Casey runs Terraform Industries, a startup focused on creating synthetic natural gas from sunlight and air. The vision involves data centers eventually replacing traditional utilities to provide cheaper local power. This shift is driven by the urgent need for power among hyperscalers who are struggling to find enough energy to meet their demands.

For the last four years I've been running Terraform Industries, my startup, to build cheap synthetic natural gas from sunlight and air. Eventually these data centers can replace utilities, give you cheaper costs for power locally, and be more of a consumer play.

The scale of solar adoption is shifting globally. China has installed more solar capacity this year than the entire United States has produced. This rapid expansion is accelerating the learning rate for solar technology. Currently, the learning rate is 48 percent. This means that every time the total amount of solar deployment doubles, the cost drops by 48 percent. This dramatic cost reduction creates a significant advantage for countries that can harness cheap, abundant power.

As of this year it's running at 48% which means that for every cumulative doubling of deployment, we are getting a 48% reduction in cost.

The United States currently maintains a lead through financial capacity and financing. However, Casey suggests this advantage may only last for another decade. The future economy will likely belong to those who can produce the cheapest energy at the largest scale.

Casey Handmer is the founder of Terraform Industries. His company focuses on building cheap synthetic natural gas from sunlight and air. With a background in theoretical physics and a stint at NASA, Casey is now applying his expertise to the energy sector. His company aims to sell synthetic gas to utilities through existing legal structures. He also maintains a blog where he makes specific predictions to test his understanding of reality.

I think it's important that if you're a public intellectual in this space, you should occasionally make specific predictions. Really that's the only test of your model for understanding reality. Does it make testable predictions or not?

Human history is closely tied to energy use. Before the Industrial Revolution, the economy was limited by the metabolic energy of humans and animals. The average person lived on just a few hundred dollars a year because they could not access more energy than they could eat. The steam engine changed this by allowing us to use coal to power machines. This effectively bypassed the bottleneck of human muscle and led to rapid economic growth. Wealth now doubles every generation or two as a result of this transition.

The current electrical grid is a complex system that balances supply and demand in real time. When someone turns on a light, it creates a path for electrons that forces a turbine far away to work harder. This system of coal plants and long wires worked well for over a century. However, solar power and batteries are now disrupting this model. These technologies are inherently superior because they can handle the timing of energy supply better than traditional infrastructure.

Solar and batteries have permanently destabilized the stable electricity monopoly markets that we think of as our electricity utilities today. We have been in that process for just over 10 years and we are probably about a third of the way through.

Solar technology was first used for space missions in the 1960s. For a long time, it was just a cleaner way to make power, but it cost too much. Things changed around 2009 when solar power became cheaper on the margins than the next cheapest form of grid electricity. This gave the industry the foothold it needed to grow beyond academic labs and government subsidies.

Casey explains that the Chinese government played a major role by investing heavily in the industry. This moved the field from a small group of researchers to tens of thousands of people working on the problem. These efforts unlocked new innovations, such as refining silicon specifically for solar panels. This investment dramatically increased the speed at which the technology improved.

For every cumulative doubling of deployment, we are getting a 48 percent reduction in cost, which is about as steep a learning rate as I have ever seen in any technology.

The learning rate for solar is now 48 percent. This means that every time we double how much solar power we deploy, the cost drops by nearly half. This is much faster than the 10 to 20 percent learning rates seen in the automotive or aviation industries. Solar is now following a path more like the rapid evolution of microchips. Because global capacity doubles roughly every two years, costs are falling by about 25 percent every single year.

While fossil fuels currently dominate primary energy production, the market is shifting toward electricity. Some parts of the world have already shown that it is possible to move large parts of industry and transport to electric power. This transition is being driven by market forces rather than just government mandates. Casey points out that for electrification to succeed, it must be packaged in a way that is actually appealing to customers. Tesla is a prime example of this. Instead of building small, limited cars that people only drive out of necessity, they created vehicles that people genuinely want to own.

The exciting thing is for the entrepreneurial class, if you can figure out a way to package electrification of some fundamental technology in a way that's appealing to customers, you can laugh all the way to a trillion dollar company because it's just such an enormous sector.

Ground transportation and much of our heating and cooling are likely to go electric. Casey shares a spicy take that we might eventually prefer resistive heating over heat pumps for homes. However, certain energy intensive applications like high speed air transport, rockets, and long distance shipping will probably continue to rely on hydrocarbons. The next challenge is to figure out how to produce those hydrocarbons using solar power as the primary input. This would effectively put fuel production downstream of the solar energy grid.

