens/$) range. For example, there are points with equal utility that have 20% less solar or significantly more availability, depending on secondary constraints such as land availability.

Given that, long term, trends will favor pure solar+batteries and the performance relative to complexity is already favorable, there is an argument that one hyperscaler should probably move aggressively in that direction, so as to obtain differentiation. 

We investigated what the performance hit would be for a pure solar+battery DC power system that deleted non-computational silicon, that is, inverters, converters, and MPPTs between the solar and battery components, and even in the racks themselves. We found that the power system performance relative to cost improved with the deletion of AC and DC-DC conversion components between the arrays and batteries. Again, this is a nod to the future we will converge on.

Space AI

Late 2025 saw much speculation about space-based AI. It seems to me that SpaceX, with their incumbent advantages in launch and Starlink hardware expertise, may be able to ship gigawatts of inference compute into Earth orbit for something like 2x the per token cost of ground-based AI, but that this would still be quite profitable. Why bother? It’s a separate delivery and distribution channel that isn’t congested by the usual permitting and regulatory nonsense at play on the surface, or at least, a different and uniform set. And if you have unlimited launch upmass it helps to have a profitable use case, like Starlink or orbital AI, to soak up that supply. 

Let’s list their respective advantages and disadvantages.

Space AI

Positives:

• Simplified and unified regulatory regime.
• High altitude (800 km – 2400 km)  dawn dusk sun synchronous orbits are never in shade (except momentarily during rare lunar eclipses) so don’t need batteries.
• Don’t need to be cheaper than ground AI, as long as they’re cash flow positive.
• Infinite source of marginal launch demand can fill in gaps from other customers for a very large rocket.
• Passive stabilization with dihedral solar arrays is possible.

Negatives:

• Launch cost is not free.
• No maintenance capability.
• Higher latency (50 ms worst case).
• Radiation can accelerate GPU degradation.
• Thermal environment requires relatively large radiators and (probably) active cooling.

Ground solar AI

Positives:

• Mostly a land development play, plenty of unused land on the ground.
• Batteries are relatively cheap compared to GPUs and enable operation overnight, getting cheaper all the time. 
• Almost certainly cheaper than launch and space rating of components.
• Don’t require a million tonnes a year to low Earth orbit to deploy.
• Easier to maintain/retrofit.
• Lower latency (closer to end users).
• Default human Earth surface environment is less hostile to hardware.
• Can easily cool using HVAC and air. 

Negatives:

• Regulatory/permitting is painful, byzantine, and locally variable.

Fundamentally this is a bet that GPUs are so valuable on a per gram basis that even launching them to space helps improve the economic utility of a Watt of solar power. 

Finally, an application of space-based solar power that can justify something like the vision of Gerry O’Neill. To be clear, this is because the value of a watt of space-transmitted microwave power encoding an intelligent token of data is about a trillion times higher than the value of a watt of space-transmitted microwave power competing with your local power plant to supply the grid.

Conclusion

Whether in space or on the ground, turbines are irrelevant to reaching Kardashev Level 1. The fastest growing AI will win, and the fastest growing AI must delete all non-essential parts. The only essential parts are a solar array, a battery, and the GPU itself. 

Aside: Carl Sagan’s extension of Kardashev Levels from integers to reals

Planetary power: 10¹⁶ W -> K = 1
Stellar power: 10²⁶ W -> K = 2
Galactic power: 10³⁶ W -> K = 3

One step on the Kardashev scale is equivalent to increasing power consumption by a factor of 10 billion. 

What is the current level of humanity?

Global electricity production: 3.5 TW -> K = 0.65.
Global fuel consumption: 20 TW -> K = 0.73.
Global planetary surface use for agriculture: 13% -> K = 0.91.

If all of Earth’s land was paved with solar PV at 26% efficiency -> K = 1.01. 
If the entire Earth including oceans was paved with solar PV at 26% efficiency -> K = 1.06.
If we fill the unshaded dawn/dusk sun synchronous orbital (SSO) band (800 km to 2500 km) with SpaceX AI satellites, 10^17 W available -> K = 1.04.
All of Earth plus the SSO orbital band -> K = 1.085.

Convert the entire Moon into 2 kg/m^2 solar inference at Earth’s solar orbital radius -> K = 1.91.
Convert Mercury into a Dyson sphere (7 kg/m^2 density) @ 26% PV efficiency -> K = 1.9998.

No way to get to K2 without a slightly more efficient solar panel!

One of the unexpected privileges of founding an ambitious, future-oriented hardware company is that I get to hire, mentor, and develop ambitious people to become the leaders of tomorrow. Earlier this year, I formalized this process and today I’m externalizing this memo. 

Why? To better prepare candidates for the challenges ahead, to share our insights into organization, and to strengthen the culture of American re-industrialization. There is no supply of hardened, seasoned veterans of synthetic fuel production, yet. We get to build them from scratch. 

Join Us

Terraform is the best place for ambitious, hard working people to become the best they can be.

Terraform has a range of positions for early, mid, and late career mechanical, electrical, and chemical engineers and technicians, and allied trades. Incredible opportunities for growth, leadership, and worthy challenges in carbon capture, hydrogen electrolysis, methane and methanol synthesis, as well as nascent efforts on metals, cement, and desalination. 

To conquer the primary energy consumption layer of our civilization, check out our open roles and apply at hiring@terraformondustries.com. 

More Context

Here are some other useful posts on career development. I like the LLM-induced trend of putting only relevant references up top.

Career Development Guide For Job Seekers

Maximizing Resume Signal Noise Ratio

The Well Rounded Engineer

Stuff You Should Have Learned By Now

You Should Be Working On Hardware

Mr Beast Memo

The Company As A Machine For Doing Stuff (Packy McCormick)

Terraform Development Responsibilities

Engineers build tools to perform otherwise impossible tasks. A tool can be a hand, a screwdriver, a line of code, a fixture, a jig, an injection mold, a factory, or an org chart. The tool is the mechanism by which an idea is made real.

Career development is meta-tooling. Terraform as a whole, and your team and responsibilities within it, are intended to be the best place on Earth for you to become the best you can be. 

As Terraform continues to grow, we need to ensure that job expectations continue to be explicit. The following levels provide a map for your development and the intensification of capability. They are intended to function as a cursus honorem, allowing any employee to easily see how to make the most of their time and opportunities here. 

The job ladder provides visibility into both technical and managerial tracks, but at Terraform, we strongly encourage people to develop these mindsets in parallel. Pay increments are roughly 10% per level, and we expect people to climb a level in no more than a year. This is explicitly “up or out.” Stagnation is not an option. 

The following is not an exhaustive list of expectations. The job titles here map to broader industry titles, and not Terraform job titles, which we prefer to make more descriptive. The increments in capability and authority are intended as a guide for career development conversations at Terraform. 

Leadership

Terraform provides early opportunities to take on leadership responsibilities and develop skills. Leadership skills are unintuitive, enormously valuable, and may be readily improved through practice. All leaders must be able to:

• Build the team: Recruit, mentor, hold accountable, and drive.
• Set the direction: Lead, motivate, explain, consult, and build consensus.
• Don’t run out of money: Sell, advocate, and understand the needs of the customer.

Some organizations prefer to hire specialists to offload some of these skills, such as sales or recruiting. People in these organizations assume that their job requires them to only handle their specialty, and their skill atrophy in other areas. This is deeply career limiting. Even if your company has a sales org, you still need to understand how sales works, because it’s how you attract top candidates, get credit for your work, drive consensus around your ideas, and secure resources.

At Terraform, we generally don’t hire non-technical specialists, because our most critical interfaces are with other technical people, and we don’t want to dis-intermediate that information exchange. Instead, everyone gets to learn about traditionally “non technical” functions, greatly expanding the scope of their competence.

Each of the levels below requires aggressive development of all of the skills associated with leadership to progress, in addition to the steady in-sourcing of responsibilities and growth of scope described. 

Levels

Table stakes at entry level

1. Integrating with the existing team. It is non-trivial, requiring a high degree of technical competence, insight, and self-awareness. Must also have sufficient ambition and determination. Remember, action precedes motivation. Enjoyment comes after sweat.
2. Can successfully complete basic tasks given detailed instructions.
3. Can successfully complete basic tasks with basic instructions.
4. Can successfully complete more complex tasks with detailed instructions.
5. Can successfully complete more complex tasks with basic instructions.
6. Can autonomously complete complex tasks.

First promotion

7. Can oversee small teams. This includes providing coaching and instructions to other less experienced team members, and taking responsibility (and giving credit) for the successful completion of their deliverables. It also includes knowing how to communicate information people need to build stuff. 
8. Can specify tools to solve problems. “We need a widget that can do X.” “We need Salesforce/CRM.”
9. Can make tools to solve problems. “I found this design for a widget, I’ll make it.”

Engineer II, Senior Lead

10. Can design tools to solve problems. “I designed this new widget to solve a problem.” Note that reusing old solutions is often better than designing new solutions, and knowing when innovation for its own sake is not necessary is a key judgment skill needed at this level. On the management side, responsibility grows to include outcomes, not just component delivery. That is, the senior lead delivers not just the component, but the solution the component is meant to provide. This requires the ability to see through the text and through reality to the solution space. It is important to cultivate the intuition necessary to judge what is necessary, or what is a viable path. Constraints often force an even better design.

Lead Engineer, Team Lead/Manager

11. Can oversee sub-systems. This is a more product-focused role than small team or component management, because here the person has visibility into their team AND interfaces with other teams, and has to intuitively understand what their customers need, not just what they say they need. At this level, the lead has responsibility not just for a component but for a product. The role includes product management aspects, that is, responsible not just for the success of the product in terms of function but also in terms of intended use cases and interfacing with the customer. This is also the level where outcomes are consequentialist. At level 11, the Lead guarantees an outcome and they are not graded on effort. A team lead/manager also has considerable responsibility for talent management, that is to say, recruiting, hiring, performance management, and career development. The team lead is expected to remain the supreme technical authority within their vertical and have the ability to operate authoritatively in adjacent verticals. 

Senior Engineer, Senior Manager

12. Mastery of a topic area, such as a particular manufacturing stack, design language, analysis tool set, regulatory framework, enterprise sales, etc. Eg tool & die, thermo/fluids, statics, etc. This level is table stakes to spend a career working on really interesting stuff. The manager has a high level of confidence in successful execution of any project, and has typically demonstrated a track record of success. At this level, the manager also begins to manage managers, a critical but very challenging transition and one that enables the operation of much larger projects.

Staff Engineer, Associate Director

13. Company authority on a particular area. The manager remains deeply technical but interfaces primarily with managers, and has developed competence in correctly intuiting issues and decisively addressing them, so as to maintain a high degree of leverage in keeping projects of large scope on track. They have comprehensive knowledge of every aspect of development they supervise, play an active role in developing junior talent, and make swift and correct decisions. 

Senior Staff Engineer, Director

14. An industry authority on a particular area. In addition to the above, they have an established track record and likely a public profile. They have developed the skills necessary to move into any project set and get things right. They assume total responsibility for successful execution. 

Principal Engineer, Senior Director

15. World authority on a particular area. “There are 10 people who might know more about this than I do, and we’re all in the same group chat.” This may also be thought of as a transitional role into an executive-style role. 

Reading List

Fortunately, we don’t have to do this in a vacuum. Millions have gone before us. Thousands have written their stories. Dozens are comprehensible and useful. Speaking of which, you should definitely write a blog. I know of no better way to increase your luck stat.

“If I have seen further it is by standing on the shoulders of giants.” Isaac Newton 1675 to Robert Hooke, a predecessor of rather short stature.

In our quest to maximize our rate of skill development, we need to aggressively consume hard won lessons by others. You should develop your own programs of reading or other meaningful media consumption to level up. LLMs can help here by listing and ranking dozens of primary sources according to your preferences. I have found that first person technical accounts are about 100x more interesting and useful than researched histories by non-technical historians, in terms of actionable knowledge for real world leadership of technical programs.

The purpose of reading is to challenge your assumptions and biases, and improve your internal world model and intuition so that you can make correct decisions more quickly and more often.

Safety

NTSB aviation accident investigation database
NTSB YouTube
The Killing Zone: How and Why Pilots Die (Paul A Craig)
Into Thin Air (John Krakauer)
Touching the Void (Joe Simpson) (also a film)
Normal Accidents (Charles Perrow)
Aeronautical decision making
Checklists
Columbia Shuttle Disaster report
Challenger Shuttle Accident report
Normalization of Deviance
Chemical Safety YouTube
178 seconds to live (controlled flight into terrain)
Steam leak horror stories (reddit)
Therac-25 software safety escape
Flying Blind (Peter Robison)
Dittemore’s Law (Casey blog)
Mark 14 Torpedo problems
TWA Flight 800
Hyatt Regency Walkway Collapse
2021 Champlain Towers Condo Collapse Report Critique

Leadership

Poor Charlies’ Almanack (Charlie Munger). Founders podcast summary
Pieces of the Action (Vannevar Bush)
More Than My Share of it All (Kelly Johnson)
Now It Can Be Told (Leslie Groves)
Rickover Effect (Theodore Rockwell)
Zero to One (Peter Thiel)
Hard Thing About Hard Things (Ben Horowitz)
Founders (podcast)
Founders: How Elon works
Founders: How Jensen works
Acquired (podcast)
– VISA
– Costco
– Enron 
– Amazon
– Walmart
– Google
Cost of glory (podcast)
Stuff you have to know (Casey blog)
Maximizing resume SNR (Casey blog)
Well rounded engineer (Casey blog)
Patio11 on COVID vaccines (Patrick McKenzie blog)
How NASA Builds Teams (Charles Pellerin)
Only The Paranoid Survive (Andy Grove)
High Output Management (Andy Grove)
Mythical Man-Month (Fred Brooks)
The Elephant in the Brain (Simler and Hanson)
LOTR (also the films)
Thunder Below (Eugene Fluckey)
The Lean Startup (Eric Ries)
Extreme Ownership (Willink and Babin)
Mr Beast onboarding memo
The Sunion
Skunk Works (Ben Rich)
Play Nice But Win (Michael Dell)
Healing Back Pain (Sarno)
On Responsibility Essay (John Grier Hibben)
Lessons of History (Will Durant)
How the world really works (Vaclav Smil)

Physics and Science

The Art of Doing Science and Engineering (Richard Hamming)
Halliday Resnick and Krane (Physics Text)
Casey’s Physics Olympiad notes
Problem Solving Algorithm (Casey blog)
Structures: Or Why Things Don’t Fall Down (J. E. Gordon)
AvE YouTube
Munro Live YouTube
Ultralearning (Scott Young)
Why We Sleep (Matthew Walker)
Feynman Lectures on Physics
Alchemy of Air (Thomas Hager)
The Pattern on the Stone (Danny Hillis)

Are there any truly terrific books I’m missing? Send your suggestions!

Journal entry from January 2021: “Something bad is going to happen this year. I just hope I don’t die.” 

And the adventure begins…

Steel pipe is an unforgiving seat, especially two hours in. The back of the truck is perforated with a few grated windows that chop the passing world into meaningless slivers of light. It’s hard to tell where we are. The guy next to me needs to pee—a delicate situation when all twenty of you are handcuffed together. A bottle was procured, assistance rendered. Thankfully his needs were not of a more substantial nature. 

We have to be close. 

At about three hours we pass through metal gates. Big metal gates. Like a castle. But this isn’t a castle. I’m relaxed. 

This is all a grand misunderstanding. 

Although my confidence dips at the sight of multiple six meter fences. 

My friend is going to pick me up. 

Then logic erodes hope: The sun is going down. Must be 6ish. Government workers leave work at 5. These are the night-shift workers. 

I’m spending the night here. Fuck. 

What do they care that I’m innocent? As they say: In jail everyone’s innocent. 

We are ushered out of the truck and taken to processing. First, to plastic chairs where inmates administer less than pleasant haircuts. 

“Strip off your clothes and do ten squats!” 

I don’t actually know what they are saying—it’s Malay—but I get the gist. One guy with a bad leg gets kicked. Then they hand us colored clothing. I’m given purple, which means foreigner. 

“Foreigner.” He pronounces that surprisingly well. 

I look around but don’t see any other purple.

We are given a bowl, a cup, and soap. They organize us in rows on the asphalt and we sit. Criss-cross apple sauce. A position I would learn to loath over the next couple weeks. After a while the ground feels like hydraulic pistons digging into your body. We pledge allegiance to the prison system’s governance. After that, it’s time to go inside.

The night is dark, but artificial light follows us. We walk bent over. At each doorway we kneel and wait for further instruction. Through the final door they hit each of us on the head with a wooden block, for good measure. I’d been feeling pretty bad for myself up to this point, but I barely contained a laugh as a thought occurred to me…

All the guards are in Nike tracksuits. Is this prison sponsored by Nike? What a turn of events. I used to make their packaging…

Part 1: Origins

Leaving home

My freshman year at Cal poly was full of great memories. But I was anxious about taking out $80K to finish. I was probably going to make a pretty mediocre civil engineer and I desperately wanted to be more than mediocre. I wanted adventure. I wanted respect. 

Dustin Hoffman once said something on Inside the Actors Studio that stuck with me. He needed to yell in a scene and couldn’t get it. Take after take. Finally in a fit of frustration at himself, because of his own incompetence, he yelled, and that was it. He fulfilled the director’s vision without a deep connection with the character’s plight, and to the viewer it didn’t matter. It still felt authentic. That resonated: Frustration at my inadequacy propelling me forward just as genuine interest would. Who can tell the difference?

I went to the study-abroad office and asked for the cheapest, most obscure place I could go. In hindsight it’s funny they recommended Taiwan but at the time my reaction was “WHERE??” 

I enrolled. On my last day of freshman year, saying good-bye, I shook a friend’s hand and he said “See you in 6 months.” “I’m not coming back.” That was the first time I articulated this, even to myself. Even if I did come back, I would not be the same. 

Now I’m going to fast-forward through 18 months of language study and four years of university—love, heartbreak, ups and downs—because I want to get to something useful. Writing this down is my attempt at a small contribution to all the internet content that has made my life better over the years. 

So here is the super-expedited version of that period. 

Start:

Finish:

Now on to manufacturing. 

I’ll start this story from when I graduated from National Taiwan University with an econ degree and an inflated view of myself, then ate shit over and over for the next six years.

Mr. M: The teacher and The Test

Mr. M’s spirit haunts his facilities. You think you are safe, talking trash about his lunatic timelines with a colleague in the back corner, and then bang, he’s right behind you, footsteps light and quick like a cat. The floor workers had a special whistle, a little bird call that meant the devil is coming. I think he loved it.

Unlike the line workers, I got to talk with Mr. M 172 times a day. In hindsight, not a great way to run a business, or maintain the vestiges of my sanity, but I also can’t discount its effectiveness. 

He’d start the conversation with a simple, infuriating question.

Saturday at 10 p.m. “Where are you?” 

Sunday at 7 a.m. “Where are you?” 

Sunday at 11 p.m. “Where are you?”

… “I’m at home in my bed…”

90% of the conversations were less than a minute. I would answer the phone and he’d already be talking. “3 things, 1……2……. 2a……..2b…….2c…….2c1……2c2……..3….. That’s it.” click.

To me, Mr. M was a mentor, father figure, boss, tormentor, all wrapped in one. He cut an unassuming appearance, but his energy was uncontainable. The center of any room, among people of any stature. Typical uniform: Sneakers, light athletic pants, an outdoorsy collared shirt and a hat. Late fifties. Jet-black hair, never dyed. Young, intelligent eyes. Photographic memory. One of the first few hundred signators of Taiwan’s Democratic Progressive Party, which was still outlawed when he initially joined. 

Old school, too. Rarely said “good job.” Occasionally he’d say how thankful he was for something. That emotional moment lasted 10 seconds. Then: “I wish I was as lucky as you. You know you should be paying me for this. You’re getting an MBA for free.” 

In many ways he was right. I did get an MBA in those years. It didn’t come free though. I paid for it. Norm Macdonald had the best response to the “what doesn’t kill you makes you stronger” line.

“Now, metaphorically that’s true.
But literally it’s not true at all.
Literally it’s that what does not kill you, makes you weaker! 
And will probably kill you the next time it shows up!”

Condensed story of how I met Mr. M

• Friend meets Mr. M’s daughter on tinder date. As most modern stories begin. 
• Introduced as a ‘white math tutor.’ The irony didn’t fall flat on people.
• Pitched for a chance at a ‘real education.’ An internship in a Chinese factory town named ‘The Ancient Pit.’ With a name like that, how could I refuse?
• After 3 months, quit abruptly. Internship fell short of A Knights Tale fantasy—no reading poetry to princesses or jousting.
• Flew back to Taiwan to apologize in person. Conscience wins out, the treacherous bastard. Get ignored by Mr. M. 
• Back in Hong Kong—broke, shame mounting. Mr. M calls with another offer. “My friend is looking for someone to help him in China.” Accepted within hours. 
• Prepped with Infernal Affairs. Maybe you know this one, the one The Departed ripped off, with double agents. Mr. M tells me to watch it before going to work at the other company. But my need for money trumps any reservations I might have. 
• Completed ‘drinking bootcamp,’ preparation for drunk negotiations with Tier-1 clients.  
• Entertain many clients. Among them, a Portland-based sportswear company.    
• Landed at SFO to a text from Mr. M. “Welcome back to the US. Stay in touch.”
• Installed to run a Milpitas design office. Slept on a cot in the office for 6 months. Purgatory with slightly better food.
• Resigned. Scouted factories in Mexico. Flew to China to finalize a contract. Then got catfished.

Malaysia by accident (or the plan all along)

Mr. M: “Go to my Malaysia factory first. Charles just left. I have no one there. Maximum one week, then we’ll go open something in Mexico.”

Two weeks later: “Just stay there a few weeks while I find someone.”

A month later: “What can you possibly learn in a month? You need to spend at least three months here. Then you’ll be ready for Mexico.”

Three months later: “What’s the difference between 3 months and 3 years? You’re still in your twenties!”

I arrived in Malaysia on January 1st. 2019. As with my previous jobs, I was representing a Taiwanese company. 

Disclaimer: Some of the anecdotes don’t flatter us, and I won’t claim innocence, but there are two sides to any story. When buyers, QC (quality control) personnel, and mid-and-upper level managers (employ-ees) threaten an employ-er whose life is wrapped up in this business, it gets personal fast. Standard disclaimers apply: Do I seem like a reliable narrator to you?

Fly on the Wall: Damned if you do, damned if you don’t

One of the easiest attack vectors is through quality control. I didn’t realize this at first, but it became obvious quickly, and it happened many times. 

We were in the QC office of a major OEM (original equipment manufacturer, the entity that does final assembly). They had locked out 70 pallets (worth $10K) and were threatening a $100K fine if we didn’t ship 70 more pallets by tomorrow. We didn’t have inventory.

The head of QC theatrically squeezed our part and dropped it on a ruler lying on a desk.

“You see? It is out of spec. The dimensions are off.”

Mr. M: “My apologies. It won’t happen again. How about dinner tonight? Do you play Mahjong?”

That night we went for dinner at a nice Japanese restaurant. Multiple bottles of Sake were opened. Our guest got very drunk. Mr. M stayed sober. He’d brought us along to drink for him.

“Oh I can’t drink because…”

He always had the best excuses. More importantly, he could stay sober in a room of drunks and remain infinitely engaging. Anyway, Mr. M lost miserably that night at Mahjong. I’d never seen him play so poorly.

It was so bad it was brilliant, like getting every question wrong on a true/false test. He let the guy know ‘I’ve got you’ while still making sure to lose a decent sum. 

Sure enough the next day we were personally escorted to their warehouse. On the spot, each ‘rejected’ label was swapped for an ‘OK’.

A year later, this QCs appetite had grown. He knew the market rate; competitors were paying. Thankfully, I was there the day Mr. M had had enough. It was equally spectacular. Same office, this time with the head of quality, procurement, a few people from the end customer and the CEO. Mr. M took the QC’s phone, threw it into a wall, flipped a desk over, and walked out. I followed. Shocked, but also in admiration. 

Part 2: The chess board

Cat and mouse… and the elephant 

The cat-and-mouse game between customer and supplier overlooks an important player: The OEM.

The end customer issues an AVL (approved vendor list) and insists on cost breakdown from the OEM. OEM pricing is expected to match prices quoted directly by suppliers. For example: suppliers A, B, and C report $1 each to the customer. The customer then tells the OEM: Your price should be $3 + cost of assembly & logistics + 3%. Something like this. 

OEMs across Malaysia, Vietnam and China are competing for the same projects. They must team up with suppliers to bid, but ultimately they’re the price makers. That power is often overlooked by those not close to it. Foxconn employed half a million people at its Shenzhen campus alone in the early 2010s. Right now in Zhengzhou there are two hundred thousand people working on the iPhone. Two hundred thousand people!

The customer wants cheap, fast, perfect—and the right to change orders without eating inventory. Owners want profit without front-loading millions for a line that might vanish to another country. That tension births some creative behavior.

The Big Vacuum Cleaner Company 

This particular vacuum cleaner company had already chosen a designated supplier in Malaysia, but the order was big, and multi-year. Packaging alone would be US$100K per month, enough to start a factory. That means $10 million a month revenue for the entire product, since packaging usually makes up 1-3%. 

The game was afoot. As the designated supplier, you’re always on defense. This time, we got to be on the attack.

We approached the OEM and said we’d produce for 10% less. Since the order has been allocated, there was no budget for tooling, meaning we’d have to do it at risk. There was no way in hell we were going to open the full set of six tools for free without guaranteed orders. 

So Mr. M made his move. Knight to F6. With a few quick calls he found who was already making the part in China, prior to the Malaysian OEM being brought in. Another call, and he had the production manager. Just 10 hours later, under the veil of night, a truck pulled up outside their facility and 40 sets of samples were placed in the truck. A few days later those samples were in Malaysia on the OEM’s table as our first article. 

“Wow Mr. M, I can’t believe you opened those tools so fast! Amazing.”

“Anything for a good friend.”

Kickbacks

I learned, too, that a kickback is not always a cut-and-dry case of greed. It is often a down payment on future competence. The most brilliant supply chain manager I ever knew had a portfolio of homes across Asia, all funded by the “generosity” of his suppliers. The board knew, of course, but the man could move mountains with a single phone call, and so they contented themselves with looking the other way. O.K. Lim, the chain-smoking maverick who could move tens of millions of barrels of oil with a phone call, wasn’t taken down until he no longer had utility. 

My 2 cents: There’s rarely grand villainy in the bribe. I wouldn’t take offense if asked. It’s routine business, just part of their quote. Funds earmarked for dinners, golf, cash, KTV (karaoke bars), etc, usually 1-3%. But my strong recommendation is that the juice isn’t worth the squeeze. I’ve watched a lot of cash change hands, and what was asked of the people taking the money was always worse than what was being given. Also that 3% becomes 2.8%, then 2.5%, then ‘a little late’, to 0.5%, then nothing.

Running a company, you have to play whack-a-mole and take a hard stance even when the employee is critical. But be ready, when a fiefdom is toppled you must take the controls and be ready for some turbulence. From direct reports to QC to production to suppliers that might have been complicit. It is not for the weak and is usually very messy. 

Attacked from all sides. “Qing soldiers are coming!”

Profits in Malaysia were too good in 2021. 30% net profit in manufacturing is dangerously high, especially as Chinese industrial overcapacity hit new highs during COVID. We were on high alert. Mr. M strung a story together that had us prepping like it was the battle of Stalingrad! We must hold our ground!

Chinese businessmen poured in from Batam, Singapore, Thailand. Some were already behind our lines, quoting crazy prices and offering generous gifts, just to get a foothold. We’d match them quote-for-quote at every turn. Profits fell, but that was still better than letting them taste blood, because allowing a Chinese supplier to get a foothold doesn’t bring in one competitor; it brings in ten. And they play to win.

Free is the most expensive

I learned this lesson before Mr. M, at my first job in China, but it’s served me well since. The FAANG guys—the initiated ones—were particularly eager to hit the KTVs (karaoke) and experience the whole dragon, head to tail, as the saying goes. If you don’t know Dongguan’s history before Xi’s crackdowns, over 10% of its GDP came from the sex industry. The infamous Prince Hotel once had over 1000 ‘technicians’ working nightly, with underground tunnels in and out. When the owner was finally brought to justice in 2014 they sent in 6000 police to surround the hotel. 6000! The legend is harder to kill, and many still land in China with lust in their eyes dreaming of the past. 

One night sticks with me. I was in a room with very senior people from a Silicon Valley company. They wanted the usual treatment, but they’d also put heavy pricing pressure on us and our allocation was down. My boss wasn’t happy. So someone in our org hired working girls to wear our uniforms and accompany these three—and video taped the entire thing.

Part 3: The COVID Pressure Cooker

The start of COVID 

For me, it started in January, 2020. I stepped off a cargo ship in Odessa, Ukraine after some vodka-filled days at sea. We had to wait at the docks for customs, outside. I was woefully underdressed. On board, I’d met a few guys who said they’d take me to Kyiv if I drove one of their cars, as one of their guys had stayed behind in Georgia (hopefully with all extremities still attached!) 

Everyone on the ship was Ukrainian. They’d gone to Georgia to buy broken cars and bring them back—either evading import taxes for foreign cars or flipping used cars as new. Whatever it was, not completely kosher. 

From the port, we drove to a small farm house in the middle of nowhere. Nothing for 10 minutes and then one house, a barn out back. We pulled into the barn, then went into the house for coffee. As we left I noticed new license plates. I just need to get to Kyiv. 

Halfway to Kyiv, the front diffuser fell off my car. I reattached it with the strap from my luggage. An hour later, we were in Kiev. We stopped at a metro station. I got out, another guy got in. I never heard from those guys again. 

Late that night, I received a text from Mr. M.

“Check for N99 facemasks.”

I slipped into a pharmacy and replied.

“They don’t have any. I’ll check once I’m in England.”

I had a flight to London in the morning to meet with engineers at a customer’s office. 

In London, Mr. M kept asking when I’m coming back. Then he got direct.

“Go look for facemasks, buy what you can, and get back here.”

There were hints that what was going on in Asia would get more serious, but it wasn’t broadly recognized. I started going to pharmacies in London. Nothing. Everyone’s inventory was zero. If I were smart, I would have shorted stocks at this point. 

It wasn’t until I was in Leicester the following day that I found a hardware store on the outskirts, owned by this little old lady.

“Do you have face-masks?”

 “Yes, all different kinds. Which one do you want?” 

“I’ll take all of them.” 

Jackpot. I changed my flight for the next day and drove back to London. I felt like a conquering hero, returning with a trash bag full of spoils worth nearly $2,000. In a large transparent hand-carried bag I smuggled them from London to Bangkok to Singapore, a journey fraught with the peril of thieves and the strange looks of fellow passengers.

Upon return, we immediately packaged the facemasks, inconspicuously in a box and shipped it off to our factory in Dongguan, China. But they never made it. They were seized by the Chinese government!

Tensions rising

For all Mr. M’s bluster, I saw the mask slip once. We were on a monotonous ribbon of highway from Johor to Kuala Lumpur. Endless palm plantations. Without a word, a shudder ran through his shoulders, and the man who in my eyes could handle anything began to weep. It was the raw, unrestrained anguish of a man pushed past his limit.

I drove in silence, blinking furiously to hold in my own tears at seeing the man I revered so greatly reduced. I knew of the private war he was fighting—a long, silent illness that had taken up residence in his home and was slowly stealing his wife away. Then it was over. He wiped his face, straightened his hat, and stared out the window. We never spoke of it. 

3/18 – Malaysia officially shuts down. 

Malaysia shut down but never really shut down. You just had to avoid check points. Customers were complicit. “You have a choice. Figure out how to fulfill orders, or we will move it to Vietnam, Thailand, China… whoever can deliver.” That threat hit the OEM, and then the OEM passed it down. So, in the beginning, we did a lot of shipping under the cover of night.

Quarantine

I spent 56 days in different quarantine hotels, usually 7-14 days at a time. I actually got pretty good at this part. Even now, I have bouts of nostalgia. 

I’d bring tools to unlock the windows. Typically, the hotels were large, 20+ floor buildings. By the second or third time, I was sending tips to delivery guys to pour out water bottles and fill them with baijiu (rice liquor), put cigarettes in food (I since quit smoking, mom), and buy more books. 

I’d bring a pile of books and spend my days with a nice draft from the open hole in the wall and read. 

I even lucked upon a beautiful romance, started by writing messages on paper signaling the girl being quarantined in a tower directly across. 

But these days were like happening upon a hot spring while lost in the mountains. 

Disinfect THE AIR!

COVID was mostly constant stress. But occasional moments still bring a smile to my face. Below is something we had to pay for, and to a proper government agency! It wasn’t horribly expensive or anything, just … well, I leave it to you to be the judge.

Our factory was 60 m x 90 m x 12 m, with 6 open bay doors. I’m sure this did the trick though.

Staying afloat 

Manufacturing is suffering, followed by more suffering. If you can withstand the suffering and have just a wee bit of luck, you are rewarded. But you MUST remain vigilant! Because as you are rewarded, you shall be tested once again!

The bit of luck was profit. Prices were great for anyone that could keep going. But everyone was without a lifevest. It was hard to get parts. Even getting basic  maintenance was hit-and-miss. So we did it ourselves. We ran our factory everyday. 24 hours. The production manager and I were there for every daily meeting: 7:45 a.m. and 7:45 p.m. When things needed fixing, they happened on the spot. The goal was simple: keep going. 

***Don’t copy this—this is incredibly fucking dumb.

***Don’t copy this—this is even dumber. 

***Don’t copy this—not safe

***Don’t copy this—still not safe

Part 4: Russian Roulette

Not so different

Most factories in Malaysia were filled with foreign workers. From Bangladesh, India, Myanmar, Nepal, Indonesia, the Philippines. Half the time, we were miming commands. If it sounds radically different from the US, the conflicts weren’t—usually caused by someone not pulling their weight or the Brownian motion set off by boy-meets-girl. Racial flashpoints happened, but rarely. A pack of cigarettes and a chilled can of Coke was an essential part of the manager’s kit.

Every morning, workers showed up dressed to the nines. Hair jelled. A visible plume of cologne could be seen around the morning meeting. 

During my time in Malaysia I happened on a “free vacation” at an immigration holding camp. What a shitshow. Guards beat people by day then brought contraband to us by night. A story for another time. When I got dumped there, a guy who’d worked at our factory spotted me and shouted “Hey boss!”

I shook my head.

“Here. You-me, same-same.” 

He smiled. 

Road stop negotiation

Our factory was less than a kilometer from our dorm. Left out of the parking lot, immediate right, straight, right and you’re there. The roadblock was right outside the dorm, and I still drove into it. The cop asked for my passport, flipped through it, then motioned me out of the car. I couldn’t speak Malay; he couldn’t speak English. He walked me to their van, where two other officers waited.

One pulled out his phone, opened the calculator, and typed: 1000myr ($250).
I shook my head dramatically. Took the phone: 100.
He typed: 800.
I typed: 200.
He typed: 500.
I tried: 300.

He snatched the phone back, done with the game, and pointed at the backseat. I said, “OK, OK, OK” and handed over 500.

I gained a strange kinship with the foreign workers in our factory through this. We were all strangers in a strange land, trying to earn a living, all one traffic stop away from disaster.

Fun and Games

We were trying to flatten the ground to pour concrete. This was our method. The music seemed fitting.

On the same day, we got 3 separate vehicles stuck. 

Things change in an instant

What I didn’t see creeping in was sloppiness. Instead of killing it at the root, we laughed at it, let it slide, and it spread—layering on top of our exhaustion. People were getting sick, dropping out, and the agency sent replacements—that were inexperienced. And we didn’t onboard them properly. One night shift, one of them flipped the oven into a manual override and forgot to switch it back. By dumb luck the factory didn’t burn down. 

And then someone did get hurt.

I was next to it. Heard the scream. Smelled the burnt flesh. I drove her to the hospital, and they amputated her hand. 

This is still hard for me to think about, and it’s woken me up in the dead of night more than once since. Maybe if the project were grander, It would be easier to make sense of, but this was so one more coffee maker would make it onto the shelves. I don’t have much to say regarding how to avoid injuries in manufacturing other than—somehow you have to internalize that the stakes are people’s loved-ones, children. You can lose everything in seconds. 

It will always be the job of the seasoned to hold the unseasoned accountable. This was one sensor that went bad and wasn’t fixed immediately. A new employee came in and did something stupid. And if there is anything that can be counted on, it is for people to do stupid things. Make it so they can’t. Some things in life, 99% isn’t good enough. 

Our customer’s buyer was on-site when this happened.

“Is the tool damaged? We are behind schedule. Keep going!”

Part 5: Expansion & In Charge

Building factories. First you have to find one. And before you do that, learn what to look for.

”Go visit 50 factories first, then we’ll decide.” Mr. M.

We were in production hell, but over 4 months, I visited 80. He was 50% stalling, 50% teaching. Every factory was his own money (no leverage) so he would only pull the trigger on a new factory when alternatives were exhausted. And even then he’d wait a few moments longer. Like redlining your car. Oil boiling, pistons rattling, then at the last second, swerving into the pit stop. Maximum pain for all involved because, too many times, customers have overstated volumes or cut and yanked entire production lines to another country. 

Any person that owns a small business and is still open after 20+ years, without giving up ownership has chutzpah. I tip my hat to all of you. 

Finally I got the message: “Settle on a factory now.” He gives me a week. Typical Mr. M. Stall. Stall. Stall. Go! All the options were bad, so I picked the least bad. The next day I selected a factory and transferred the deposit from my personal bank account, just because it was quicker. I had all the contracts ready and I camped outside the landlord’s office. “Have you signed it yet? Have you signed it yet? Have you signed it yet?” Sometimes you can annoy people into action. 

That weekend, another agent messaged me. “There’s a factory next to your customer with 500KVA and enough space. Want to take a look?” I didn’t, because we’d already settled on another factory but said “OK”. Sometimes it’s the mere act of deciding on the final path, this is final and no one can change my mind otherwise!!, that flushes the right solution to the surface.  

The factory was in disarray and had been on the market for a while. It didn’t have floor plans. No fire system. Everything was in disarray, but in an instant, I could see our entire production inside the factory. I knew this was it. We’d have to redo everything, but god dammit. Who cares?

I sent my boss a message. “I found our second factory.”

His response was typically short. “OK. Go get that deposit back”  

Thankfully my ego was long gone. I went back to the first landlord and made sure to embellish my boss’s lack of sanity and my impending doom if I didn’t get the money back. I returned triumphant. 

Setting up the factory

In 3 months during COVID, amidst an even stricter lockdown, we went from this: 

To this:

No pain involved whatsoever!

COVID barbecues and calling ahead to change the production schedule when you’re in a pinch

COVID barbecues after 12-hour shifts are some of my best memories. I wouldn’t allow beer on-site now, but back then it was different. Special. In my head it lives as the best of times and the worst of times—straight out of Dickens. We’d invite customers, and it became something people looked forward to. They’d message asking when the next one was, if they’d made the invite list. 

The kernel here is that through those barbecues, we stopped being suppliers and buyers—we became friends. And that simple thing, being friends with our customers (what a radical idea), went a long way. Actually liking your customers, feeling like you’re on the same ship together, is still worth something. 

Because sometimes, when you’re in a pinch, you need help. 

Two examples: We were massively behind for multiple clients. On paper at least. OEMs routinely build a little inventory so there are no last minute surprises. As the supplier when you are hand to mouth, it can be the difference between being placed on the chopping block or not, knowing who is building inventory and who is actually using what you are sending. This buys you just a little more time. This friend would send me their internal production schedule and tell the warehouse to hold off on escalating if our shipments were late. If there was a problem, he handled it with us directly. If this isn’t amazing to you, it’s because you’ve never had to rely on forecasts for production. 

On another occasion, we were again headed towards cosmic death. All of our orders for clients tripled overnight, and we had no way to run all the tools at once. Don’t oversell your capacity, you might be thinking. Hold more inventory. Yeah, yeah, if you think that, just wait. Customers won’t reimburse you for the inventory you build, even when guaranteed by their purchase orders. Or they pay you 18 months later. Also, 9 out of 10 times their orders don’t live up to their forecasts. 

The same friend saved us again. He shifted production that day to a different model, one we had a little inventory of and could produce faster. Several other suppliers were hanging on the edge. He simply decided who would take the full wrath of the CEO and the customer, and steered the fire away from us. The supplier he turned the fire on suffered greatly. 

Haunting the factory

Mr. M was slowly passing the baton. I was also starting to find my rhythm, doing things the way I wanted to. Now it was me haunting the facility, and I did. I would stop by at 2 a.m. and see if things were working. You’d catch people sleeping all the time on the night shift. I’d show up right at 4 a.m. and make sure the 15-minute break didn’t bleed into twenty. It’s the curse of running something 24/7. On one hand, it’s hard to make money running one shift, but that money comes at a cost: your sanity.

The local guy 

I’ve found investing in a supplier that can be a fixer when you’re in a pinch pays dividends. My guy was Bin. He owned the shipping company we used. Bin was a solid figure with dark eyes and tattoos edging out from under his sleeves. A man with presence. A good guy with a bad background, filled with stories he didn’t volunteer even after a few beers. Good maybe isn’t the right term, but he was rock solid. His drivers knew everything that was happening, as drivers often do, and he made sure that information got back to us. If I called, he answered. 3 in the morning, if I needed a truck, he’d light up a cigarette, slip on his flip-flops, and be on his way.

During COVID, we were short-staffed for eight months. I didn’t have a warehouse manager or anyone to load the trucks. We were still shipping 100-200 pallets per day, and through these eight months, Bin’s drivers labeled and loaded all of our finished goods. I don’t recommend this as a strategy, but leaks in the ship are unavoidable. Having really good people to help plug holes is a lifesaver.

Drive-by forklift 

I could write a few chapters about Bin saving our ass, but for time’s sake, here’s one more.

Most OEMs require JIT (just-in-time). We were in Johor, but the OEM was in Penang: 12 hours by truck. A truck to Penang cost $500 USD. The twenty pallets on board were worth maybe $5,000. The truck left Wednesday night, delivered Thursday morning. Inventory: zero. We were shipping hand-to-mouth. Then we got a call. Shipment locked out.

AQL 2.5. From the lot, they sample 200 products, with a max of 5 defects. They found 8 or 9 pieces under/overweight. For pulp trays, if you weigh them right off the press, they can be 5-10% lighter than a day later. Especially if it rains, because they absorb moisture. Normally, for non-egregious stuff, you can make a call and someone will help, especially if you have good rapport, but we were the new supplier, and the QC was very disgruntled about management bringing us in. (I wonder why…) They gave an ultimatum. Replace immediately, or they’d escalate to the end customer.

We couldn’t admit we didn’t have any stock, because that would get escalated too. I called Bin to send a truck with four pallets up to Penang. Rented a forklift for a couple of hours. We rendezvoused on a rural street. Bin’s driver, Aman, unloaded the 20 pallets, stripped the top layer off each, swapped them for in-spec parts, relabeled, reloaded, and sent them back in.

All in a day’s work.