Batteries are currently following the same growth curve as solar technology. The production rate is ramping up quickly while costs are falling. This trend is driven by a high learning rate where battery prices drop about 35% every time production doubles. Because of this, the addressable market grows much faster than actual production every year. This suggests a continued acceleration in how many batteries are produced and used.

The battery prices are falling like 35% per doubling, which means that in any given year, the addressable market is growing far faster than production. We're only going to see continued acceleration of battery production and adoption.

Casey suggests that we are still far from reaching a saturation point for battery use. When looking at materials like lumber or steel, the United States uses several tons per person every year. We will likely reach a point where we have tons of lithium-ion batteries per person as well. This indicates that the industry is poised for explosive growth because current levels are nowhere near that scale.

While the basic battery cell is important, there is significant room for innovation in packaging and software. A cell must be encapsulated in a pack that provides safety and usability. It also needs software to give it economic utility for the user. Companies like Tesla and Basepower are leaders in this area. The industry is becoming more horizontal and fragmented, which makes it easier for new companies to enter. Entrepreneurs can now partner with existing manufacturers to get products to market quickly without needing to refine lithium from scratch.

It's no longer necessary to follow the deep tech development dream and figure out lithium refining from scratch. You can generate significant value without having to go through the hassle and the decades of pain and strife to get there.

Hyperscalers are currently desperate for power. They often require hundreds of megawatts and are willing to do whatever it takes to get it. This includes bringing in natural gas generators or buying out existing contracts for equipment like transformers. This urgency creates a clash between two very different types of organizations. On one side are the hyperscalers, who move quickly and have nearly infinite demand. On the other side are traditional utilities, which prioritize reliability and move slowly due to regulatory and infrastructure constraints.

It is kind of like an unstoppable force meets an immovable object kind of situation. It is hard to imagine two more different companies in terms of the way they do business. A utility concession has been doing one job for more than a century while a hyperscaler is running around with their hair on fire trying to get power.

Utilities are often hesitant to build massive new infrastructure for data centers. They worry about the long term risk of signing up rate payers for expensive projects when data center equipment might depreciate in just a few years. Additionally, modern permitting and land use rules make it much harder for utilities to expand their networks quickly. While some hope small modular reactors can solve the power gap, they are difficult to scale. Casey notes that solar and battery production is already operating at a terawatt scale, making it a more viable solution for the immediate energy crunch.

The future belongs to those with cheap and abundant power. Currently, the United States uses its financial strength to cover for energy gaps. However, this strategy has a limited lifespan. Within a decade, the approach must change. Instead of outbidding others for gas, companies will need to secure large areas of land and cover them in solar panels as fast as possible. Once the land is secured, the rest of the infrastructure is relatively small in comparison.

There are concerns about overbuilding or creating a bubble in AI infrastructure. Casey remains optimistic about the underlying economics. He points out that nearly 90 percent of the capital spent on a large data center goes directly to Nvidia. Because the chips are so expensive, the cost of the supporting infrastructure becomes secondary.

You can double the thickness of your road and double the size of your solar plant and double the number of free doughnuts in the break room and it won't really change your economics.

The main goal for an operator is to maximize the tokens produced per acre of land. This drive for absolute revenue dictates how land is used. Once a site is filled with solar, the focus turns to fine-tuning. Operators must then balance battery capacity and GPU power use to get the best results from their land.

Terraform operates on a model of using available solar power during the day to create cheap, high value primary materials. While this works for natural gas and chemicals, there is a massive opportunity in metals. Basalt is an incredibly common and inexpensive rock. A ton of basalt costs only fifteen dollars, but the metals inside that ton have a value of about thirteen hundred dollars. This nearly 100x increase makes it a compelling target for a company that can process material at scale.

Once you have a terraform sized hammer, a lot of problems start to look like nails. And so I'm kind of obsessed with trying to figure out what is the correct way of grinding up basalt and then sorting it out by atomic number.

Casey describes a process of grinding the rock and sorting it by atomic number to create individual piles of silicon, iron, aluminum, and other elements. While this method is very energy intensive, Casey views that as a strength rather than a weakness. This technology would be essential for building on Mars. Mars has an abundance of basalt but lacks the high quality mineral ores found on Earth. If explorers can unshuffle the minerals in common Martian rock, they can produce whatever materials they need to survive and build.