Part 6: Personal Payoff

Attaining Wisdom

God, I hope I got wiser. The alternative is too horrible to think of…

​​Mark Twain said it best in Roughing It:

“The moral of it is this: if you are of any account, stay at home and make your way by faithful diligence: but if you are “no account,” go away from home, and then you will have to work, whether you want to or not. Thus you become a blessing to your friends by ceasing to be a nuisance to them–”

You’re here, so I’ll assume you coughed up the same ticket I did for this rust eaten startup freighter—steaming toward glory through a fog of self-inflicted pain. The only certainties on board are leaks in the hull and midnight doubts about why we ever left port. 

So here’s my first piece of advice. 

Many suppliers carry PTSD from brutal OEM treatment. Every exporter has their war stories. You might be thinking: I’m a great customer, so why don’t they treat me better?  It’s because they’re waiting for the other shoe to drop. Building rapport takes time. Remember this. 

Second: There is a Chinese saying that goes “A distant relative is not as good as a close neighbor.” Yes, Chinese shops are faster and cheaper. Sometimes just getting quotes from US suppliers feels like pulling teeth. But be wary my friends, when the roof leaks on a Friday at 5 pm, that 8 000 mile relative can’t mop the floor, or your tears. 

Use Alibaba as a crash course, not a lifeline. Need prototypes? Fine. Need production critical to your launch? Layer backups. Place the same RFQ with three vendors. Then sleep with one eye open. Sometimes you’re up shit creek and have to throw a Hail Mary pass. Fine. But make sure your bases are covered. Imagine the flip side: Some stranger across the globe, barking demands in broken English, expecting a quote by tomorrow. Add in shifting politics, where the fading brand of the American Dream is now being overshadowed by TikTok clips of homelessness and gun violence.  

“The equipment of guerrillas cannot be based on what the guerrillas want, or even what they need, but must be based on what is available for their use.” “On Guerrilla Warfare” by Mao Zedong, Chapter 5, Part 3.

My third piece of advice is keep a check-list for talking with suppliers. Not every supplier will check all the boxes. Sometimes you have to do a risk buy, but do so consciously. 

Add to it, delete it, tailor it however you want, but then paste it to the side of your computer. 

• Where is this supplier from? Is that product known for being produced in that state/province? 
• My volume vs. their stated capacity. Am I ordering a box of screws from an iron ore company that’s listed on an Exchange? 
• Who is this? The factory? Trading company? An apartment?
• Can it legally ship? Any export controls?  
• Packaging Spec? Wooden box or plastic box? 
• Am I sending aerospace specs for a garden rake? Tolerances match process capability; if the CNC can only do ±0.1 mm, don’t send ±0.01 mm drawings.

• Have I checked the raw material prices? If a kg of aluminum is $2.50 and I’m buying a block that costs a dollar, I’m in for a surprise.
• Did they say it will be done “Friday”? This often translates to “Monday-ish.” Holidays? 
• Is this your forwarder? If you care about time, handle the forwarding yourself. 
• Has the supplier given the packing list, invoice, information about the product to the forwarder? This delays more orders than you can imagine. 
• Did you check that the raw materials were purchased? Photo! Progress of the products? Photo! Progress of packaging? Photo! Did your forwarder pick it up? Photo! 

Fourth – first appearances can be deceptive.

We were running 24 hours a day, seven days a week, and still couldn’t meet our supplier’s demands. The customer had buyers sitting on our factory floor with stopwatches.

My phone rang. It was Mr. T.

“Let’s have dinner tonight. Meet us at Spicy Momma’s Hotpot.”

Oh, shit.

Mr. T is the most humble customer CEO I’ve ever met, and not in a way that puts you at ease. It makes your skin crawl. He was six feet tall with a medium build, graying hair, a uniform of shorts, flip-flops, and a QC hat, with a cigarette always dangling from his mouth. He could out-drink anyone in Malaysia, was a fearsome gambler, a genius with numbers, and even better at reading people. And his bite was vicious. He would bring the wrath of God down on anyone who crossed him. Fear, that’s how he managed. Mr. T radiated warmth, but if you touched his skin, you’d pull back from the cold.

That night at the restaurant, I dutifully drank myself into a stupor without a word being said about my recent inability to deliver, which only made the reprimand worse.

At some point in the haze, I stumbled up to the counter to pay. The only thing I remember saying to the poor woman behind the register was a complaint about her beer. She didn’t have any in stock, so she had personally walked 400 meters to a store and carried two 24-packs back herself. I thanked her by asking, “Those are $2 a bottle. Why are you charging us $3?”

Here’s her side of the story: Sophia opened her restaurant in January 2020. Flawless timing. She was just trying to keep the lights on, and Mr. T was an important customer who used her restaurant for meetings/punishing suppliers. Sophia was secretly keeping the place open despite the country-wide lock down, just to survive. On this particular night, Mr. T brought in a particularly obnoxious foreigner—her words were “scum from the bottom of the barrel.” This asshole staggered up to her register, looked her dead in the eye, and complained that her beer was too expensive.

I radiate charm and charisma. Reader, I married her. Yes, really.

Fifth, make sure to check your own hubris every now and then.

…from personal experience… 

My Triumphant Return to the States. It’s been 13 years and I’m finally moving back home. I tell my wife–why use air freight when we can save all this by sea.

“I’ll handle it.”

I get multiple quotes. At this point I’m an expert. I know what to look for. Everything is smooth. I brag to my friends how much I’m saving.  $1.40/kilo. Damn good! Then the add-ons start. Fabrics, +$0.07/kg. Residential address, +$3.6/box. Oh, you have brand-name items, +$0.07. For 61kg of stuff, I’m ultimately charged for 78kg. Total cost: $135. Ha! I knew that was coming. No problem! I pat myself on the back for saving $500 and a job well done.

Two and a half months later, two withered packages showed up on my doorstep, one with my books, and the other with 1,000 plastic kid tiaras inside. My wife’s shoes, clothes, and bags are gone forever, but at least they eventually gave me a refund. $78.  

Which in a broader context sums up this entire piece quite nicely. I left Cal Poly in order to save $80,000.

And look at me now! Think of all the money I saved!

Part 7: How I got out of a Malaysian prison

Kneeling at the final gate, I looked out into an expansive, two-story corridor with cells on both sides. The silence changed to a hum, then a buzz. A cacophony of yelling inmates erupted as they saw the lonely foreigner being escorted in. Oh, fuck. I’ve seen the movies. 

They led me to a cell and slammed the door with extra force. A mixed-race tree trunk of a man walked over, wearing nothing but a pair of tighty-whities. Next to him were what I gathered were his henchmen, one with a long gash across his head. Both looked dumb, but the way their bodies were scarred up, it was clear they didn’t care much for their own well being. So they probably wouldn’t care much for mine, either.

50 inmates to a cell. Something like 4 steps by 15 steps. I paced this many times.  Here’s what Gemini says: 

“I’m Big Baby,” the tree trunk said. If I wasn’t scared out of my mind, I might have found this funny.

Wait… did he have a Chinese accent?

I said something to him in Chinese. “Nǐ hǎo, xiōngdì…” (Hi, brother…).

A smile broke out on his face.

Then, a beautiful voice cut through the noise. The chaos of the prison hadn’t stopped since I’d arrived; men from other cells were still yelling to get my attention. Another voice joined the first, in perfect harmony, and the sound became unmistakable—it was the last call to prayer of the day. It’s hard to put into words, but the acoustics of that prison were wonderful. The sound echoed smoothly off the walls. In each cell, voices began to chime in. I forgot how long it lasted but it ended with everyone shaking each other’s hands. I’ve never been so grateful to stumble upon religion. 

To some, two weeks in prison sounds like a lifetime. To others, they’ll scoff that I’m making such a big deal of it. Neither of these relate to my point. I never felt pity for myself. I did cry once, a moment of emotion I was sure would make me a target, but I was immediately comforted. “This place breaks everyone down,” an older inmate told me. What I felt then and still feel today is some weird gratitude. I was there by accident, but my trajectory would have put me there eventually—or in some place much worse. In hindsight, it was one of the best things that ever happened to me. The stay was short, but it was just the nudge I needed. I got COVID inside and spent a week in a delirious state, probably closer to death than I like to admit. The experience didn’t make me stronger—it definitely made me weaker, ha!—but it was all a small price to pay for a new lease on life. 

Part 8: How I’m currently wasting my life

For a brief period, in this niche field, I was known. Mr. M was looking for a successor and thrust me to the edge of the industry. “I was 34 when I started this company!” he’d say. “You’re not even 30. You’re barely at the starting line.” 

What future do you want for your kids? Many have said something similar but I first heard Elon say this.

Sometimes you just hear something at the exact right time that you need to hear it. I’d always wanted children, but thought marriage would be better if treated like a factory contract: fixed amount of time with the option to renew. In practice, 18 months into raising my first child, I’m damn happy it turned out a bit different in practice!

I left the packaging industry not because I think there’s more honor in tech. Packaging, the noble art of preventing catastrophic disassembly during transit, will still be important when we’re living among the stars. What drew me back, and ultimately to Terraform, was the chance to work with people who are either optimistic about the future or pessimistic enough to have the agency to try and change it, which the US still has the highest density of. Everyone told me I was dumb. I was throwing away my career, everything I’d built, just to start over.

There’s an ancient Chinese saying: “Break the kettles and sink the boats.” It refers to Xiang Yu’s order at the Battle of Julu (207 BC); by fording a river and destroying all means of re-crossing it, he committed his army to a struggle to the end with the Qin and eventually achieved victory. I took this approach.

It wasn’t a straight line. For a quick minute I was lost—I nearly pivoted into manufacturing dried orange peels, then tried to rope a friend into a dumpling franchise. Thankfully he had the good sense to decline. I had decided early on I was going to read my way out of this fog. A book that I credit with setting my new trajectory in motion is Deutsch’s Beginning of Infinity. The entire book is phenomenal but the first seven chapters are some of the best and most inspiring pages ever written. Don’t trust me though, Naval speaks ad nauseum about this book. Then I found myself in the tech-manufacturing corner of X, where Jason’s S3 pointed me to Terraform. 

So here we are. The world is burning, the AI is coming for us, and we’re all probably going to die in a fiery heat death by the sun. But hot damn, it’s exciting! Because it is a thrill of a lifetime to be at Terraform Industries with so many top shelf atom-manipulating MF’ers fighting to live another day. 

I finally found the right castle.. 😎

Terraform has a wide range of life-ruining, career-ending opportunities for anyone sufficiently psychotically ambitious to, like Ken, believe that yes, really, we can upgrade the way humans get raw materials. Come over and build synthetic natural gas, oil, ammonia, steel, non-ferrous metals, desalinated water by the millions of acre-feet, cheap solar, AI power systems, and so on!

“The difference between screwing around and science is writing it down.”

— Adam Savage

If you’ve ever tried to drill 5 evenly-spaced holes around a circle, you know exactly why I’m writing this. If you haven’t, let me save you some pain: it’s where elegant mathematics collides with shop floor reality, and reality always wins.

Actually, let me back up. I know this particular pain intimately because I once decided—in what can only be described as a fever dream of automotive obsession—that my 1962 Mercedes-Benz W110 190D absolutely needed to wear 1990s Mercedes monoblock aluminum wheels. Not just any wheels, mind you, but the specific AMG-style monoblocks that came on the W124 500E—that Porsche-built super sedan from the Bush Sr. era that rolled on the iconic “Gullideckel” wheels.

These weren’t your garden-variety rims; they were the distinctive 15-hole, flat-faced alloy wheels that defined Mercedes-Benz in the late ’80s and early ’90s, nicknamed ‘manhole covers’ because some German engineer apparently looked at street drainage covers and thought, “Yes, this is exactly what our luxury cars need.” I needed them because I’m the kind of weirdo who thinks a 40-horsepower diesel sedan built during the Kennedy administration deserves the same rolling stock as this automotive masterpiece. Some people call it auto sports, but I know it’s actually called Autism.

Here’s the thing about being probably the only human on Earth with this particular flavor of auto-ism: nobody makes adapters for this. Nobody. Because who else would look at a car that takes ≥37 seconds to reach 60 mph and think, “You know what this needs? Wheels that were designed for Autobahn cruise missiles.” 

But I had a vision. And access to time shares on CNC mills. And more importantly, I had the kind of stubborn determination that makes you spend three weeks figuring out how to perfectly adapt a wheel bolt pattern from 1962 to wheels from 1992, because Mercedes changed the hub bore, the offset, and literally everything else except the 5x112mm PCD (Pitch Circle Diameter) in those 3 decades.

The Problem (That I Created for Myself)

Here’s what every engineering textbook tells you: 360° divided by 5 equals 72°. Simple. Clean. Done.

What I imagine they don’t tell you: when you’re standing at a mill at 2 AM on a work night, trying to position those holes in a custom aluminum spacer that needs to perfectly mate two different generations of German engineering—while maintaining proper hub-centric fitment because you’re not a complete barbarian—you’re working with coordinates like:

- Hole 1: X = 56.000, Y = 0.000 (the only nice one, and yes, 56mm because that's half of the only solid, the 112mm PCD)
- Hole 2: X = 17.305, Y = 53.246 (welcome to decimal hell)
- Hole 3: X = -45.305, Y = 32.920 (it gets worse)
- Hole 4: X = -45.305, Y = -32.920 (at least there's symmetry)
- Hole 5: X = 17.305, Y = -53.246 (are we done yet?)

Unlike 4, 6, or 8-hole patterns that give you nice round numbers or simple fractions, 5-hole patterns force you into irrational number territory. It’s sin(72°) = 0.95106, and no, that’s not going to simplify nicely. [Editor’s note: It’s simply sqrt(5/8 + sqrt(5)/8) for those of you keeping score at home.] And when you’re trying to hold tolerances tight enough that a German wheel from the era of “Rhythm Is a Dancer” will center perfectly on a hub designed when “Moon River” was topping the charts, every decimal place matters.

I think I spent two solid week nights just figuring out the proper math. Not because I couldn’t do the math—I can calculate sin(72°) in my sleep at this point—but because I kept second-guessing myself. When you’re about to drill into a $200 mill cut of 6061 aluminum round, and you know that one misplaced hole means starting over, you check your math. Then you check it again. Then you have an existential crisis about how food would probably be a better way to spend your money, and whether you really understand the unit circle. Then hungrily, I checked it a third and fourth time.

Why This Obsessive Precision Actually Matters

You might think this level of obsession over wheel adapters is excessive. And you’d be right. But my manic attention to detail—my refusal to accept “close enough” when precision is possible—is exactly what’s needed when you’re trying to revolutionize energy production.

Today I’m at Terraform Industries, where we’re building machines that literally pull fuel from thin air. We’re converting sunlight, water, and CO2 into natural gas through processes that would make your thermodynamics professor weep with joy. The engineering involved is legitimately rocket science levels of complicated. But you know what all those brilliant chemical reactors, heat exchangers, and pressure vessels have in common? They’re held together with bolts. In holes. That someone had to drill. Precisely.

The difference is, when a wheel adapter is off by a few thousandths, the car vibrates a little. When a pressure vessel flange is off by a few thousandths, well… let’s just say the consequences can involve more paperwork and fewer fingers. The stakes are higher, but the fundamental challenge remains the same: turning theoretical perfection into physical reality.

We live by a simple truth: the fastest path to breakthrough innovation is through rapid physical iteration. Not endless modeling, not perfect simulations, but actual hardware that we can test, break, learn from, and improve. And when your iteration speed is limited by how fast you can turn ideas into testable prototypes, every bit of practical knowledge matters.

My Mercedes wheel adapter saga taught me this lesson years before I joined Terraform. Those adapters? They worked. Perfectly. My ’62 now rolls on period-incorrect but aesthetically perfect 1990s monoblocks. But more importantly, the process of making them taught me that the gap between “I know the math” and “I can make the part” is vast, treacherous, and filled with expensive scrapped aluminum and countless broken end mills.

The Documentation Gap

I’m not a classically trained millwright, and I would hesitate to associate as a machinist. I’m a hacker who learned by doing, breaking things, and most importantly, by documenting both successes and spectacular failures. Every crashed tool, every scrapped part, every “that’s not how that works” moment became part of my dataset.

After years of accumulating this knowledge—and months spent recently watching all these brilliant engineers who surround me, often contending with the gap between theory and practice—I realized something: we’re all solving the same problems, making the same mistakes, learning the same lessons. But we’re doing it in silos. We’re struggling in isolation.

Think about it: somewhere right now, someone is standing at a mill, trying to drill 5 holes and a center hub bore, doing the same trigonometry I did, having the same crisis of confidence, making the same mistakes. They’re probably not making Mercedes wheel adapters (because again, that level of specificity is probably unique to my particular brand of obsession), but they’re fighting the same fundamental battle between theoretical knowledge and physical reality.

So I started writing it down. All of it. The tribal knowledge that usually takes decades to accumulate:

### The 5-Hole Bolt Pattern (Have Mercy On Your Soul)

5-hole patterns are rare for a reason. They're a huge pain in the ass.

Starting at 0° (3 o'clock position) and going counterclockwise:

- Hole 1: 0° = X: radius × 1.00000, Y: radius × 0.00000
- Hole 2: 72° = X: radius × 0.30902, Y: radius × 0.95106
- Hole 3: 144° = X: radius × -0.80902, Y: radius × 0.58779
- Hole 4: 216° = X: radius × -0.80902, Y: radius × -0.58779
- Hole 5: 288° = X: radius × 0.30902, Y: radius × -0.95106

Write these multipliers down. Tattoo them on your arm. Whatever it takes.

Pro tip from my Mercedes adapter adventure: Always drill a test pattern in
scrap first. Always. That $200 aluminum billet slab doesn't care about your
confidence in your maths.

Interactive: coordinates for any number of evenly spaced holes. Open the interactive on the blog.

Knowledge Beyond Bolt Patterns

Here is a bit of backstory. Ever since childhood I’ve been documenting my mechanical adventures. Long before the Mercedes wheel obsession, and long before Terraform. It started when I was maybe six or seven years old, and Mom made the critical error of giving me an expensive radio-controlled monster truck for a birthday gift. While other kids would have been jumping it off ramps and racing it up and down the driveway, I found a screwdriver set and methodically disassembled the entire thing. Not because it was broken—It was new and it worked perfectly. I just needed to know HOW it worked to be able to fully enjoy it. For me‒enjoyment in life comes through thoroughly understanding how and why things work. 

This approach works great with machines. Less so with people. You can’t take people apart to understand them, and even if you could, I’m pretty sure it’s illegal. Plus they never seem to come with service manuals or wiring diagrams, and they are next to impossible to reassemble.

I still remember the panic my poor Mother experienced when I couldn’t figure out how the plastic gears in the differential went back together. I cried and she had no consolations to offer. I took it apart, and that was the end of the RC Monster-Truck. But all was not lost. Because I filled my notebook. I sketched every gear in crayon, every bushing, every mysterious spring whose purpose my seven year old mind could not divine. Long after the other kids RC truck batteries were dead, or the toy broken, I was transfixed, filling my notebook with every intricate detail of the toy’s design, as I tried to grasp its function. That spiral bound notebook was one of many.

That panicked kid trying to reassemble a differential grew up to be someone who documents everything—not because anyone asked me to, but because I physically cannot NOT document it. It’s compulsive. It’s who I am.

The documentation evolved with the technology. Started with spiral-bound notebooks through the ’90s. Then around 2000, I discovered the power of digital documentation—first Emacs, then Vim, because why use a mouse when you can keep your hands on the keyboard? Suddenly I could grep through years of notes, cross-reference failures with successes, build my own knowledge base that grew with every project. PS: ripgrep! Welcome to 2025!

A laptop became as essential as any tool in the shop. I’ll have vim open in one terminal, FreeCAD in another window, and usually some forum post about why someone else’s similar project exploded. Every experiment gets its own set of files. Looking back‒my file names alone tell the stories of my many frustrations and breakthroughs.

My 5-hole pattern documentation became a tutorial. It is just one of the chapters in what’s become an encyclopedic collection of hard-won knowledge. Because here’s what I’ve learned after tens of thousands of hours of figuring out how stuff works: everyone hits the same walls. The intern trying to find the edge of their workpiece is having the same crisis I had when I discovered that “eyeballing center” is a recipe for scrap. They’re learning what I learned the hard way—like that spindle speed isn’t just a number, it’s a contract between your tool and the material you’re working.

We’re all just trying to turn ideas into reality, and reality has very specific, non-negotiable requirements that nobody teaches you in school. Theory says “apply force F to achieve result R.” Reality says “apply force F at the wrong angle and your part becomes a projectile, your tool becomes shrapnel, and your day becomes painful and expensive.”

The gap between knowing something should work and knowing how to make it work shows up everywhere:

• Edge finding (you can’t drill a hole if you don’t know where you are—learned this the hard way when my first adapter’s holes were all shifted 0.030″ because I trusted my eyeball instead of knowing how to use precision measuring tools to make it correct)
• Dial indicators (the difference between “yeah, it looks straight” and “it is straight”—critical when you need that hub bore to be concentric within 0.001”)
• Speeds and feeds (why that horrible squealing means you’re about to break a $150 end mill—RIP to the many that have made the ultimate sacrifice)
• Work holding (because parts launching across the shop is bad for morale—and bad for windows)
• Surface finish (when you’re making cosmetic parts that will be visible every time you walk up to your car)
• And yes, compound angles (for when one angle isn’t enough suffering)

Each tutorial exists because I’ve personally screwed it up, figured it out, and realized someone else is going to face the same problems.

A Transformation

Here’s what happens when you bridge the theory-practice gap: people start designing differently. They iterate faster. They try crazier ideas because they know they can test them today, not next month.

Competence is contagious. When we see colleagues confidently turning ideas into hardware, everyone wants that superpower. The building bug spreads through teams like wildfire. Suddenly, “let me try something” leads to immediate action, not hours of technical meetings.