There is a common fear that data centers and artificial intelligence will break the power grid due to their massive demand for electricity. However, looking at history suggests a different path. Energy-heavy industries, like aluminum smelters, historically built their own captive power plants. We are likely moving back to this model where data centers buy thousands of acres to install solar and batteries. To ensure constant power, these facilities must overbuild their capacity, resulting in vast amounts of excess energy most of the time.

As soon as it becomes obvious in the market that the utilities are actually a net buyer rather than a net seller of power and that the local communities will end up benefiting enormously from being right next door to these basically self funding, arbitrarily green, non polluting power sources.

This shift turns local land into an economic engine far more powerful than traditional agriculture. Casey notes that in places like western Ohio, the revenue per acre for solar energy can be 100 to 1,000 times higher than growing corn or soy. This creates an opportunity for economically neglected regions to thrive by hosting self-funding, green power sources. Beyond infrastructure, building this future requires a specific culture. Casey emphasizes the importance of leveling up his team at Terraform Industries to create a legacy of talent similar to the PayPal mafia.

You're asking people to come in here and work for years of their life on a project that pretty much all the smart money thinks is impossible. So what can give these people they can take with them? Even if the worst should come to pass? What do you give them to make sure it's worth their while?

By giving employees frontline responsibility and pushing them to stretch their abilities, startups can ensure that even if a project fails, the people involved have gained invaluable skills. This intentional focus on growth ensures the team is ready to lead as the company scales.

Casey has spent the last four years running Terraform Industries, a startup focused on creating cheap synthetic natural gas from sunlight and air. The energy landscape is shifting rapidly as massive tech companies search for power. China is currently leading this change by installing more solar this year than the total electrical capacity produced by the United States. The cost of solar is dropping faster than many expected because the learning rate has steepened to 48 percent.

The learning rate for solar actually steepened quite dramatically. It's running at 48 percent which means that for every cumulative doubling of deployment, we are getting a 48 percent reduction in cost.

The future belongs to those who have access to cheap and abundant power. While the United States currently has financial advantages to bridge the gap, that advantage may not last. Casey suggests that the United States has less than a decade to maintain its superiority in this area before the global landscape shifts completely.

Casey is building Terraform Industries to produce cheap synthetic natural gas from sunlight and air. His background includes theoretical physics and work at NASA. The company uses existing legal structures to sell gas to utilities. This allows them to enter a mature market without changing how contracts work.

Energy history shows that economic output was once limited by human and animal metabolism. People could only produce what they could digest. The steam engine changed this by using coal to provide mechanical energy. This bypassed the biological bottleneck and started a period of growth that doubles wealth every few generations.

The advance of the steam engine was that it allowed us to use coal, a fossil form of energy, to increase the mechanical energy available for our civilization by bypassing our guts. It is like routing around the bottleneck of a human's ability to digest food and then do things with their muscles.

The electricity grid is a complex system where supply and demand must match at every second. When you turn on a light, the grid immediately adjusts the torque on a distant turbine. For a century, this relied on large power plants and long wires. Now, solar power and batteries are changing the industry. These technologies allow people to store energy and use it when needed. This shift is disrupting the traditional utility monopolies that have existed for over a hundred years.

Solar cells began as a niche technology for space applications in the 1960s and 1970s. For decades, they were a low carbon way to make electricity, but the costs were too high to compete with the grid. Everything changed around 2009. Solar power finally became cheaper on the margins than other forms of electricity. This gave the technology a foothold in the market.

In about 2009, something interesting happened. We obtained line of sight to solar photovoltaic power being cheaper on the margins than the next cheapest form of grid electricity, which is really all you need in order to get a toehold in that market.

This shift led the Chinese government to invest huge amounts of money into the industry. The workforce grew from a small group of scientists to tens of thousands of experts. They developed new ways to refine silicon specifically for solar panels. Casey explains that the learning rate for solar is now 48 percent. This means that every time the world doubles its solar capacity, the cost drops by 48 percent. This is much faster than the progress seen in cars or planes. It is closer to the rapid growth of microchips.

As of this year, it's running at 48%, which means that for every cumulative doubling of deployment, we are getting a 48% reduction in cost, which is about as steep a learning rate as I've ever seen in any technology.

Solar power is moving toward becoming the primary source of energy production. This shift requires the total electrification of transport and industry. Some early predictions suggested that heating and cooling would lead this transition through heat pumps. However, the adoption of heat pumps for heating has been slower than anticipated. In contrast, the automotive sector saw a massive shift because companies like Tesla created electric cars that people actually wanted to drive rather than just providing basic utility.