It’s the same phenomenon I’ve experienced before. Once I successfully made those wheel adapters—once I proved to myself that I could take an absurd idea and manifest it in aluminum—everything changed. Suddenly, every weird idea becomes possible. Custom electric drive-train conversions? Sure! Bespoke transmission brackets? Absolutely! Adapter plates to merge a Toyota / Lexus electric drive unit with a 1960s Swing-Arm rear differential? Already done!

Why Open Source?

It’s well known that Millwrights in traditional machine shops guard their knowledge like state secrets. But we’re not trying to corner the market on machining wisdom—we’re trying to solve climate change. If my tutorials can help another hard tech startup iterate faster, that’s a win for all humanity in my book.

Think about it—if historical innovators like Edison had Git and GitLab, they could have version-controlled their 10,000 failures. Instead of hoarding knowledge in personal notebooks, they could have collaborated across continents. Today we have that opportunity. We can share our successes AND our failures, let others fork our mistakes and merge their improvements.

Because at the end of the day, hardware is hard enough without artificial barriers. And saving the world is going to take every advantage we can get—including knowing that sin(72°) = 0.95106, and that you should probably just use 6 holes instead.

(Seriously, use 6 holes. Your future self will thank you. Unless you’re adapting vintage Mercedes wheels, in which case you’re stuck with 5, and you deserve the suffering you’ve signed up for.)

The complete Machine Shop Tutorials are freely available at github.com/JoNilsson/machine-shop-tutorials. Everything from basic tool use to advanced techniques, written in plain English by someone who’s made every dumb mistake so you don’t have to. MIT licensed because the future is too important to gatekeep.

Hey! Are you the kind of person who wants to know both the theory AND the practice? Terraform Industries is hiring engineers and hackers who understand that the best ideas are the ones you can actually build. Drop us a line at hiring@terraformindustries.com with evidence of something you’ve made. Bonus points if it’s something nobody asked for but you built anyway because your vision demanded it!

I knew the moment we first met you were a very unusual person. When did you realize that … you weren’t like the others?

That’s an interesting question. What’s unique about me is that I never adhered to the conventional tradition or stereotypes of my generation. While my female cousins were playing with dolls, I was reading books on how to become a scientist and the impact that scientific research has on the world. Marie Curie was my north star, while the cutest pop singer was the one for most 8-year-old girls of my school. Ultimately, I joined Terraform because I knew it was unique and it would allow me to test the limits of my ambitions. But I grew up in a pretty isolated environment and didn’t recognize that my attitude was unusual until later in life. 

Tell me about your childhood.

I come from a very small village in Bourgogne, (Burgundy in English), located in the middle of nowhere, pretty flat, in France. I grew up on a flower farm that my dad built when he was 17 with his brother, a family-owned business. I spent the first 18 years of my life working at the farm in all my free time: taking care of the flowers, chopping trees, building irrigation systems so my uncle didn’t have to water 10,000 plants every day by hand, or just selling the flowers and answering people’s questions: any type of advice they would need about sun, shade, what comes back every year, what to use for their garden.

During that time, I was also pretty curious, and I realized we had some unforeseen events happening at the farm that my dad couldn’t explain: severe drought and wind that took out the greenhouse. “In 30 years that I’ve done this business, I’ve never seen this before,” he would say (in French, of course). I asked myself why. At school, my science teachers were pretty adamant about answering some of these questions. The answer was climate change.

I decided that I was going to dedicate most of my life to solving that problem. At 18, I left the farm and went to the big schools in Paris to study maths, physics and chemistry. I fell in love with chemistry, which is one of the most applied sciences.

When did you realize that you had the ability to contribute to the sum total of humanity’s knowledge?

Before I signed up for my PhD, I was curious about everything. I didn’t have the maturity to sit down and actually think beyond myself. Only when I realized that the work I could do could benefit both me and others did I look outwards and shifted toward contributing to society.

What were the basics of your PhD?

I started my PhD in September 2015 at the University of Calgary in Alberta, Canada, under Dr. Warren Edward Piers, a very well-known inorganic chemist who specializes in understanding how reactions take place. My contribution was to take a new family of molecules that employed Earth-abundant metals and study them for alternative energy reactions. Most chemical transformations are made using rare metals that are expensive and finite. Could iron drive the same transformations? Could we modify the conditions to make it work? (Spoiler alert: Yes.)

How long did it take?

Five years, including the COVID pandemic which made it a good time to finally write my thesis.

What did you learn beyond the pure details of chemistry?

How to be organized, detail-oriented, persevere, and understand when a project had to be cut off and rerouted from one path to another. These are the meta tools of the trade. I also learned how to mentor people and educate the next generation of scientists coming through the university, not only teaching them in courses but working with other folks in the lab with me.

How did the mentoring you did in your PhD compare to managing the reactor team at Terraform Industries?

My PhD gave me a basic understanding that the fundamentals of mentoring and management come back to taking care of the human side first, before the technical experience. You can’t teach people without realizing that they are humans.

From then to now, it’s the same thing: Make sure people are heard, understood, have correct directions, and that you care for them. If you don’t do that first, you can bombard them with technical directions, but they will lack trust and won’t be receptive. You have to establish a relationship based on trust first. Without that, nothing else matters.

What other activities depend on trust?

When I met my husband, he was (and still is) crazy about extreme sports. So I got into skiing and rock climbing. Both require (and challenge) the ability to make the right decisions, and trust your partner with your life, literally. It’s the same way you trust a pilot to fly the plane. You haven’t interviewed them before, but you trust they care enough for the passengers to make the right call.

Sometimes rock climbing is terrifying. You could be protected by a rope but still risking a giant pendulum or breaking a hold. If you fall, you could take a giant whipper 30, 40, 50 feet into some adjacent rock face. Sometimes it’s very focusing, especially in a no-fall zone. Like engineering chemical reactors, climbing is a team-based cerebral risk management exercise, not just screwing around trying to die.

What happened after your PhD?

In my 4th year of grad school (2019), I went to a conference and learned about using ammonia as a synthetic fuel. There was a groundbreaking study published in Science that someone presented and I got inspired. Could ammonia become, indeed, the fuel of the future? I decided to go back into the lab to see if the molecules I’d been working with could be used for ammonia activation, which they could. The brief highlight of my last PhD work was that Earth-abundant iron can be used for ammonia oxidation. This is necessary because burning ammonia can create NOx, an air pollutant, so first we want to catalytically decompose it so we only burn the hydrogen part.

But we also need to make green ammonia in the first place. Synthetic ammonia was invented in about 1912, through the Haber Bosch process, but it’s still an incredibly energy-intensive industry. I had a desire to continue my career in academia, which requires going through a big university in the US. I applied to Caltech, was accepted to work in Jonas Peters’ lab in chemistry, and worked there for two years on the electrification of ammonia synthesis directly from nitrogen and water.

How did you find out about Terraform?

The most unusual referral ever. I went rock climbing at Black Bluff, a very tiny wall close to Big Bear in California. I was at the base of the route with three other guys who were also rock climbing. Their belayer started talking to me, asking basic questions. Who I was, what I did, whether I was looking for a job. Maybe, I said. Why? He was working for a small startup currently looking for a chemist to help with a carbon capture and utilization project. That’s how I met Graham Kersey. This happened on a Saturday. On Monday, I received a DM from Casey asking if I was available for a chat. 

Then what happened?

I went through the interview process, probably one of the first 10 people interviewed at Terraform. It was very exciting once I realized that I wanted the job and was ready to leave academia. Now the shoe is on the other foot, since I’m doing so much interviewing to continue building the team at Terraform.

What’s Graham doing now?

He’s building Mechazilla at Starbase. He really enjoys the “high angle” work. But secretly I know he misses Terraform Industries. 

You often talk about your five core values, what are they? 

I’ll list them and then unpack them one by one. Safety First. Dedication. Organization. Team Spirit. Honesty. 

Safety First: How should you approach safety in such a high-risk environment?

Working in a chemistry lab and rock climbing have taught me that nothing can justify cutting corners, being complacent, or ignorant. At Terraform, I was, until recently, the safety officer, and my role since Day 1 has been to build a strong, positive, and successful safety culture. Prevention and education are key to achieving these goals. However, too much safety prevention and training can lead to zero progress, and that’s not helping anyone either.

That’s why I decided to empower everyone at the company to share in our safety culture and actively participate in it. I’m not running around with a police hat telling people what to do. Instead, I use my expertise and that of my colleagues to educate each other and exchange knowledge about safety hazards, risk assessment, and mitigation. When near misses or incidents occur, because they inevitably do, we learn from them and do better. We iterate. So far, I’m proud to report that the worst thing that’s happened in our hardware shop is a badly bruised thumb caused by a dropped wrench.

Dedication: What drives our passion for synthetic fuels?

I cannot dedicate myself to something I’m not passionate about. How did I become passionate about making synthetic fuels? It seems counterintuitive, and it was initially after spending years looking for alternative fuels to replace hydrocarbons. During my PhD, I worked on nitrogen-based fuels and other fuels that wouldn’t create carbon emissions. I joined Caltech with the hope of making green ammonia a reality.

All of these projects were fantastic to work on, and I do believe they will become a fundamental pillar of our energy system. But unfortunately, they require years, decades of development to reach the industrial level of hydrocarbons. Additionally, most of them are currently, one way or another, derived from fossil energy.

We have three problems to solve that are tightly intertwined:

• Our society requires a continuous flux of energy, and this energy demand will continue to grow.
• Because of our linear relationship with carbon—we extract it from the ground, burn it, and it stays in the atmosphere—our energy crisis is quite literally fueling our climate crisis, and we’re running out of time to do anything about it.
• Renewable and most “green” energies do not meet the requirements of continuous energy supply, meaning our energy transition is slow and expensive.

Casey Handmer, founder of Terraform, realized this years ago and proposed a solution to bridge two worlds that interact very little: making cost-competitive synthetic fuels using excess solar energy that would otherwise be wasted. We’re bridging the gap by making the same molecule that comes from the ground, but instead, we’re using captured carbon dioxide as our feedstock. We solve all three problems: we’re not fueling our climate crisis; on the contrary, we’re recycling CO2, we’re using renewable energy, and we’re providing the world with cheap and continuous energy.

I’ve been excited about accomplishing this goal since Day 1, and I’ve been dedicated to our mission and to sharing it internally as well as externally.

Organization: How do you manage the complexity of your role?

I see organization like a chess game: I need to be three moves ahead to stay in the game. I’ve been playing the game for a while, so experience helps here. Google Calendar, spreadsheets, roadmaps, whiteboards, sticky notes, clear communication, and clear expectations are all the tools I use to win the game. So far, it’s paying off.

Team Spirit: What’s your philosophy on collaboration?

The most rewarding experience of maturing was when I realized I didn’t have to be the best all the time and work alone. You can have five individuals working on their own projects, or you can have the same individuals working together to tackle a giant mountain. I’m convinced that the big problems at TI, and in general, can only be solved if we work together, share ideas and opinions to come up with the ultimate solution.

I keep repeating this, but we’re not competing with one another. The only competition out there is climate change and our energy crisis that urgently need to be solved. I’m proactively building team spirit at TI: I give each member of my team the opportunity to provide feedback, share ideas, and be part of the decision process. Ultimately, I find it most rewarding when my team accomplishes a big project together and gets to share the success thanks to collaboration.

Honesty: Why is direct communication so important to you?

People don’t know if you don’t talk to them. To the best of my knowledge, no one can read minds. That’s why I pride myself on honesty, in giving people feedback and having honest conversations. Clear communication and avoiding buildups is crucial for a small startup to survive. We’re spending 8+ hours a day together, we’re building relationships, and it takes honest and constructive feedback to get to an A+ team.

That’s a great set of values to bring to work. What have you learned here in the last three years?

Five more values I learned at Terraform are: Delete. Simplify. Spend Money. Iterate. Think Big. 

Delete: How does this principle work in practice?

The first question during our design reviews is: “Why do you need this part/system/valve?” If you cannot justify it, then you delete it. This concept is easy for me to adhere to from a minimalistic mindset, since it’s akin to “refuse, reduce, recycle.” But for many engineers, this is extremely counterintuitive, as many tend to solve complex problems with complex engineering solutions.

One example of this in reactor design concerns mass flow controllers. This is an expensive and finicky part that is used to meter precise ratios of reagent gas flow into the reactor. Casey had always insisted that we delete this part to reduce cost and improve reliability. It was a scary step to contemplate but we have ultimately been successful in obtaining this functionality from other components that are needed anyway. 

Simplify: What does “good enough” mean at Terraform?

Our machines only need to be “good enough.” Let me explain. One hundred years ago, we developed the lawn mower. Roughly speaking, it has a blade, two wheels, and a handle, and more importantly, it gets the job done cheaply: cutting the grass for $50. Over 10 years, that’s a cheap machine. Sure, it’s not the most efficient device out there, it might leave 10% of uncut grass, but who cares?

These days, electrical lawn mowers with cameras and motion sensors are flooding the market, costing two orders of magnitude more than the good old manual lawnmower. While most engineers would think that the electrical, fully automated, more efficient lawnmower is a great advancement, we at Terraform Industries see this as an over-engineering solution. Delete parts, maintain simplicity to get the job done as fast as possible.

Over the last 100 years, enormous quantities of effort have been expended on catalyst research and design, including my own PhD work. There are at least a dozen catalysts that can perform the Sabatier reaction we use to make methane from sunlight and air. Some of them are extremely exotic, containing all kinds of rare and expensive herbs and spices and achieving spectacular, albeit not exactly robust, performance. At Terraform, we adopted the oldest, simplest, cheapest, and most robust catalyst, because we can engineer around its shortcomings and then have total confidence that it won’t be poisoned by some parts-per-billion contaminant. 

Spend Money: How does this philosophy change the development process?

When I interview folks, one of their pet frustrations is how long it takes to procure anything. When I ask “How long did this take?” they often respond, “Weeks, because I needed approval to order a $150 part.” Right there, we’ve deleted this delay at Terrafrom. We have a company card and we spend money to save time or save CAPEX.

From before I started working at Terraform, Casey had written our monthly, weekly, daily, hourly, and minutely burn rate up on a whiteboard, both on a company basis and a per person basis. We constantly remind people to recalibrate their intuitions about saving money vs time. We love nothing more than a nasty technical problem you can solve simply by spending what might seem like an unreasonable quantity of money. 

Iterate: Why do teams get stuck in analysis paralysis?

Because 1) they want the best solution on the first go and 2) don’t want to fail. At Terraform, I encourage everyone to fail fast, so that we can learn and finally celebrate when we’ve failed enough times to make it succeed. It’s too easy to spend months obsessively listing everything that could possibly go wrong and then getting worked up over trying to retire this risk on paper. It’s far better to build a scale prototype with your hands in the workshop, break it, and learn. 

I would rather have my team improve by 4% every month instead of them spending 12 months for a 60% improvement. Both lead to the same improvement amount (60%) but the latter will result in substantially less uncertainty due to the real time nature of iteration and much more accumulated learnings and more constant motivation. On top of this, marginal returns to more “thinking” drop off very quickly in the absence of real world data, while it’s almost always possible to think of some short term incremental improvement, by which you can quickly chain your way to success.

Think Big: What’s limiting you?

I constantly have to remind myself why I do the everyday tasks: because they are a part of the big picture. I joined Terraform because of the huge positive impact it can have on the world. I’m not spending pocket money I’ve made on the farm anymore, I’m responsible for millions of dollars of investment, which will be used to supply the planet with synthetic and renewable fuels. It’s inspiring to realize that a good decision made today might have a 2x impact on outcomes in a couple of years. To get there, though, I have to expand my capabilities and push my own limitations constantly. It takes self-reflection and even listening to feedback. 

Is your team too big, Lucie?

Not at all. We’re hiring for a variety of roles. We have about half a dozen projects in various states of maturity, ranging from an electrolyzer going into mass production all the way down to benchtop scale experiments. We’re currently assembling our first full scale methanation reactor, which will be capable of producing over two million cubic feet of artificial natural gas per year! My team is also looking to ramp up work on a methanol reactor and an ammonia reactor. At the same time, the reactor injection system is a standalone subsystem which can meaningfully absorb more talent. Lots of roles are listed on our website at terraformindustries.com.

What sort of talent are you looking for?

We like people who are ambitious, self aware, and have a track record of success. Terraform is, I think, the best place for people to rapidly level up their skills and really test what they are made of. We reward ambition, competence, and growth. We’re actively inventing an entirely new primary material production layer, one that is independent of fossil fuels and thus future proof. I’m confident that in 10,000 years, humanity will derive their stable material abundance from the foundations we’re developing here.

Any final words?

From a flower farm in Burgundy to synthetic fuels in the Mojave—this is what relentless dedication to solving climate change looks like. If these values resonate with you, maybe you should be building with us.

Xavier Dedenbach, Terraform Power Electronics Engineer

Originally posted on April 20, 2025.

Summary

This article takes you on a journey to explore whether East-West fixed solar arrays are now more economical than the dominant variant, single-axis tracked arrays. If you just want the answer, feel free to jump to the end or read this summary. But if you have 15 minutes to follow along, the juice is worth the squeeze.

We’re hiring at Terraform!

💡Today’s Claim
PV modules are cheap enough today that the simple fixed East-West arrays are cheaper and faster to install than the industry’s darling, the single-axis tracked array.

Arguments Against

• Single-axis tracked arrays still significantly outperform East-West fixed arrays in energy production on a per-panel basis.
• There is one weather pattern (severe hail) so destructive that I would choose the more expensive single-axis tracked array if I experienced it regularly.

Arguments For

• East-West arrays require less material and labor to install.
• East-West arrays use substantially less land for the same power output.
• As PV modules become cheaper, the economic benefits of East-West arrays get even better.

Conclusion

• If price per watt delivered were the only important factor, East-West arrays would be the clear choice everywhere.
• However, nature is metal, and in some regions, we currently lack sufficiently resilient panel technology to support East-West arrays.

Introduction

Like many technologies, photovoltaic (PV) cells, which convert sunlight into usable energy, started out expensive and uncompetitive. They had to compete in one of the most competitive industrial sectors: the energy sector, which is filled with behemoths like coal, natural gas, and hydropower. On top of that, photovoltaic technology was handicapped by limited daylight hours and an above-average vulnerability to weather conditions.

PV systems did not have it easy, but a collective push from green energy companies, governments, and their citizens willed them into existence. They shared in the renewable energy promise to build an economy powered by sunlight, wind, water, and atoms.

So began the decades-long push to improve module efficiency in every way imaginable: new materials, new PV cell architectures that convert more photons to electrons, and tracking systems that keep modules aligned with the Sun’s position. These trackers gained industry dominance by offsetting their costs through improved efficiency of the PV modules.

Today’s story examines what happens when the very components being optimized, the PV modules themselves, become cheaper than the technologies designed to improve their efficiency.

Tracking technologies are moderately complex, requiring actuators, sensors, stronger structures, and additional energy storage. These technologies fall into two camps: single-axis and dual-axis systems. Single-axis trackers handle intra-day solar movements, following the Sun as it rises in the East and sets in the West. Dual-axis trackers adjust for the array’s latitude, the Sun’s intra-day movement, and inter-day movement, due to Earth’s orbit 1. Unlike single-axis trackers, dual-axis solutions haven’t seen mass adoption due to more complicated mechanisms and increased spacing requirements between rows to prevent shading. For this reason, we’ll focus solely on the costs and benefits of the single-axis tracker, the current king, if you will, and its relation to its much simpler cousin, Earl. Ugh… I mean the East-West array.

The Rapid Decline in PV Module Prices

Why ask this question? When I took over the solar and power systems at Terraform Industries, I was asked to architect a photovoltaic system that capitalized on one of the most remarkable learning curves in modern history: a 44% cost reduction for every doubling of cumulative PV module production. The East-West array was a clear winner, despite not being highly adopted in the United States. I began to investigate why, and this article is a byproduct of this effort.

Starting with PV module costs, PV benefited from the trifecta of expanding economies of scale, increased competition as more suppliers rushed into the market, and increased supply as high-grade silicon became abundant due to the demand for PV modules and integrated circuits, aka chips. That level of cost reduction is hard to compete with. In the 2010s, the cost of PV modules fell by 85%, and so far this decade, they have fallen an additional 58%. All this in the face of the worldwide epidemic of increased inflation in recent years.

Chart: Module Price Reduction 2010-2025

Table 1: Module Price Reduction 2010-2025

Over the last decade, trackers rose to dominance, and by 2015, they accounted for more than 50% of the cumulative installed capacity for photovoltaic systems. At that point, they had usurped the previous dominant array type, the south-facing fixed array, to become the new king. Today, they continue to dominate newly installed capacity by a wide margin.

As trackers productized and their manufacturing scaled, they achieved massive cost reductions of 85% in the 2010s. But trackers have a price floor due to the large amount of steel, actuators, and sensors they require. In the 2020s, they have seen a 17% reduction in cost, compared to a 58% reduction for PV modules, making trackers an increasingly dominant line item in the overall budget of single-axis arrays.

Chart: Tracker Price Reduction 2010-2025

Table 2: Tracker Price Reduction 2010-2025

One must stop and wonder: are trackers still worth keeping around, as the very modules they were designed to improve in efficiency fall to near cost parity with the tracking systems themselves?

A Technology in Transition?

Photovoltaic technologies are no longer a fledgling idea. They have grown into a robust and scalable technology that delivers increasingly significant amounts of energy. We can compare them to the rise of the computer, the Internet, or mobile.

If we take the internet as an example, there were technologies along the way that filled in initial shortcomings, which were no longer necessary once it scaled. Netflix’s mailing of DVDs made much more sense before the proliferation of fiber and high-speed internet that led to the streaming revolution. I am here today to explore the possibility that trackers may one day be thought of this way, as a stop-gap solution that will no longer be necessary.

Choosing your Characters

Single-axis Tracked Arrays 👑

Single-axis tracking systems are essentially long steel torque tubes, motors, sensors, and algorithms that work together to align the PV module with the incoming direction of sunlight. The most popular of these systems rotates from East to West following the Sun’s position. To support these forces of rotation and resist wind loads, the torque tubes are long and heavy. They mount onto deeply driven piles, piles are to the earth what nails are to wood.

Tracking systems function by minimizing the angle of incidence between the Sun’s position and the direction the PV module is facing, later referred to as the PV module unit vector. Ensuring each PV module receives maximum exposure to sunlight, and expanding the power profile by capturing more energy in the morning and afternoons.

The components of the single-axis tracker

• The foundation holds the array to the Earth, typically using piles, which are large steel posts driven into the ground.
• The fixed structure holds the sensors, the fixed side of the actuator and bearing system, and the tracker power system. Modern systems, like those of Array Technologies or Soltec, can be installed over minimally modified terrain, which significantly reduces the cost and time of ground preparation.

• The bearing assembly affixes the rotating structure to the fixed structure.

<!-- -->

• The actuator drives the rotation of the rotating structure. It can be a linear actuator and lever arm, or a rack and pinion system. These actuators are sized to withstand lift forces from windy days and have sufficient speed to quickly stow when faced with adverse conditions.
• The rotating structure holds all the PV modules of the row at the same angle, including components such as the torque tubes—the main tube, and the purlins—the arms that hold the solar panels.
• The alignment system processes data such as direct irradiance, time, wind speed, weather conditions, and location to determine the correct angle of the modules. It includes a controller, multiple alignment algorithms, and some electro-optical sensors for detecting sunlight intensity.
  • Solar position tracking: aligning the module face with the incoming sunlight, used during clear days.
  • Backtracking: deviating from the solar position to collect more sunlight, such as lying horizontally on cloudy days to capture ambient light or moving to avoid shading from one row to another during early mornings or late afternoons.
  • Stowing: moving the modules to a safe position during an adverse weather event. This includes a near horizontal position on high wind days or away from the direction of hail during a hailstorm.
• The tracker power system (not pictured) contains the energy to drive the actuators of the single-axis trackers, including independent PV modules, charge controllers, and battery packs.

The tracking system’s competitive advantage is to enhance the effectiveness of each solar panel, requiring fewer of them to achieve a certain target power. However, this is achieved at the expense of higher structural complexity, installation times, and land consumption per panel.

East-West Arrays

Despite being a simpler system, East-West solutions are relatively new on the utility-scale solar scene. Previously, PV modules were so expensive that East-West arrays were not economical due to their lower energy collection per panel. However, they eliminate a ton (literally) of material and assembly time from the system. Solutions, such as the Jurchen PEG system, use the PV module as the primary structural component, which removes 90% of the typical structure of other solar array variants.

The array scales in both directions, like a mesh network of panels, held together by clamps and supported by rebar. This network of PV modules doesn’t have the same risk of shading each other, due to low tilt angles, so they can pack 250% more installed power into the same space when compared to a single-axis array.

East-West arrays are simple. They consist of parallel strings of PV modules that are oriented in opposing directions, one facing East and the other West. The current of the whole array is the summation of these string currents, effectively letting East-West arrays capture sunlight from dawn till dusk, similar to a tracked array.

The energy profile of the East-West array depends on its tilt angle. The more aggressive the tilt, the larger the difference in time between the East and West peaks, but the more that one string shades its adjacent neighbor. The ideal angle extends the time of near peak solar irradiance by increasing the time between the East-facing peak and the West-facing peak.

The components of the East-West structure

• The substructure of the Jurchen PEG array is rebar rods driven into the ground, held in by the friction between the rebar and the earth. There is a bottom plate crimped at the base of the rod to prevent the rod from sinking into the earth, and a “top hat” on the rod to clamp the panel onto. That’s about it. Rooftop East-West array structures can be even simpler, using bent sheet metal pieces or custom molded bricks.
• The modules themselves.

These systems require minimal land prep, reduced land usage, shorter installation time, and lower material consumption. The question is, can these savings make up for using more solar panels to achieve power parity with their tracked cousins?

How do these structures compare?

Number of Panels Required For Equivalent Power

It is important to note that the ratio between the size of the single-axis tracker and the East-West solar array varies from location to location, depending on latitude and weather conditions. Understanding this requires simulation to account for the impact of changing sunlight vectors and the effects of weather on array performance. To learn more, check out this supporting article where I step through the math to show the relationship between the angle of the panel and the angle of the incoming sunlight, as well as the underlying reasons why single-axis trackers are more efficient.

The baseline performance we were looking to achieve from each array is 1 megawatt of DC power. We’ll simulate the performance at four locations in the US that differ.

• Denver, Colorado (Cold and Dry Climate)
• Mojave, California (Hot and Dry Climate)
• Buffalo, New York (Cold and Wet Climate)
• Houston, Texas (Hot and Wet Climate)

The assumptions I made during this are as follows:

• The array voltage is 1500 volts, consisting of 27 panels in series.
• Panel size is 550 watts.
• Panels are 21.5% efficient.
• Changing the array size can only be done by adding or subtracting parallel strings.
• The power target for each array is 1 megawatt of DC power, defined as an array that averages 8 megawatt-hours of energy per day across the year. We assume all power generated over 1 megawatt is unusable.

Important concepts before we get started.

The energy available to photovoltaic systems depends on the Sun-Earth relationship and the specific location of the system on Earth’s surface. Before we start, let’s get a couple of the basics of the Earth-Sun relationship down:

• The Earth’s orbit around the Sun is slightly elliptical, leading to variations in the Sun’s position in our sky by a certain hour from day to day. The equation of time corrects for this variance.
• The Earth rotates on a 24-hour schedule.2
• The solar constant measures the power per area (solar irradiance) of sunlight as it hits Earth’s atmosphere: 1,361 Watts per square meter.

The amount of sunlight available is determined by the relationship between the local position on Earth, the orientation of the PV module, and the solar position vector, based on the relative positions of the Earth and the Sun.

• The solar position vector magnitude is the solar constant, with a direction determined by the time of day and the Sun’s declination angle, determined by Earth’s axial tilt and orbit.
• The local position vector extends from Earth’s center to the PV’s position on Earth’s surface, which is determined by its latitude, longitude, and elevation.
• The PV module vector is based on the azimuth and module tilt.

The solar position vector and local position vector both affect the way sunlight appears at any one spot. The latitude of the position and the declination angle determine the elevation and duration during which the sun appears in the sky. A shallower angle, common in winter in the Northern hemisphere, will lead to a lower elevation and a larger percentage of Earth’s 24hr rotation with the Sun appearing below the horizon. This scenario is the exact inverse in the southern hemisphere, with increasingly lower latitudes leading to the Sun appearing increasingly more northward at a lower elevation.

Impact of latitude and declination angle

I first investigated performance impacts based on the photovoltaic installation’s position and panel orientation. Both arrays were installed as 1,200 kilowatts DC for an apples-to-apples comparison.

Interactive: pick a city and a day, then set the module price to see which array is cheaper to build. Open the interactive on the blog.

Switch between cities in the explorer above and what jumps out is the difference in performance based on the latitude of the array’s location. Buffalo, having a higher latitude, has a more southerly Sun and fewer sunlight hours. Meaning, less time in which to capture sunlight, leading to an increased importance in the ability to capture sunlight in the mornings and afternoons, the strong suit of single-axis arrays. Houston has the lowest latitude of this group and hence sees the smallest variance in performance. For all cities, we see the impact of the declination angle, as performance varies from season to season, since this also affects the duration of the sun in the sky.

Performance including weather

To capture real-world solar panel performance, we must account for clouds and adverse weather conditions. I updated the tool to add weather factors and re-ran the simulation at 1,200 kW for each array type.3

The most immediate impact is in the cities with the most cloud cover. Clouds don’t delete all sunlight, but they do they scatter it, turning direct sunlight into indirect or diffuse light. The best way to catch scattered sunlight is to face horizontally, which gives the highest chance of diffuse sunlight to bounce into each respective PV module. East-West arrays are always ready for high diffuse light periods, allowing them to perform relatively well when clouds are overhead. Single-axis trackers will backtrack to the horizontal position to capture diffuse light once they determine that the sky is cloudy.

The high performance of East-West arrays in inclement weather makes performance in places like Mojave look comparatively worse, where the largest differences are still based on the advantages the single-axis tracker has during early morning and late afternoon periods.

Meanwhile, Buffalo goes from being the largest deviation in performance, when we ignore weather, to the smallest when we account for cloudy days.

Sizing the array

A note on installed capacity—arrays that deliver 1,000 kW of power are often not installed at 1,000 kW. The actual size is referred to as installed capacity, which is defined as the total of all PV modules installed in an array. They are used by solar designers and installers to determine the number of materials and labor required for the array installation. The critical dimension, measured in dollars per watt, by which all arrays are evaluated, uses the installed capacity to determine the system’s wattage.

After accounting for weather effects, the goal was to find the PV array configurations that deliver 1000 kilowatt of DC power for both East-West and single-axis tracked systems. This simulation assumes that a power conditioner or inverter is being used, limiting the power to a maximum of 1000 kW of DC power. 4 It is worth noting that typically larger sized East-West arrays will perform better when used in a direct DC connection where more of the middle of the day peak power can be utilized effectively [5]. The sizing calculations ensure that all proposed configurations deliver annual energy yields within 2% of each other across all test locations, allowing for fair comparison of performance between the East-West and single-axis tracked systems.

Table 3: Installed capacities per city.

If we roll up these array sizes into costs, we see that single-axis trackers are quite effective at saving PV module costs. Which looks pretty good for the single-axis array. Using fewer modules to get the same amount of power is a big benefit of these structures. However, these structures also have increased material, labor, and land costs per installed watt when compared to the simple East-West array.

Table 4: Costs of PV modules per array type based on city.

We will explore these costs in the next sections and ultimately answer the question—are PV modules cheap enough today that array types that use more of them, like the East-West array, are cheaper than those that save on module costs, like the single-axis tracked array?

Balance of system material and construction costs

The balance of systems includes all components beyond the PV modules and inverters. The most relevant to today’s article is the structure itself—whether it’s a single-axis tracker or the simpler East-West array structure that we’ve been comparing throughout this article. Other components include fuses, wires, and all other systems that enable an array to function. There are two types of costs in this bucket: material costs and labor costs associated with assembling and installing those systems. Like all other solar costs, they are typically listed in USD per watt terms. The single-axis array is more complicated, resulting in more parts, which in turn leads to a higher cost of both materials and labor for assembly in the field.

Table 5: Costs of the balance of system per watt.

If we plug these numbers into our array sizes per state, we see that East-West arrays are, in fact, cheaper; they consume less material and much less labor in the field. Skilled installer labor is especially in short supply, even with the boosts in wages to be eligible for IRA tax credits. So the benefits extend beyond just the cost savings, by improving the speed of installation, the team can be smaller, leading to less management overhead and increased delivery speed.

Chart: Costs of the balance of system for the East-West and single-axis tracked arrays

Table 6: Costs of the balance of system for the East-West and single-axis tracked arrays

Land Consumption and Prep

East-West arrays may require more PV modules, but they are far more land-efficient. Single-axis tracked arrays must be spaced out row to row to prevent panels from shading one another. Trackers, depending on the manufacturer, can have maximum tilts of 45-60 degrees towards the rising or setting sun, extending the shadow each row can cast based on its height. The superpower of East-West arrays is their moderate tilt angle, ranging from 8 to 15 degrees, which limits their shading to this tilt angle or less.

At dawn or dusk, when the Sun is just above the horizon, the light cast is considered low productivity. Sunlight at angles below 15 degrees is considered nearly uncapturable due to the effects of the ozone layer scattering sunlight. Therefore, the East-West arrays lose very little opportunity to capture light.

The difference in land efficiency is significant, with East-West arrays able to pack in 250% more installed power (at peak) into the same land area compared to their single-axis trackers or southward fixed facing counterparts. In an area where we install 1 MW of DC power using a single-axis tracker, we can install 2.5 MW of DC power using an East-West array. East-West arrays have a lower cost of land and, more importantly, land preparation.

Chart: Land consumption per array type

Table 7: Land consumption per array type

The cost of land varies per state. Solar installations can occur on Bureau of Land Management (BLM) land, converted agricultural land, or M3 heavy industrial land. The most common type is converted farm or pasture land. I’ve listed the costs below.

Chart: Ag land costs per state

Table 8: Ag land costs per state.

In general, the cost of land prep is much larger than that of the land itself. These are the common types of land improvements required for a PV array construction project:

• Land clearing – the demolition and removal of vegetation and structures to clear the land of obstacles.
• Site stripping – the removal of loose topsoil with root structures.
• Rough cutting and grading – the removal of dirt from large hills and the addition of dirt from large divots that exceed the array’s maximum allowable slope angle.
• Road construction – excavation, layering in 3-4 different gravel sizes, grading to have a crown to manage water runoff, and making a ditch to drain water into. Roads are required in most counties to give access to emergency vehicles when fighting fires on the site. Strangely, this is not required for row crop agriculture, despite the fact that corn burns far better than glass and aluminum.

The most competitive and modern solar structures require as little land preparation as possible, typically stripping no more than 5% of the land, and only when absolutely necessary. In general, I provide the costs of land clearing and stripping on a per-acre basis.

The costs of cutting and filling are typically per cubic yard or cubic meter. The low preparation solutions minimize the amount, so I will assume another 5% of the land will need 1 yard of depth cut or filled.

Finally, roads are cost-based on a per-mile or per-km basis. Most counties in the United States require a 20-foot fire access road that supports a fire truck along the entire perimeter of the solar array.

Chart: Land prep costs

Table 9: Land prep costs

The East-West array looks quite attractive when considering land costs, both in total project cost and on a dollar-per-watt-installed basis.

Chart: Total land costs per state and array type

Table 10: Total land costs per state and array type

Environments with challenging weather

Arrays have a design life of 20-30 years, over which they will experience a large variety of weather events. The most significant are lightning strikes, wildfires, high winds, and hail.

Heavy rains don’t harm panels and, in some ways, make them perform better since it’s ultimately just a cloudy day with a free panel cleaning service. There is a risk of scouring, which is the slow erosion of dirt around the piles due to wind and water. Still, this is typically accommodated during the design phase by determining the depth to which the piles or rods will be driven, so that they can survive the maximum scouring. Therefore, survivability in heavy rains is no differentiator between these array structures.

Lightning strikes pose similar challenges to both structures. Both have methods of conducting high current surges to ground. Both rely on surge protection systems in the electrical balance of system to reduce the risk of surges frying individual panels. This, too, does not give us any means of differentiating the system performance.

Designers and system operators mitigate wildfires by clearing the setback distance around the solar array. However, if a wildfire near the array is hot enough on a windy day, those ashes can fly right over the setback distance into the precious PV modules.

Surviving wind, caused by tropical storms, near hurricanes, or tornadoes, is a moderate differentiator of the systems. The East-West arrays perform best due to low angles of attack, resulting in less lift. The small gaps between panels prevent the formation of large vortices on the leading edge of the panel. So while the East-West array has less material affixing it to the ground, it experiences far less lift in high wind scenarios, leading to very high survivability against high winds, whether from Santa Ana winds or a hurricane. One company, Jurchen, claims that its PEG system can survive 180 mph winds, which exceeds the speeds expected in a category 5 hurricane. However, to my knowledge, no real-world testing has verified this claim.

Single-axis tracked arrays are less reliable in extreme wind scenarios, but have added stow features to reduce their angle of attack during high wind days. The individual rows of the single-axis arrays are prone to more aerodynamic loads, leading to fluttering, which can change the panel’s tilt angle, create ground vortices, twist the torque tube, and, in the most severe conditions, eventually cause array collapse. Check this out for a more in-depth understanding.

Figure 15: Simulated wind flow analysis on the single-axis tracker

Modern solar tracker designers understand these risks and have incorporated aeroelastic loads on the tracker, leading to stiffer torque tubes or the addition of dampers. These upgrades add to the system’s costs, but allow single-axis tracked arrays to withstand up to 120 mph winds, such as those seen in a category 2 hurricane, and all winds expected under normal conditions.

Single-axis tracked arrays excel in one type of weather pattern, however: large hailstorms. Hail has been a problem in the solar industry since its inception, leading to huge insurance payouts, especially in Texas, such as $70 million for the Midway Solar Project in 2019 and $50 million for the Fighting Jays solar project in 2024.

Texas may be especially at risk today, as it is located in Tornado Alley and has some of the worst hailstorms in the country, and due to its high solar density. A storm with hail larger than 1.8 inches (45mm) used to be considered a once-in-five-hundred-years storm, but such storms are becoming even more common due to the impacts of climate change, with at least two occurring in the last five years.

To combat this, single-axis tracker systems have implemented hailstorm stow protocols that reduce impact energy. During a hailstorm, the PV modules are oriented facing away from the hail’s angle of attack based on the wind speed and direction, turning a direct impact into a glancing blow. Three solar plants located near the Fighting Jay project, which experienced the once-in-500-years storm, survived the same event with minimal damage. The limited losses they did incur were confined to tracker systems that were either not equipped with hail stow features or were disabled at the time.

Trackers are not the only way to address this issue; the industry has demanded more from its manufacturers. For the past couple of years, PV modules have required testing to demonstrate their resilience to hail up to 1 inch in diameter under UL-61730, with additional tests defined (but not mandated) for hail up to 3 inches in diameter.

The dominant PV module in the US is a bifacial panel with 2 mm of tempered glass on both sides, supported by an aluminum frame. The modules are stiff and do relatively poorly when pelted repeatedly with balls of ice. Lab testing has shown that these panels typically break when impacted by ice balls of 1.8 to 2.4 inches in diameter, hitting them dead on, which represents a 12 to 30 Joule impact, respectively. The ability to withstand hail up to that size ensures these panels are safe for all array configurations in areas such as the southwest and northwest.

New PV modules are being marketed with higher impact ratings due to an increased thickness of the glass sheet, from 2.0 mm to 3.2 mm or even 4.0 mm, which should enhance survivability enough for the northeast and southeast in any array configuration. Future improvements hope to reach the survivability levels needed for Texas and the rest of the Midwest.

The hail stow feature is a relatively new addition to single-axis tracked arrays, but it constitutes one of the largest moats this configuration has against the faster, simpler, and cheaper East-West arrays in areas where hail is likely. There are instances of PV array operators reducing their insurance premiums by up to 72% through the construction of hail-resistant systems, single-axis tracked solutions with a validated hail-stow function, and higher impact-rated panels. This can save a lot of money when insurance costs average 1.5 cents per watt installed per year. Since the example listed went to significant effort to negotiate these insurance savings, it would be incorrect to assume all single-axis arrays automatically qualify for reduced premiums with hail-resistant systems.

What happens as modules get cheaper?

As PV modules continue to fall, the situation for single-axis tracked arrays only gets worse. And I’m betting on this with Terraform Industries, where we aim to leverage cheap solar to manufacture cheap natural gas and other products. So let’s find the crossover point, the cost of the PV module, where the East-West array reaches cost parity with the single-axis tracked array.

Chart: Chart

Table 11: Array cost per module price ($/W).

Chart: The module cost where East-West array costs fall below the single-axis array

Table 12: The module cost where East-West array costs fall below the single-axis array.

From just an economic point of view, modules have gotten cheap enough that it will be advantageous to build an East-West array anywhere.

Conclusion

In summary, the single-axis tracker configuration’s higher performance per panel is true, but the cost savings associated with that performance efficiency do not outweigh the additional costs in parts, labor, and land.

Table 13: Final cost comparisons between the East-West array and the single-axis tracked array

What surprised me is the impact of the weather. It has created a really defensible moat for the single-axis tracked array in the areas that can experience large hail. Today’s 10-20% savings don’t pencil out when you risk losing a significant portion of the array. Stronger modules are the only path forward I see for the East-West array in these areas.

• If I were building photovoltaic arrays in Texas, I would take that threat seriously and be willing to pay the higher premium for component costs to extend the array’s survivability.
• If I were building solar anywhere on the West Coast, I would be hyper-focused on East-West arrays using the cheapest panel options available.
• If I were building on the East Coast, I would seek out high-impact-rated PV modules and build East-West arrays. High-impact modules would have a 10-20 percent premium on the module price, but would enable the use of the East-West array, saving cost on the overall project.

1 The Earth is tilted on its axis by 23.45 degrees, relative to its orbit around the Sun. Most believe this tilt was caused by the impact between Theia and our young Earth, the same impact that set the stage for the Moon’s formation. In the Northern Hemisphere, the tilt provides longer days around the July Solstice and longer nights around the December Solstice. This tilt leads to the Northern Hemisphere experiencing increased amounts of sunlight between March and September, while the Southern Hemisphere experiences more sunlight from September to March. On average, the Southern Hemisphere will experience more radiation from the Sun, resulting in increased thermal energy in the southern regions. This increased thermal energy, combined with the higher percentage of the south being covered in water, which stores thermal energy, are the primary causes of the higher average temperatures in the south compared to the Northern Hemisphere.

If we stand in the same location across the entire year, we will see the Sun shift North and South based on the Earth’s axial tilt and its orbit. If you are above the Tropic of Cancer in the Northern hemisphere, the Sun will be to our South throughout the year, but it will be more overhead during Summer versus more Southward during winter. If you are between the Equator and the Tropic of Cancer, then the Sun may appear North of your position during the peak of Summer. The same goes with the Southern Hemisphere, if you are south of the Tropic of Capricorn, then the Sun will always be to your North.

Your latitude is important. If your latitude is 35 degrees North, then the Sun’s average position across the whole year will be 35 Degrees South. If you are between the Tropic of Cancer and the Tropic of Capricorn, then at some point during the year, the Sun will be directly overhead. If you are North of the Tropic of Cancer or South of the Tropic of Capricorn, then the Sun will never be directly overhead.

2 Actually, our rotation takes 23 hours, 56 minutes, and 4 seconds, but since the Earth is orbiting the Sun, it needs to rotate a little further to reach the same position relative to the Sun. Aka 1/365 of each day is needed to adjust for the Earth’s orbit around the Sun.

3 Notes on the simulation:

The best source of weather data is a study conducted by NREL in 2006, which covers 5-minute incremental performance from arrays across the continental United States. One limitation: most arrays at that time were southward-facing fixed-tilt arrays, which means most data is from southward-facing fixed-tilt arrays. This isn’t ideal for our comparison of the East-West and single-axis tilt mechanisms.

I normalized the performance by capturing a week’s worth of data around each point, then using the difference between the maximum and average performances to determine the weather effect on direct irradiance capture, but abstracting the array size and PV module efficiency. This weather effect was then applied to the simulation of the different array types at their respective scales.

4 Direct DC connection

If we can use all of the DC power, like the direct-connected system we use here at Terraform Industries, larger East-West arrays will be able to deliver more power overall compared to their single-axis counterparts.

For example, if we use the same array power levels of 1000 kW for a single-axis array and 350 kW for an East-West array in Mojave, but utilize the power in the middle of the day, the output of the East-West array increases drastically.

Last year, we made cheap pipeline grade natural gas from sunlight and air and expanded our hydrocarbon fuel road map to include methanol, a versatile liquid fuel and chemical precursor for practically every other kind of oil-derived chemical on the market. Unlimited synthetic methane and methanol underpinning global energy supply is a good start, but it is now abundantly clear that we have a strategic need to re-localize our supply chain for all critical minerals.

We wanted to understand the effect of geography on fuel production cost, so we performed a series of detailed simulations in a range of locations. These simulations allowed us to calculate, for any given location, solar PV, battery, and load cost, the optimal array and battery size, asset utilization, and overall effective power costs. Each simulation took real world solar PV production data at 5 minute intervals over an entire year. 

[Edit August 2025: This analysis has since been replicated by Kevin Do here and here and Rik de Kort here. Thank you gentlemen!]

One key question for equipment operators considering an off grid power system using only solar and batteries is: How much will power really cost? It depends on how sunny and seasonal your location is, but it also depends on how much you want to operate your equipment in order to maximize revenue per dollar spent. This graph solves that problem, automatically calculating the optimal asset utilization for any asset cost class, and displaying the corresponding cost of power. 

More formally, the graph shows the optimized effective amortized power cost per MWh consumed by a load characterized by its capex measured in $/MW. For example, an inexpensive electric kettle is about $20/kW, while an AI training datacenter is more like $50,000/kW. Utilization is given by the saturation of the data point, and we’ve selected six locations for a representative range of weather and seasonal conditions. 

Power system costs are baselined at $100/kW for solar PV and $100/kWh for the batteries, which we think are representative of near future conditions. For reference, the cheapest solar modules are currently $82/kW and cheapest batteries are $94/kWh. The results scale smoothly, so if you strongly believe that the true numbers are $0.20/W and $200/kWh, simply multiply the axes values by two – but it doesn’t change our conclusions. 

There are several salient features. 

First, in nearly all places and for nearly all load types, solar+batteries are cheaper than any other potential power source, replacing the concept of an expensive, undesirable “green premium” with a “green dividend,” a rare win-win for using advanced new power technologies. We have always believed that the revealed preference of the market is that less polluting technology is enthusiastically adopted, if and only if it also saves the customer money! 

The second is that power costs do increase as we drive to a large number of “9s” of reliability towards the right edge of the graph, but even for loads as expensive as AI training centers, the optimal utilization is probably closer to 99.9% in terms of weights updated per dollar spent. In practice, rather than the datacenter going dark for 8 hours a year, 99.9% utilization means some partial throttling in winter over longer time periods. 

Third, there is a clear bifurcation between battery supported loads over $2000/kW that run at utilization greater than 75%, and diurnal, battery-free solar loads under $400/kW that run at utilizations below 30%. 

We’ve seen recent studies on off grid solar power and the cost of reliability with solar in general but as far as we’re aware, ours is the first study to comprehensively demonstrate the tremendous opportunity in low capex intermittency-friendly matched loads. We intuitively understood this for synthetic fuel, but fully generalizing this for all consumers of energy in all geographies was… non-trivial. 

WIth the benefit of this insight, we can redraw the graph above with some directional commentary. We now have conviction that low capex, intermittency-friendly loads can unlock a roughly 5x bigger, greener dividend. Green is the color of money, after all!

Interactive: set solar and battery prices and see the optimal power cost and utilization for every kind of load. Open the interactive on the blog.

But what are these loads? What do they do? What do they look like? They must be relatively small, power intensive, cheap, and intermittency-friendly. Their value add is due to privileged access to cheap power. But colocating some miniaturized chemical plant, factory, or other business right next to a solar array is not without its challenges, so what processes and products are worth the hassle?

Let’s take a look at the energy landscape of our industrial economy. 

This graph shows which industries derive their value from consuming energy (bottom right) vs some other high value-add operation per unit energy (top left). They are materials, utilities, and oil and gas. The primary production layer of industry, which is concerned with processing naturally-occurring raw materials into much higher value intermediate commodity products. 

For example, of the ~6 processes involved in making aluminum cans from red mud, 60% of the total energy consumption occurs just in the Hall-Héroult process, in which the refined alumina is electrolyzed. Not coincidentally, this is also the step with the single largest value add, so exploiting cheap energy reduces aluminum production cost, improving access for buyers worldwide.

This insight prompts the following recipe for processes that Terraform’s unique capabilities provide exclusive economic access to. 

• They have relatively generic raw inputs that are readily available: sunlight, air, water, various commonly occurring rocks and minerals.
• They adapt high maturity existing industrial processes to consume 2-3x more electricity in order to get to lower capex, cutting down from $2000/kW to $200/kW.
• They adapt existing industrial processes to be intermittency-friendly, that is, capable of throttling up and down to match variations in available solar power due to weather and the day/night cycle.
• They adapt existing industrial processes to operate at the 1 MW scale immediately adjacent to the array, which is much smaller than typical plants. Miniaturization enables cost control through manufacturing rather than construction. It also helps with intermittency, since it’s easier to turn a small ship than a big ship.
• The output products are relatively valuable, storable, stable, and now contain most of their embodied energy. We are not advocating relocating the entire industrial stack to endless fields of solar panels! But Terraform’s products will require relatively little additional energy input in high value-add secondary manufacturing steps, which will occur in more conventional centralized factories with less sensitivity to power costs relative to other factors of production. Most of our value add is occurring in the energy intensive processing step using Terraform’s proprietary technology.

Let’s examine the set of energy-intensive primary products for compatibility with this recipe. I’ll include the chemical formula of the most energy intensive steps in production.

• Hydrogen production (2 H₂O → O₂ + 2 H₂)
• Synthetic natural gas (2 H₂O → O₂ + 2 H₂, CO2 + 4 H₂ → CH₄ + 2 H₂O)
• Synthetic methanol (2 H₂O → O₂ + 2 H₂, CO2 + 3 H₂ → CH₃OH + H₂O)
• Synthetic coke/graphite/graphene (pure carbon) (CO₂ -> C + O₂)
• Ammonia (6 H₂O + 2 N₂ → 4 NH₃ + 3 O₂)
• Cement (CaCO₃ → CaO + CO₂)
• Steel (2 Fe₂O₃ → 4 Fe + 3 O₂)
• Non-ferrous metals (eg 2 Al₂O₃ → 4 Al + 3 O₂)
• Desalination (something like Na⁺.2X H₂O → Na⁺.X H₂O + X H₂O)
• Exotics (secret projects)

There are some other highly energy intensive primary industries that do not match this recipe, either because their energy inputs (ie. low grade heat) are cheap to store or their value creation is not sensitive to energy cost.

• Paper. Energy consumption is mostly heat, which is easy to store, so there’s no benefit from intermittency-compatibility. That said, it is possible to make a water-proof paper substitute from CaCO3 and polyethylene. 
• Glass. Similarly to paper, there’s no energy intensive chemical reduction step occurring.
• Data center operations. While energy intensive, the load is so expensive there’s no benefit to operating intermittently and the product is relatively insensitive to power cost, provided that there is power supply.

The energy intensive step in common is chemical reduction, or ripping the oxygen atoms off, which is the inverse of combustion. Cement is an exception, but our direct air capture process performs calcination (ripping CO₂ molecules off of limestone) so we already have a mechanism for this highly energetic process. Desalination is another exception, but in this case we’re ripping hydrated salt ions off respectively the oxygen and hydrogen parts of water, and keeping the water, and the business model is a bit different. 

These products underpin our entire civilization. They consume a prodigious quantity of energy, usually in the form of coal. Or, in the case of coke, literally are coal. In all cases, these are commodity products, with zero market risk and steadily rising consumption. Their price and scarcity are fundamental constraints to the progress of the human condition.

We see a path to cost parity with all existing production processes in the US by 2031, often sooner. Terraform’s proudly energy intensive philosophy positions these processes to reap ongoing and substantial production cost advantages courtesy of solar’s continuing price declines.

Since the beginning of the industrial revolution, growth has been constrained by energy availability, and primary energy production was the imperative. With the solar PV revolution, we have an inversion of this paradigm, similar to the effect described by Aggregation Theory. The new imperative is to find better ways to consume this unexpected bounty in productive ways.

To occupy the solar cost sweet spot at 30% utilization with $200/kW capex loads obliterating oxides forever. 

To conquer the primary energy consumption layer of our entire civilization.

Reflecting this strong conviction, Terraform Industries is ramping up maximally aggressive development efforts for solar-adapted processes to make all these products. Terraform is the place where the world’s most far sighted visionaries come to build for a millennium of material abundance. Our pioneering work in this sector has made us the natural leader and ideal partner to develop this portfolio and allocate capital across each of these verticals. 

Here’s a non-exhaustive list of current career opportunities at Terraform. 

Technical Chief of Staff
Technical Recruiter

Electrolyzer System Lead
Electrolyzer Production Mechanical Engineer
Electrolyzer Product Technician
Electrolyzer Test Technician

Direct Air Capture System Lead
Direct Air Capture Cyclonic Calciner Engineer
Direct Air Capture Mechanical Engineer
Direct Air Capture Technician

Reactor Injection System Lead
Reactor Injection Compressor Engineer
Reactor Injection Controls Electromechanical Engineer
Reactor Injection Technician

Methane Reactor Lead Engineer
Methane Reactor Design Engineer
Methane Reactor Mechanical Technician

Methanol Production Lead Engineer
Methanol Reactor Chemical Engineer
Methanol Reactor Design Engineer
Methanol Reactor Mechanical Technician

Photovoltaic Systems Lead Engineer

Ops/GSD Lead
Project Development and Construction Lead
Terraform Sales Lead
California Permitting Lead

Coke Project Lead Engineer
Coke Mechanical Technician

Ammonia Project Lead Engineer
Ammonia Reactor Chemical Engineer
Ammonia Reactor Mechanical Engineer
Ammonia Reactor Mechanical Technician

Cement Production Lead Engineer
Cyclonic Systems Mechanical Engineer
Cement Process Technician

Steel Production Lead Engineer
Steel Production Metallurgist
Flash Ironmaking Engineer
Electric Arc Furnace Design Engineer
Steel Production Technician

Silicon Production Lead Engineer

Aluminum Production Lead Engineer

Titanium Production Lead Engineer

Generic Mineral Inputs Development Lead Engineer

Desalination Lead Engineer

Secret Projects Lead Engineer

The Real Blueprint for Hard Tech Success

Hard tech startups occupy the critical frontier of industrial progress, from reshoring production to securing strategic access to the energy and materials needed for a more abundant future. Yet promising industrial innovations stall out before reaching commercialization at an unacceptably higher rate than the broader startup ecosystem. 

It doesn’t have to be this way! The “Valley of Death” refers more to a yawning chasm of excuses than a diagnosis. Yes, many capital-intensive startups flounder and struggle with the higher upfront investment required to reach commercial viability, but plenty have also succeeded. The question isn’t just why startups fail, but why some manage to win—and how we can engineer more victories. 

Those that succeed are focused from day one on developing technology and building companies that deliver real economic value per dollar spent, not by leaning on government crutches, rich subsidies, or vague promises of economies of scale. They develop and build rapidly and iteratively with a maniacal focus on the cash they spend and an unmatched urgency to generate revenue. 

At Terraform Industries, we are building a company that is deliberately structured to overcome these obstacles from the outset. Our mission, to produce cost-effective, carbon-neutral synthetic fuels, requires a fundamentally different approach to scale, capital deployment, and industrial execution. If we were to simply cargo cult other companies, we would not succeed. Instead we draw inspiration from past industrial giants who mastered the art of resource-constrained execution to help us exploit the unexpected bounty of cheap solar power.

Our rapid iteration, operational efficiency, and disciplined capital allocation mirror Henry Kaiser’s revolutionary wartime shipbuilding, with a vertically integrated supply chain and modular construction techniques. Our lean, high-speed development cycles allow us to make real progress without being bogged down by bureaucracy or inefficiency, much like Kelly Johnson’s Skunk Works, which used to be just across the road from our location in the Burbank castle!

Terraform is always eager to develop top talent, so if you are super smart and like the sound of rebuilding the entire foundation of our industrial civilization, shoot us a compelling one pager.

Energy, Growth, and Industrial Resurgence Matter

Energy abundance fuels human prosperity. The ability to produce materials, transport goods, and manufacture at scale underpins economic strength and national security. The U.S. must remain at the forefront of industrial capability, and so it must master the art of scaling transformative 21st century industrial technologies without succumbing to the waste and inefficiencies that doom so many promising ventures.

Yet, energy alone is not enough. The ability to harness it effectively—through superior industrial processes, vertically integrated supply chains, and cost-efficient manufacturing—is what determines economic leadership. Countries and companies that fail to innovate in these areas become dependent on fragile global markets and unreliable (or outright hostile) counterparties, while those that take control of their energy and production ecosystems secure strategic and economic independence. Terraform Industries is ensuring that America leads this new industrial age, not by relying on subsidies or regulations, but by building a scalable, market-driven model that works.

Hope is Not a Scaling Strategy

When faced with the failure of Cleantech 1.0/2.0 and the failure of the VC start up model to generalize much beyond SaaS, industry and government leaders will typically bemoan the “Valley of Death”. The Valley of Death is a purported capital-starved gap between the pilot-scale development of some new process, and the industrial scale investment necessary to see meaningful returns on investment. Instead of asking for funding to go from 0 to 1, hard-tech startups are often asking investors to trust them to go from 0 straight to 100 in a single step. Without that intermediate risk-happy and capital-flush customer and investor coming into perfect alignment, we are told, it’s nearly impossible to get the new industrial processes to escape velocity and so we are stuck with legacy plants and processes that somehow our grandparents were able to build with slide rules and high interest rates, and that just happen to make money.

These arguments rely too heavily on purported “economies of scale.” In this scenario, an academically-derived process may be demonstrated at pilot scale, retiring much of the perceived technical risk, while still remaining industrially immature and still being far away from positive unit economics. Perhaps this value deficit can be made up for by a purported willingness of the market to bear a massive green premium, despite this almost never occurring in practice.

Instead of charting a course that allows them to take smaller steps, rapidly iterate, and demonstrate measurable progress, we tend to see most companies instead devoting resources to extended solicitation of massive investments with technology and teams that are simply not prepared for it. While $10M can prove the science works, at least with the help of a team of PhDs, commercial viability may still be an afterthought.

We know from experience that, for example, chip and car manufacturing achieve low unit costs through enormous, capital intensive mass production factories, while container ships achieve low unit costs through enormous, capital intensive ships. It’s not too hard to cook up an Excel model that shows that the piloted process will magically stop bleeding red ink after someone else puts in a billion dollars. History is replete with examples of megaprojects that failed to deliver for all kinds of exciting reasons. 

What is going on here? 

While it is true that some costs can go down with scale, it is equally true that other costs go up. It’s also true that new processes always have unexpected sources of cost growth during scale up, which is why conventional wisdom, and historical precedent, calls for scaling up after demonstrating positive unit economics. 

Hope is not a strategy. At Terraform, we don’t believe customers are prepared to part with their hard-earned money to purchase a steep green premium and that is resoundingly reinforced in the market today. History shows, however, that we will have no difficulty selling a green dividend. Superior value rules. This means that any new process needs to actually deliver value to its customers, ideally from day one. 

Ongoing declines in solar PV module costs are the single most extreme value generation opportunity that has ever occurred to humans, and business models that exploit this steady progression need to find some way to get the unit economics the right way up at the beginning. It turns out that capital has a finite appetite for billion dollar experiments.

Superior Value Rules the Day

The hard tech startup landscape is littered with companies that raised millions, built promising pilots, and ultimately failed to cross the chasm from ideation to commercial viability. Successful hard tech ventures don’t wait for perfect conditions or unlimited capital. They engineer paths to market that generate revenue while proving their technology and hacking their way to success. The difference between breakthrough technologies that change the world and those that remain perpetually “promising” often comes down to strategic choices made in the earliest stages that separate the commercial projects from the science projects:

• Find a Beachhead Market: Don’t aim for a massive plant before proving the economics at a smaller, profitable scale. If the tech is wildly out of the money at pilot scale, iterate until it isn’t. Identify niche applications where your technology provides immediate value. SpaceX, for instance, didn’t start with Mars colonization; it began with Falcon 1 and Falcon 9, servicing smaller satellite launches that provided revenue to fund larger ambitions. Terraform Industries follows a similar approach with modular synthetic fuel production units to generate revenue at smaller scales before expanding.

<!-- -->

• Generate Cash Flow Early: The belief that hard tech requires billions in upfront capital before reaching ramen profitability is flawed. The companies that succeed start by dominating a valuable niche and scaling from there. Terraform follows this principle by ensuring that each deployment of synthetic fuel production generates revenue, avoiding the trap of infinite fundraising cycles.

<!-- -->

• Avoid the Capital Trap: More funding is not always the answer, nor is it an excuse for why you can’t otherwise succeed. Overcapitalization leads to bloated company structures, consumption of resources by inefficient internal interfaces, slower iteration cycles, and a lack of urgency. Stay lean, and let demonstrated success drive capital inflow.

<!-- -->

• Prove Real Unit Economics: Fancy Excel models don’t substitute for actual cost reductions. If a process isn’t cost-competitive in the pilot phase, it likely won’t be at scale either. Companies that master efficiency at small scale and grow sustainably avoid the common scaling trap.

Execution Over Excuses

Winning in hard tech isn’t about waiting for perfect conditions—it’s about psychotic determination to execute, iterate, and provide value every step of the way. The companies that have succeeded in reshaping industries didn’t do so by relying on external lifelines; they did it through relentless focus, superior execution, and an uncompromising commitment to delivering real value.

Too many hard tech ventures fall into the trap of believing that capital, policy, or luck will determine their fate. But history shows that the companies that dominate do so because they build cultures of engineering excellence, operating efficiency, and financial discipline. This is not about creating grand visions without a foundation—it’s about making meaningful, measurable progress every day, ensuring that every dollar spent generates tangible returns.

Terraform Industries is operating with these principles at its core. Just look at our website! Our approach to synthetic fuel production is not dependent on subsidies, speculative capital, or vague promises of future economies of scale. Instead, we are methodically proving our technology at small scale, delivering real economic returns, and scaling only when the economics justify it. We are not just theorizing about the future of energy—we are building it and we need a lot more people to help us get there.

We have opportunities in all aspects of our business, including mechanical engineers, electrical engineers, technicians, business specialists. To work with us, shoot us a compelling one pager!

I’m Chandhana Sathishkumar and I have an odd curiosity for plastic. I was nerd sniped randomly at a hackathon in 2022, my team’s goal was to make a biomaterial. It was very rushed but our pitch deck received a grant from Shruti Rajagopalan at Emergent Ventures, precipitating a set of unlikely coincidences that landed me at an alchemy castle called Terraform Industries. 

Plastics are everywhere. They’re wrapped around our food, woven into our clothes, stitched into medical devices, and scattered across grocery store shelves in bright, crinkly packaging. We’ve engineered thousands of variations making them arguably the most versatile material class we have. But here’s the thing: they don’t just disappear. They break down into smaller and smaller pieces, slipping into our food, our water, even our bloodstreams…with no reliable way out. And while that sounds like a sci-fi horror plot, it’s real. 

We’ve gotten better at medicine, we’ve beaten a bunch of infectious diseases, and yet, somehow, hormonal issues are on the rise. Fertility rates are dropping worldwide. It makes you wonder, are we poisoning ourselves without realizing it? It’s easy enough to test for plastic additives in your system, and let’s just say, no one has ever been excited by their results.

In mid-2024, microplastics and endocrine-disrupting chemicals (EDCs) were all over the news again. By then, I was at university in the US, doing my best to stay off social media, but people kept sending me posts about plastics. One of them caught my attention, and before I knew it, I was deep in conversations with Yaroslav Shipilov and Nat Friedman. Initially we weren’t sure whether Nat was going to support it, but it didn’t matter. Yaroslav and I hit it off and we decided we were going to work on it together no matter what. Don’t wait for external validation once you’ve reached conviction. But he did end up supporting it and this led to the creation of PlasticList along with Richel Murata and Claire Zhang. We spent months digging into the plastic content of foods in the Bay Area–something that, somehow, hadn’t been done before. 

But the real issue with plastics (and much of modern chemistry) goes deeper than waste. It’s about where we get our carbon from in the first place. Before joining TI, I hadn’t thought much about the chemical industry beyond plastics and pollution. I didn’t realize just how much of modern life relies on chemistry. Converting carbon into polymers isn’t inherently harmful. Plants have been doing this for millions of years. But we are taking carbon from deep within the Earth and putting it into the atmosphere with no plan for return, comparable to exactly what we’re doing with EDCs and our bodies.

When I first met Casey he was helping us PlasticListers understand the intricacies of a Gas Chromatography column. What was meant to be a brief encounter turned into awe at how someone knew so much about…so much. A few months later, PlasticList wrapped up, and I wondered what to do next. I had taken a gap year from school and had developed a desire to learn to build physical things. A shot-in-the-dark application, which turned into an interview, which against expectations turned into an offer. I wasn’t remotely qualified in hardware, and I knew it. But I wanted to learn. And somehow, that was enough to get my foot in the door.

Casey believes that everyone starts from zero, and he was willing to take a chance on me. So, I packed up my life in Palo Alto, crammed everything I owned into a few bags, and moved to Burbank, California, for a hardware internship.

Burbank, Los Angeles.

The work was tough though. When you jump into something completely new, step one is just mapping out everything you don’t know. Terraform strongly encourages (and gently hazes) new MechE interns with a day-one project. Mine? Installing a massive 20-foot-wide whiteboard for the magnetic kanban cards. Simple enough right? Except I had never used a power drill or jigsaw before, and every scene of Final Destination flashed in front of me as I picked them up. It was frustrating watching the rest of the team move effortlessly while I struggled to figure out which way the bit was supposed to turn. 

Whiteboard – 5, Final Destination – 0.

I thought the rest of the team would be tough on me, but they turned out to be some of the kindest and most helpful people I’ve met. I failed, learned, failed again, and kept going. It was frustrating at times, but every mistake pushed me forward. At one point, Casey told me something that stuck—learning at the fastest rate feels a lot like failing all the time. He wasn’t wrong.

This cycle repeated itself: struggle, ask questions, fail, and then eventually figure it out. Last week, I finally digested all the information relayed to me through months of whiteboard sessions with the team—messing up, figuring things out, reaching dead ends—integrated everything I had understood about different parts of the electrolyzer and learned to assemble a few cells from scratch. Although physically small, it was a mighty piece of chemistry unfolding before me. As the Binary Gas Analyzer numbers ticked up and indicated how much hydrogen and oxygen gases were being made and the electrolyte bubbled, I couldn’t help but gawk at how hard, and yet how simple, it was to make that reaction happen.

It’s pretty wild to work at Terraform. Every day, normal seeming people are going about their business just casually transforming air and water into fire at a castle! I knew all this stuff on a theoretical level since ninth grade, but now I know what it means to convince atoms to go where you want. 

Six months later, I’ve learned enough to hold my own in my area, but more importantly I’ve gained the confidence to ask questions without hesitation, to turn to my trusted advisor Ms. Claude more effectively and to know exactly who to text when I hit a wall. It has rewired how I think about the physical world, how things fit together, how they break, and most importantly, how to fix them.

Five year old Chandhana’s wish to be a princess in a castle came true in a different way than expected.

Other highlights? Hitting team milestones. Terraform has three core teams—direct air capture, hydrogen electrolysis, and the Sabatier reactor. Every month, they set themselves sprint goals that feel borderline impossible, turning it into a kind of competition. When a team pulls it off, the lead hands out prop cash, which we slap onto our desks as a reminder that we do miracles on a regular Tuesday.

One unexpected twist is I might have found a long-lost cousin? Swetha, another intern, was on her third internship when we realized our families are from the same village in India. The odds are slim, but we just sent in our 23andme tests so stay tuned.

Casey’s whiteboard sessions felt like a fever dream of ducks, springs, jets, unicycles, and ship crashes. But somehow, it all made sense by the end. Dave threw me into a world of fittings, and finding leaks turned out to be way more detective work than I expected. Russ had me flipping through entire textbooks just on nuts and bolts. Lucie made chemistry safety click. Stephanie helped design a pressure system out of janky bike pumps and plastic buckets. Xavier and Jon handed me wires, tools, and just enough patience to make sense of circuits. Trevor and Sarine helped piece together the electrolyzer, turning a pile of parts into a working machine.

Don’t ever talk to me or my son or my son’s son’s son again.

It’s funny how a simple equation like pV = nRT ends up shaping real-world engineering. I used to see it as just numbers on a page, but now it’s everywhere; in test setups, in electrolyzer challenges, in the messy process of taking a scientific concept and making it actually work. When the temperature changed in the stack and pressure spiked, purging hydrogen and oxygen accidentally pulled a vacuum in the bubblers. It was one of those moments where a textbook equation suddenly became very real. I wouldn’t have expected a high school equation to be at the center of a real problem a company is trying to solve. But if we crack it, we move the technology forward. And that’s pretty wild.

Most things really are a lot simpler yet a lot harder than they seem.

The teams flags and a functional reminder of my week one escapade.

I’ve been ridiculously lucky to spend the past six months at Terraform, working alongside some of the best people I’ve ever met. My quest continues beyond the castle into the realm of plastics.

By Johanness Alarik Nilsson – Sat Mar  1 02:01:27 AM UTC 2025

❝ You have to act as if it were possible to radically transform the world. 

And you have to do it all the time.❞ 

—Angela Davis.

The Viennese ballroom remains gilded, but the chandeliers flicker. The music of a bygone era still plays, yet the scent of burning drapes lingers in the air. That’s what it feels like to stand in an outsourced America—watching a nation waltz through a grand hall as the walls crumble, and the air thickens with smoke and flame.

Once, I was part of the old dance, a polymath software engineer who spent years optimizing databases, APIs and cloud bound deployments. I spent my working days chasing the promise of infinite scalability.  But the illusion shattered, and I stepped away from the ballroom and into the workshop. Now, I’m a Senior Mechanical Technician at Terraform Industries, an ambitious energy startup building the future from the smoldering ashes of the old world.

For decades, I bought into the software-first dream—the idea that code could solve anything, that problems existed in abstraction, that real progress came from elegant algorithms and clever cloud architectures. My peers and I were sold the fantasy of a world that didn’t need factories, turbines, or steel foundries—just cloud computing, VC-backed disruption, and ever-expanding digital platforms. The Instagram Nirvana. But when I looked around, I saw that the infrastructure holding everything together was crumbling. If you’ve ever driven the freeways winding their way through the heart of the Los Angeles Megalopolis, you’ve felt it.

Once these concrete ribbons were marvels of the 1950s American Dream. Masterpieces of massive civil engineering, now just battered, neglected, and cratered with potholes that rattle both my spirit and the suspension of my stock, unrestored 1962 Mercedes-Benz with its 4 wheel disc brakes, independent suspension, and 40mpg consumption on home-made bio fuels. 

The American decay isn’t just cosmetic; it’s a symptom of a nation coasting on its past achievements while failing to maintain the motive forces that keep it moving. Energy grids buckle under increasing demand, supply chains suffer, and the physical world reveals itself as the unshakable foundation of all things digital falls away. In Southern California, this failure isn’t theoretical—it’s apocalyptic. A power grid stretched beyond capacity, neglected and vulnerable, has ignited disaster after disaster. The same complacency and corruption that led to rolling blackouts and unreliable energy delivery has also fueled wildfires that consumed entire communities, reducing thousands of homes to ash and turning the city’s skyline into a dystopian haze. When aging transmission lines, battered by deferred maintenance and corporate greed, spark infernos across drought-ravaged terrain, the result is a hellscape born not just of climate change but of a system unwilling to address the rot at its core. We are not merely watching the collapse—we are inhaling it, sifting through its embers, driving through its charred remains.

It’s not that I ever thought hardware was unimportant—far from it. I built and ran a successful side hustle that grew into an internationally recognized leader in vintage analog and digital electronic music repair and upgrades: Synthesizer.repair. Through that work, I spent years deep in the guts of legendary instruments, made by American companies like MOOG and ARP. First wave Silicon Valley companies, like Sequential Circuits, and E-mu Systems.  

It has been decades since most of these great companies went bankrupt or were subject to corporate acquisitions and mergers. In two of those past decades I’ve been restoring and enhancing this legacy hardware.  Hardware that was built to last, often with craftsmanship and design principles that today’s disposable tech culture has long since abandoned. I saw firsthand how technological hardware, when designed with clear purpose and longevity in mind, could outlive the businesses and engineers that designed and manufactured it. I learned that no amount of software optimization could compensate for de-valued components or bad engineering decisions at the physical level.

But despite this, the industry I was primarily working in—the world of high-tech software—continued to sideline hardware, treating it as a necessary but inconvenient layer beneath the real world work of “abstraction and digital scalability”. We let financialization gut domestic manufacturing, let critical systems fall into obsolescence, and convinced ourselves that software could replace physics. But physics doesn’t care about abstractions! The age of hyper-growth SaaS models and platform monopolies is fading, and what remains is a new industrial reality. It’s not enough to engineer software that moves money around. We need to move molecules, build infrastructure, and reclaim what was lost!

That’s why I left the waltz. That’s why I joined Terraform Industries. We’re not here to optimize ad delivery algorithms or streamline subscription funnels. We’re here to reengineer energy itself—to take carbon from the air, turn it into synthetic natural fuels, and rewrite the relationship between civilization and its fossil fuel legacy. It’s not a ‘neat problem’. It’s not infinitely scalable in the way software pretends to be. It’s messy, ugly, physical, and constrained by real-world limitations. And that’s exactly why it matters.

We aren’t chasing fleeting trends or digital virality; we’re engineering permanence. The next wave of innovation won’t be measured in clicks or ad impressions but in megawatts, molecules, and matter brought to order with blood, sweat, and mathematics! While the old guard clings desperately to spectacle, we are designing the future with our hands—wiring circuits, welding frames, and forging solutions from the raw elements of reality itself. This isn’t a game of abstractions; it’s a battle of physics, precision, and perseverance. And we will win!

The transition isn’t seamless. The old system refuses to acknowledge its obsolescence. But I can see it happening—an exodus of software minds moving toward hardware, manufacturing, and deep tech. We were told for years that the future was virtual, that the next frontier was a digital metaverse where we’d transcend the physical world. Neal Stephenson imagined it as speculative fiction; Mark Zuckerberg tried to brute-force it into reality. But the truth is, no one wants to live in a headset, strapped into an abstraction while the real world crumbles. The metaverse was never a vision of the future. It was a coping mechanism for a civilization unwilling to maintain its foundations.

Now, we see reality, and we know the real frontier isn’t a simulated one. The future isn’t in escapism; it’s in infrastructure, in power grids and carbon and calcium loops, in the material mastery of energy and entropy. We aren’t here to build digital theme parks for billionaires; we’re here to rewire the physical economy, to re-engineer the molecular flows that civilization depends on. Carbon capture isn’t a buzzword—it’s a thermodynamic necessity. Calcium looping isn’t a theory—it’s an industrial imperative. The future will be built by those who understand chemistry, metallurgy, and physics, not those trying to sell us a pixelated escape from reality. Mark my words: The future is industrial!

If you’re out there—if you feel this shift deep in your bones, if you see the writing on the wall and are looking for a way to put your skills to work—join us. The world doesn’t need another app, another algorithm, another digital metaverse. It needs hands-on minds, polymaths who aren’t afraid to think across disciplines and build real, lasting solutions. Send us a one page introduction. The future isn’t waiting for permission, and neither should you.

As the ballroom of the old world continues to burn, I no longer feel the need to do that dance. Instead I’m dancing in the lab. I’m out on the workshop floor. I’m designing and building. And that’s where the real future begins.

After the success of our team in building an end-to-end demonstration of our fuel synthesis process, aspects of the future are coming into focus.

Terraform is emerging from its R&D phase into its value generation phase. That’s right – it’s time to make money! This necessity is not motivated by greed, but by the recognition that it is impossible to rapidly deliver a complete industrial stack for synthetic fuel production at global scale without churning out enormous value to every person on Earth. Terraform’s success has positive externalities, including solving long term CO2 excess and energy scarcity, and looking like a gold mine.

While the team is hard at work building our first commercial-scale Terraformer and preparing for field deployment at a nearby site, it is time to lay out a comprehensive road map for the successful execution of our vision.

Terraform started, for pure technical expediency, with synthetic natural gas, but we’ve made no secret of our longstanding ambition to own the production of every kind of hydrocarbon in every market on Earth (and beyond!).

For each of the following entries, we list market size in annual sales in 2023 – not including the substantial future growth in many of these industries. We also list the year when we intend to externalize this capability as a standalone product.

Near-Term Priority Commercial Opportunities

Electrolyzer: The Cornerstone of Green Hydrogen Production
The Terraform Electrolyzer produces the cheapest green hydrogen on Earth.
Market size: $3b. Market entry: 2025.

Hydrogen Gas (H2): A Key Input for Industrial Applications
Terraform’s green hydrogen is competitive with gray and blue hydrogen, and will serve applications in chemicals, fertilizers, and metallurgy.
Market size: $240b. Market entry: 2025.

Methane Terraformer: Clean, Synthetic Natural Gas from Anywhere
Combining our electrolyzer and CO2 direct air capture (DAC) with a methane reactor, the Terraformer converts any patch of land into a carbon-neutral and fossil-free natural gas well.
Market size: $1.5t. Market entry: 2026.

Natural gas (CH4): A Drop-In Replacement for Fossil Fuels
Our synthetic methane provides a chemically identical drop in augmentation for fossil gas supplies, carbon neutral, and manufacturable anywhere on Earth – not just in oil-rich places.
Market size: $1.8t. Market entry: 2025.

Mid-Term Priority Commercial Opportunities

Direct Air Capture: Scaling Carbon Removal Solutions
Terraform’s proprietary DAC captures CO2 from the air anywhere on Earth.
Market size: $3b. Market entry: 2025.

Carbon Dioxide Gas (CO2): A Growing Market for Industrial Carbon
Relatively small as an isolated market, DAC will become humanity’s plants as the default source of industrial carbon, a critical feedstock for synthetic fuels and chemicals.
Market size: $11b. Market entry: 2025.

Methanol Terraformer: A Versatile, Renewable Liquid Fuel
Integrating our Electrolyzer & DAC systems with a methanol reactor, this system converts sunlight into a valuable liquid chemical feedstock that is energy-dense and easily transportable globally.
Market size: $2.5t. Market entry: 2027.

Methanol (CH3OH): The Foundation of Future Hydrocarbons
Serves as a versatile chemical feedstock for numerous applications including, fuels, solvents, or as a precursor to other hydrocarbon products via well-understood processes.
Market size: $31b. Market entry: 2026.

Methanol-Derived Products: Unlocking the Remaining Petrochemicals and Fuels
Methanol can be readily converted into numerous petrochemical products via well-understood catalytic pathways with acceptably high yield. 12-15% of current petrochemical use goes into polymers, durable stores of carbon that will previously have been in the atmosphere, and one of the few cash-flow positive mechanisms for carbon dioxide removal (CDR). A non-exhaustive list:
– Dimethyl ether (DME, CH3OCH3). Market size: $9b. Market entry: 2027.
– Ethylene (C2H4). Market size: $140b. Market entry: 2027.
– Propane (C3H8). Market size: $100b. Market entry: 2028.
– Gasoline (C7-C11). Market size: $1.4t. Market entry: 2028.
– Kerosene (C9-C16). Market size: $390b. Market entry: 2028.
– Diesel (C10-C22). Market size: $225b. Market entry: 2028.

Off-Grid Solar Development: One-Stop-Shop for Cheap Electrons
In addition to selling Terraformers to solar developers, we are also in the process of standing up our own solar development capability similarly focused on radically accelerating cost-reductions ahead of the market so that the entire process from sunlight and land through to hydrocarbon production is provided by a single entity.
Market size: $250b. Market entry: 2026.

Longer-Term Priority Commercial Opportunities

Terraformer Factory of Factories: Building the Future of Global Fuel Production
Over the next two decades, the world will need hundreds of Terraformer factories. Terraform Industries will complete the journey from making cheap green hydrogen with our electrolyzers, natural gas from hydrogen, Terraformers to make natural gas, factories to make Terraformers, through to factories to mass produce the tooling necessary to set up more factories.
Market size: $300b. Market entry: 2029.

We currently have no plans to verticalize into polysilicon production.

In the intervening years, though, I’ve come to realize that obvious things aren’t actually obvious, in general, so here I’ll explain why Terraform’s radical architecture is both unconventional and also inevitable.

We want to increase the supply of cheap, energy intensive materials that are foundational to our civilization.

Incidentally, finite and expensive fossil sources for hydrocarbons cannot scale sufficiently, so solving this problem will also solve carbon emissions among a bunch of other problems. In fact, I’m convinced that a cheaper source of carbon neutral fuel is the *only* thing that can displace fossil fuels.

The most polluting industries are coal, oil, natural gas, concrete, and steel.

I don’t mean to exclude anyone, but solving any one of these problems takes a big chunk of the problem out. We can have as much carbon-neutral vanadium as we want but it won’t make a dent in emissions and it isn’t a path to post-fossil abundance.

Solving fuels helps to solve steel and cement too, so let’s focus on fuels.

Product must compete on cost.

Energy intensive fuels and materials are commodities whose cost is driven by energy opex and whose cheap reliable supply is essential to our civilization and life as we know it. There is no economically or politically sustainable way to make synthetic fuels at scale unless they are net generators of value, which means they must at least be cheaper on the margin than drilling.

Go to cheapest energy source, focus on geographically distributed, unlimited feedstocks.

We want to make oil and gas, which are hydrocarbons, chemically speaking. We need a source of hydrogen, carbon, and power. All must be as cheap as possible, or we will have no chance of competing with drilling as a source of fuel, and thus no chance of delivering value to customers.

We can make an argument that point source carbon capture is cheaper than direct air capture (DAC) but point sources are limited, geographically localized, and have their own externalities. They also require the captured CO2 to be transported somewhere else. So we’re going to have to solve DAC.

Same with hydrogen – we will need to make green hydrogen from water and power, which is easy enough to do, but it has to be done cheaply, which means we need lots and lots of cheap power.

How do plants do it?

The atoms in your body mostly come from food which in turn came from plants. Plants are mostly made of carbohydrate biopolymers, and the carbohydrate is derived from water and CO2, powered by sunlight. Photosynthesizing plants cover almost the entire Earth. Their mass derives from water and CO2 which themselves come from the air. Almost all the hydrogen, oxygen, carbon, and nitrogen atoms (>96% by mass) in our body were, relatively recently, part of Earth’s atmosphere. Half of the nitrogen atoms came via the Haber-Bosch process for making synthetic nitrate fertilizer.

That’s interesting. But if plants could meet humanity’s energy needs, we would never have needed to mine coal to start the industrial revolution. If we fed all the world’s plants into biofuels it would meet barely 10% of demand – it just doesn’t scale.

Cheap solar photovoltaic power is the key.

This is not the place to sing the praises of solar – I have done so elsewhere. Suffice to say that solar PV is far cheaper, and growing far faster, than any other energy source in history.

Ordinarily, oil is burned as fuel to make electricity, with about 35% efficiency. We can convert electricity back into oil, with about 35% efficiency.

The golden rule of business is to buy inputs where they are cheap and sell products where they are expensive. Synthetic e-fuels are a terrible idea, unless solar PV electricity is about 35% x 35% = 12% of the cost of oil-powered electricity. As of now, it is crossing that threshold with no signs of slowing.

This is a key tipping point. A cusp. A transition.

Find some way to harness the cheapest possible solar, and you can undercut oil drilling, if you do it right.

Solar PV unblocks the land constraint.

Agriculture is solar powered, but its useful energy productivity per acre is about 1000x lower than solar PV. Food is much more expensive (~100x) per calorie than oil, but even so solar is about 100x more economically productive than agriculture.

If any nation on Earth can be self-sufficient with respect to food, it can be self-sufficient with respect to oil and gas with about 10% of the land. It doesn’t have to be arable, irrigated farm land either! Deserts, swamps, mountains, brown fields, rooftops, mountains, lakes, oceans all work just as well for solar.

Solar is intermittent, so work with intermittency.

Solar PV is the cheapest energy in history – during the day. Batteries can spread the load, but add unacceptably high costs for synthetic fuel.

So, we must adapt the synthesis process to work with intermittency – just like plants do.

Technical intermittency compatibility.

Most chemical processes do not like intermittency. Tough. Find ones that do. Build electrolyzers, DAC, and reactors that can cope with variable or zero power supply, in order to harvest the bounty of cheap solar. Internal combustion engines can throttle up and down in seconds, so it’s not forbidden by the laws of physics.

Architectural intermittency compatibility.

We must mercilessly reduce parasitic costs within the system. To achieve cost parity with fossil fuels, the Terraformer can’t afford to transport power, hydrogen, or CO2 from place to place. The Terraformer must be an integrated system localized with the solar array, displacing (and thus deleting) the expensive electrical inverter.

The biggest mistake a competent engineer will make is to optimize a part that should not exist. All those parts cost money. Delete them all. If you’re not adding deleted parts back in, you’re not deleting nearly enough.

Sunlight falls on the ground for free, heating dirt. The solar arrays sit on the ground and convert sunlight into electricity. The electrolyzer sits next to the solar array and consumes most of that electricity to make hydrogen. The reactor sits next to the electrolyzer to consume the hydrogen and make natural gas. The reactor sits next to the DAC to receive CO2.

All three subsystems form an integrated unit that converts commodity solar electricity to valuable natural gas, anywhere on Earth. It is an economic engine that performs an arbitrage over the physical form of energy and captures all the upside of the continuing decosting of solar PV panels.

The Terraformer minimizes the costs of moving power, hydrogen, and CO2 to convert solar energy into a high value, backwards compatible drop in, storable commodity (natural gas) in the array – a solar gas well that requires no drilling and no favorable geological structures.

Financial intermittency compatibility.

Solar arrays average 25% utilization, with less than 100% maximum sunlight for about 12 hours per day. To achieve comparable project economics, the Terraformer must have 4x lower capex. This would be impossible, except that solar is so cheap the Terraformer can afford to be a little less efficient.

Extremely cheap primary energy supply covers a lot of sins.

A discrete, modular, manufactured, scalable 1 MW electrical synthetic fuel plant.

Utility scale solar arrays standardize at 1 MW to minimize the economic cost of power transport from the array to the substation. 1 MW is also the natural size for a Terraformer. The smaller size helps us in multiple ways. It is easier and safer to build, test, finance, and iterate.

A smaller reactor is more nimble and easier to throttle.

A 1 MW Terraformer is the size of two shipping containers. It is movable by truck, skid mounted, and requires minimal in field construction or maintenance. It is mass produced in a factory by a trained workforce, enabling low cost and higher reliability.

Methanol is our approach to liquid fuels.

Natural gas provides roughly half of US hydrocarbon energy consumption, and is growing. It has substantially higher energy content and lower CO2 emissions than liquid fuels, as well as lower toxicity, albeit some handling issues relating to being a gas. Liquefied natural gas is an emerging fuel for high performance aviation, which is poised to become the greatest source of hydrocarbon demand as passenger hypersonics reach maturity.

Half of US hydrocarbon consumption is liquid petroleum, and our job will not be done until we can displace that too. Does this mean we need a small, throttleable Fischer-Tropsch reactor in the Terraformer? Not exactly.

The job of the Terraformer is to transform light energy into a low hassle, high energy density, transportable, storable fuel. After methane (CH4) the next most useful chemical is methanol (CH3OH), a liquid at room temperature.

So we’re starting work on our methanol reactor. It shares much commonality with the methane reactor, and works essentially the same way.

CO2 + 4 H2 -> CH4 + 2 H2O (methane)
CO2 + 3 H2 -> CH3OH + H2O (methanol)

After methanol is produced at dozens of Terraformers positioned inside their respective arrays, it is moved by narrow diameter pipe to a central gathering point, where it is stored in a tank. A methanol tank farm can store energy hundreds of times more cheaply than H2 storage or battery storage.

Once we have a large supply of methanol, it can be chemically converted into any desired chemical using tried and tested continuously operated large scale chemical synthesis. Dimethylether (DME), propane, ethylene, longer chain alkanols, even gasoline, diesel, kerosene/SAF/JP-8 using the MTG process. There are many ways to perform these synthesis steps, once the Terraformers provide a large, storable supply of low cost methane and methanol.

End to end demo.

Idea mazes are thought experiments. Nice in theory. Plenty of clever ideas crash upon the rocks of reality. Now try it in the real world and find out how good your internal model of the universe actually is.

So we did.

In March, we validated our intuitions by building our own capex-optimized electrolyzer, DAC, and Sabatier methanation reactor, connecting them together, and producing pipeline-grade synthetic natural gas.

As far as we know, this is the first ever pipeline-grade natural gas ever produced with a scalable architecture. Low capex, DAC CO2, cheap solar hydrogen.

The Terraformer is inside the tent of capitalism.

One thing, and only one thing, separates neat ideas which move the needle on climate and neat ideas that make no difference: Large scale adoption, catalyzed by transformative value generation in enormous markets. Tesla electric cars yes, hydrogen cars no. Solar photovoltaic power and wind turbines yes, nuclear cargo ships no.

To make a meaningful difference on energy and climate, solar synthetic fuels must be economically compelling to billions of people. All the clever technology in the world matters not a bit if production requires a premium, rather than delivering a dividend.

In this post, I’ve laid out how Terraform thinks about this problem and how we’ve positioned the Terraformer in design space to solve it. The Terraformer is designed to convert economically unproductive and neglected land into oil and gas wells, which is about the largest value upgrade imaginable. The Terraformer generates value for its owners and its customers. The Terraformer is driven into the market at scale by its ability to attract project capital with attractive returns.

Join us!

We’re hiring! We are always on the look out for exceptional engineers to join us on our ambitious project.

Last Wednesday March 27, 2024, we commissioned our integrated power-to-gas demonstrator. The system converts electricity and air into affordable fossil-free natural gas using proprietary subsystems, developed in-house, for hydrogen electrolysis, direct air capture, and methanation. This marks a major milestone in the technical and commercial development of our technology and another step on the path to our positive vision for the future where everyone has access to plentiful cheap, reliable energy, harvested from renewable carbon and power, with lower ecological and climate impact.

Our core innovation is the utilization of proven and well-understood industrial processes implemented within innovatively simple and cost-effective proprietary technology systems adapted to the daily variations of renewable power for producing cheap sustainable natural gas from sunlight and air. These design and process innovations are responsible for driving production costs down for both their green hydrogen and atmospheric CO2 capture to levels previously thought unattainable. Combining super-cheap solar power with low-cost green hydrogen and CO2 capture are the key to producing commodity-grade electro-fuel alternatives at competitive commodity prices.

Our elite team of engineers have made significant progress towards our stated goals of radically cheaper green hydrogen and direct air capture CO2. We follow a simple playbook. Adapt a known industrial process to ultra-cheap solar power by a) radically reducing capex to compensate for lowered utilization and b) developing processes that are robust to variations in power due to weather, as well as diurnal and seasonal power shifts.

First, our innovative electrolyzer converts cheap solar power into hydrogen with current production costs at less than $2.50 per kg of H2. How? To first order, our electrolyzer capex costs <$100/kW, and we’re currently baselining solar PV DC electricity at $20/MWh. There is no hand waving about economies of scale or subsidies here, though we are eligible for the full IRA 45V green hydrogen tax credit, worth $3/kg-H2.

Second, the proprietary direct air capture (DAC) system concentrates CO2 in the atmosphere today for less than $250 per ton. How? Our DAC capex costs <$600/T-year-CO2. Like the electrolyzer, this represents a substantial improvement on the state of the art.

Finally, our in-house multistage Sabatier chemical reactor ingests hydrogen and CO2, producing pipeline grade natural gas, which is >97% methane (CH4). Our reactor uses no rare or expensive catalysts nor cherry-picked data – it handily operates for hours in steady state.

The Terraformer combines these three subsystems into one modular chemical plant, skid-mounted for deployment alongside a solar PV array.

It is particularly cool to develop and build all three of our sub-systems in our castle in Burbank, CA, right across from the long-time home of the fabled Lockheed Skunk Works. In-house production & manufacturing enables faster iteration, lowered supply chain risk, and the successful demonstration of the complete techno-economic model for producing cost-effective fossil-free natural gas.

While I generally prefer understatement, it is hard to overstate the significance of this achievement, which has been two years in the making. Humanity today depends on traditional oil and gas to provide them with a cheap and easily transportable source of energy-dense hydrocarbons, to the tune of 100 million barrels per day, which unfortunately also generates ~50 gigatons per year of CO2 emissions.

The Terraformer unlocks an industrial-scale source of renewable hydrocarbons that breaks our dependence on fossil fuels and lays the foundation for another century of spectacular growth, broad-based wealth creation, and global environmental restoration.

What’s next? We’re taking reservations for our first production Terraformers, and ramping production to meet demand.

• Fuel is cheap and economically desirable.
• Fuel is backwards compatible with existing infrastructure, distribution, and uses.
• Terraformer is scalable with a fast ROI, comparable to drilling but with lower risk.
• Synthetic fuel needs only sunlight and air, so no strategically fraught dependence on geology and foreign security issues.
• Clear path to unconditional energy abundance for all humans.

In addition to these, it is generally underappreciated that Terraform’s synthetic fuel has significant environmental benefits, summarized in the chart below.

Let’s go through them one by one, organized according to the various sustainable development goals.

7: Affordable and Clean Energy

Fossil fuels are ultimately derived from ancient plants while fuels such as bio-ethanol are made from agricultural products, so both are ultimately derived from solar power. Unfortunately, there is simply not enough of either, at any price, to meet the demand of a growing human population in the process of destroying poverty forever.

The Terraformer is able to convert solar energy to fuel with ~20x higher per-acre productivity than corn->bio-ethanol, meaning that our technology can deliver abundant, cheap, clean chemical energy for the masses.

9: Industry, Innovation and Infrastructure

The Terraformer, installed within utility-scale solar arrays, produces a predictable and constant supply of natural gas that is cross-compatible with existing fossil gas. That means we can use the same gas gathering, storage, injection and transmission pipelines in the same gas production areas. As we ramp up solar gas production and fossil wells naturally deplete, we can smoothly transition our entire industrial economy onto a new, sustainable energy source requiring:

• No new infrastructure build out, avoiding the political pain of eminent domain, not to mention the incredible cost and delay.
• No interruption in supply.
• No increase in price, minimizing economic disruption.
• No change in downstream applications, because the fuel is chemically identical.

Only the Terraformer can offer a smooth, swift supply side substitution for fossil fuel.

13: Climate Action

Conventional wisdom on climate has been unable to solve the collective action problem. Fossil fuels underground represent enormous wealth and short-term benefits to humanity, and no government on Earth (except perhaps North Korea) has been able to unilaterally impose the sort of grinding poverty on its populace that would be necessary to reduce consumption to a level low enough that the biosphere can naturally absorb the extra carbon, stabilizing the climate. And nor should we want to, since starving billions to death is almost certainly a worse outcome even than dealing with the impact of unmitigated climate change.

Fortunately, the Terraformer presents another path forward. The Terraformer produces fuel cheap enough that its supply will naturally displace fossil production without any heavy-handed economic policy. The way out is up – producing more fuel, but without extracting it from the ground. The Terraformer sources its CO2 from the air, so even if all our fuel is burned, it will not result in the transport of carbon from the crust to the atmosphere. It will not result in new CO2. The carbon flow is circular within the atmosphere. Of course, a decent fraction of hydrocarbons are turned into chemicals such as plastics, adhesives, medicines, carbon fiber, and so on – that effectively sequester that carbon for a long time. So Terraformer fuel is actually slightly carbon negative in practice.

Another under-appreciated benefit of the Terraformer is that it creates an economic forcing function for carbon capture via direct air capture (DAC). Other potential approaches to carbon capture depend on carbon users voluntarily purchasing offsets or paying for sequestration. This is unable to scale because almost all fossil fuel is consumed by processes that are very price sensitive. The chart below shows that the companies that consume the most fuel generate very little profit per tonne of CO2 produced, and so there simply isn’t the economic buying power to pay for sequestration at scale. As described earlier, the answer isn’t to simply ban any lower-profit use of fuel, since we all depend on fuel to produce the food we eat, the clothes we wear, the houses we occupy, and the factories that the other, more profitable companies depend on.

The Terraformer instead places DAC upstream of the point of sale, turning CO2 capture from a cost center to a profit center. It is an economic engine to scale up the deployment of DAC systems worldwide, and we don’t even need a carbon tax to do it.

Ultimately, I believe that while synthetic fuels will rapidly crush new carbon extraction from the crust, deploying enhanced weathering to deal with legacy CO2 emissions and stratospheric sulfur injection to keep a lid on heat for the next two decades will save millions of lives by accelerating the re-stabilization of Earth’s climate.

8: Distributed Economic Growth

Where do people live? Most of us live in cities and are accustomed to thinking of land as scarce and houses as expensive, but much of the Earth’s surface is completely unpopulated. Why? 99% of population density reflects where people’s grandparents were able to grow enough food to survive, before refrigeration and trucks. This requires the things plants need – sun, water, arable land, and a sensible climate.

Looking at a chart of photovoltaic power potential, one could be forgiven for thinking that the only good places for solar are the Andes and Himalayas. Solar power is so cheap that the cost of transporting it to the end user is relatively high, and so instead what we find is that the most local solar is almost always the cheapest – rooftop solar in the limit. A closer look at the solar power map reveals that the places where nearly everyone lives – the yellow area – is only 20-30% less sunny than the most scorching high altitude desert, where no-one lives. Solar PV panels are falling in price by about 15% per year, so in any given market just a few years separates solar electricity price parity.

This has two profound implications for the future of energy.

First, as we’re already seeing in the market for electricity, local development is king. Battery deployment is already undermining the need for radical new investments in electrical transmission capacity, previously thought necessary to decarbonize the grid.

The Terraformer extends this trend to chemical energy too. Soon, we will be able to deliver farm-to-table food, electricity, and hydrocarbons. The implications are profound. Even relatively small cities, anywhere on Earth, will be able to develop and deploy local supply chains for their energy and own local manufacturing, which translates into increased local demand for jobs in the green energy sector and downstream uses. High productivity jobs everywhere.

Second, the Terraformer’s ability to produce abundant hydrocarbons anywhere on Earth that people live means that the age of globalized strategic interdependence on thousands of enormous oil tankers, guarded by the US Navy, bringing oil from the Middle East to the rest of the world is coming to an end. No longer will your country’s access to energy depend on foreign buying power, beneficial geology, or arcane international relations. This will have all kinds of interesting geopolitical implications but we’re headed for this world whether we like it or not. We believe that a world with abundant local production of cheap hydrocarbons is a peaceful, prosperous world, but we’re not there yet.

10: Realignment of Insecure Energy Supply Chain

In addition to the geopolitical implications of localized energy self-sufficiency, we will also improve the lives of the marginalized and precarious people who currently live at the ends of the supply chain. In particular, Terraformers are at least 20x more productive, on a per-acre basis, than biofuels, and they don’t require A-grade Iowan corn growing conditions to operate. To a remote desert community perpetually hovering on the brink of economic disaster, the Terraformer acts like an incredibly lucrative cash crop, as transformational as the discovery of oil in Arabia or Texas, but without needing to win a geological lottery. Anyone can deploy solar arrays and Terraformers and achieve local energy abundance and economic security. For arid areas, the Terraformer can also produce excess water as a DAC byproduct.

11: Clean Energy Production Near Cities and Communities

We already understand that every deployed gigawatt of solar PV saves one hundred lives per year, mostly from displaced coal-fired electricity and its pollution. Last year, we deployed about 450 GW globally, up from 250 GW the year before. Cumulatively we’re over 1 TW of solar, saving more than 100,000 lives per year from respiratory failure-induced premature death.

Terraformers extend this benefit to chemical energy. In displacing biofuels by using 20x less land, Terraformers also produce no dust, air pollution, or water pollution. This is in marked contrast to fossil fuel production, as well as some methods of biofuel production.

Our DAC carbon capture system scrubs most of the CO2 and some water out of the air, incidentally removing dust and smog pollutants through our filtration system. The products are carbon-neutral methane (natural gas), oxygen, and optionally we can unbundle to produce fresh water, hydrogen, CO2, or even deuterium.

2: Increased Food Security and Effective Land Use

Terraformers do not cause the destruction of arable land and even reduce demand on productive land. The US is anomalously agriculturally productive, but even so, several states worth of land are devoted to ethanol production – all of which can be substituted by Terraform’s solar synthetic product with 20x less land consumed. Even better, the land which is used for ethanol production can be put to more productive uses, since the Terraformer happily operates on arid, unproductive land.

Even more than ethanol, enormous tracts of land are used to produce livestock feed. Terraform Industries is pioneering direct solar synthesis of molecular fuel, and natural gas is just the first in a long line of reduced carbon products which will soon include gasoline and kerosene and eventually include starches, fats, and proteins. In other words, we can produce more animal feed, more cheaply, at higher quality, and with less land using the solar PV-synthesis route, than by planting corn and waiting for it to grow. There isn’t enough productive land on Earth to agriculturally produce enough food to supply eight billion humans at US standards of living, but the Terraformer shows how we can produce much more with much less.

15: Reallocate Production and Enhance Biodiversity

The incredible productivity of modern industrial agriculture has created a world largely without hunger, a towering achievement in direct contrast to our 20,000 generation legacy of near constant hunger and starvation. Endless rows of wheat, corn, soy, and winter vegetables at affordable prices would have boggled the minds of our great-grandparents, let alone our bronze age ancestors. Nevertheless, such abundance covers a disturbing truth – modern agriculture is a pesticide and herbicide-supported monoculture in the place of pre-existing forest, prairie, swamp, or other ecosystems.

There is no path to energy abundance via biofuels which does not extend this agricultural monoculture to every possible fertile corner of the Earth, due to the low intrinsic CO2-capture efficiency of even the best plants. The only way to achieve post-carbon energy abundance is through electrical synthesis, which is much more productive than plants, mostly because the equipment doesn’t have to keep itself alive! Cheap energy requires cheap power, and the cheapest power is solar power, from which we get 20x higher productivity than agriculture. If someone finds an even cheaper source of power, we’ll use that too!

Even solar deployment on arid land is often a net biodiversity benefit, since the panels shade the parched ground “for free”. Shading improves moisture retention and increases insect biodiversity, effectively halting or reversing the process of desertification which has steadily progressed since the end of the last ice age and accelerated with global warming and increased agricultural pressure.

6: Efficient and Clean Water Usage

Communities, particularly in arid areas, are justifiably concerned about water scarcity, especially as rivers begin to dry up, glaciers melt, and traditional water allocation methods hit their limits. Biofuel production is a climate-healthy alternative to fossil fuels but often requires more water, not to mention more land, than we can readily make available.

In contrast, the Terraformer sips water, only requiring a few thousand liters per day per MW – barely enough to irrigate a lawn on the same small foot print. In areas with insufficient water to supply the Terraformer, its DAC component can be adapted to recover the water it naturally captures in its CO2 sorbent material, producing as much water as it needs and, if desired, even more. That is, a Terraformer installation can provide not only local employment, local energy, and a local source of wealth generation through energy sales, it can also provide a source of water.

Unlike conventional gas production, the Terraformer involves no drilling into subsurface rock, so it cannot disturb groundwater. The gas we produce is free from the usual fossil contaminants, so we have no difficulties with spill containment or waste disposal. Unlike agriculture, we do no spraying or irrigation, so there’s no contamination of surface water either. Unlike biofuel production, we have no toxic byproducts to treat or dispose of. We aspire to deliver a community-friendly local energy production system that actively improves the environment in which it’s placed.

12: Responsible Consumption and Production

In a pre-Terraformer world, sustainable growth presents us with a contradiction in terms. From a humanitarian perspective, we want to increase the supply of cheap energy and cheap goods to people all over the world, by which they can lead longer, happier, more productive lives. From an environmental perspective, there is simply no way this can be done without further rampant destruction of limited natural resources and the ongoing pollution of the atmosphere with CO2 by which we are heating the Earth beyond any sensible limit.

With the Terraformer, this impossible contradiction goes away. Why? It is fundamentally a more efficient, more productive root source of wealth for humanity. It allows us to price the production of more energy at the rapidly falling cost of energy, rather than the cost of exploring and drilling into ever deeper, ever poorer oil-bearing rock. Its fundamental inputs, sunlight and air, are essentially unlimited on any scale meaningful to the next few centuries of growth and development. We can actually “have it all” – democratized renewable energy production and consumption driving global growth without the environmental horrors of oil drilling and melting ice caps.

Like any machine, the Terraformer does not last forever, so we should look to its full life cycle to ensure we’re not bequeathing some nasty toxic legacy to our children and grand children. The Terraformer is deliberately designed to use only generic, abundant materials for its construction, so mass deployment does not require enormous new mining development or new supply chains for exotic materials. The Terraformer and its solar panels are so environmentally inert and non-toxic that they could be land filled without any concern, if they weren’t instead recycled. If, after some decades of cheap solar synthetic production, we discover some other, superior way to make energy, it is trivial to de-rack the solar array and return the land to its primordial condition, only now persistent shading will likely have increased plant and animal abundance in the area. The Terraformer will have lower persistent impact on the land it uses than almost any other human land use, including development, agriculture, mining, grazing, or even extensive hiking.

Vision For The Future

Let’s integrate these disparate parts of the vision to understand how the Terraformer’s approach to synthetic fuel will transform our future this century. Instead of a miserable century of rising temperatures, increasing geopolitical strife, fuel scarcity, and economic malaise, we are poised to address the climate problem head on, provide copious energy with simplified supply chains to every human on Earth, and ensure that the bounty of cheap solar flows to all our energy sources.

In the short term, we’ll see growth of solar PV, and solar synthetic fuel through the sunnier and more developed parts of the world, eventually extending everywhere on Earth except perhaps the UK and Nordic countries. Of these only the UK has significant population and demand for energy and may end up relying on wind – but it hardly matters in the context of climate.

Energy abundance in the 2030s will also greatly increase the availability of low cost plastics, metals, cement, and desalinated water, among many other bulk products.

Through the 2040s, as the renewable electricity and synthetic fuel supply chain is completed, we’ll see lateral growth into production of more complex molecules, including animal feed and healthier staples, allowing us to focus agricultural effort on the highest value products and free much of Earth’s highly ecologically productive areas for cultivated restoration of a curated wilderness, giving humans the choice to live in large, dense modern cities or in some kind of anarcho-techno-primitivist state, only without involuntary hunger or tooth decay, or anywhere between.

The key unlock is that the natural world, which depends on photosynthesis for its energy, can decouple and co-exist with the industrial human world, which will depend on much more productive solar PV and eventually perhaps fusion for its energy. Because solar PV is a thousand times more productive than plants, our civilization can continue to grow its energy consumption far beyond what the natural world can provide without consuming more than a small fraction of the available land.

Here are some renders from Midjourney to illustrate our vision of a future that’s abundant for both humans and nature.

Here’s Yuma in 2040.

This is Tuscon in 2045.

This is Mojave in 2050.

And Barstow in 2060.

Solar is now cheap enough that synthetic fuels are coming no matter what! Our choices can only affect how quickly we’re able to accelerate the retirement of catastrophically harmful legacy energy sources, and how beautifully we cultivate the world that comes after.