The exciting thing is for the entrepreneurial class, if you can figure out a way to package electrification of some fundamental technology in a way that's appealing to customers, you can laugh all the way to a trillion dollar company because it's just such an enormous sector.

Casey expects that most ground transportation and cooling systems will eventually run on electricity. He suggests that heating might move toward resistive heating instead of heat pumps. Some energy-intensive sectors like long-distance air travel, shipping, and rockets will likely continue to use hydrocarbons. The goal for the future is to produce these hydrocarbons using solar power as the primary energy source.

The real question that I have to answer here at Terraform is how long will it take us to put hydrocarbons downstream of solar power as well? That's what we're working on.

Battery technology follows a growth curve similar to solar energy. Production is increasing rapidly while costs fall. Every time production doubles, battery prices drop by 35 percent. This high learning rate means the market for batteries grows faster than the rate of production. We can expect to see a steady acceleration in battery adoption for years to come.

Batteries prices are falling like 35% per doubling, which means that in any given year, the addressable market is growing far faster than production, which means that we're only going to see continued acceleration of battery production and adoption.

Casey suggests we might eventually need tons of battery storage per person. This is similar to how we use tons of steel or lumber for infrastructure. Most of this capacity will exist outside of individual homes or cars. While the battery cells are a commodity, there is still room for innovation in packaging and software. These elements provide safety and economic value to the user.

The battery industry is now fragmented and horizontal. Companies like Tesla and Basepower are performing well, but new opportunities exist for entrepreneurs. It is no longer necessary to invent new chemistry or start from lithium refining. A new business can partner with existing manufacturers to get a product to market quickly. This allows for faster iteration and value creation without the pain of deep tech development.

Hyperscalers are desperate for power and will do almost anything to get it. They are buying out existing contracts for transformers and even bringing in gas generators to keep up with demand. This creates a clash between two very different types of businesses. On one side are tech companies moving as fast as possible to build data centers. On the other side are utilities which are designed for reliability and slow, steady growth. These utilities are often skeptical of data centers because they provide few jobs and use equipment that depreciates very quickly.

So you want me to go and build a huge amount of power lines and basically sign our ratepayers up for their care and feeding operation, financing, et cetera, essentially indefinitely on the basis of a promise to take a bunch of stuff here. But you're also very mobile, and in five years time, you'll have fully depreciated all your GPUs.

While many people look toward Small Modular Reactors as a future solution, Casey points out that they cannot scale fast enough. Building enough reactors to satisfy current demand would require producing millions of them. In contrast, solar panels and batteries are already being manufactured at a massive scale. China alone has installed more solar capacity this year than the entire United States produces in baseload electricity. This makes solar and batteries the most viable path forward for powering the next generation of data centers.

The future of global power belongs to those who possess cheap and abundant energy. While the United States currently maintains a superior position, this advantage is fragile. Financial capacity can temporarily mask the gap in energy production, but this buffer will likely vanish within a decade. The current approach to energy must change significantly to maintain a competitive edge.

If the future belongs to the people who have cheap, abundant power, there is only so long that the United States is vastly superior. Financing and financial capacity will paper over that gap, and at this rate, I would say it is less than a decade.

To address this challenge, the focus should shift away from being a marginal gas turbine bidder. The priority must be to acquire large areas of land and cover them with solar panels as quickly as possible. Most of the necessary equipment is relatively compact and can be transported in a few containers. Casey suggests that rapid assembly and massive scale in solar infrastructure are the most viable paths forward.

There is a question of whether developers might overbuild or overspend on infrastructure for large data centers. Casey believes the risk is low because of the specific way these projects are funded. About 90% of the total capital expenditure for a large scale data center goes to Nvidia for hardware. This creates a unique economic situation where the cost of the physical infrastructure is relatively small compared to the cost of the chips.

You can double the thickness of your road and double the size of your solar plant and double the number of free doughnuts in the break room. And it won't really change your economics.

The main goal for an operator is to maximize the number of tokens produced per acre of land to increase absolute revenue. Once a developer has used all available land for solar power, they must focus on a different kind of optimization. They have to balance battery size and GPU power consumption to ensure they are getting the best possible return on their investment.

A trillion dollar market exists within common basalt. This volcanic rock is incredibly cheap and abundant, yet it contains a vast array of metals. While a ton of basalt costs roughly fifteen dollars, the total value of the metals hidden inside that same ton is about thirteen hundred dollars. This represents nearly a hundredfold increase in value if the materials can be extracted efficiently.

I can buy a ton of basalt for 15 bucks. And the net present value of all the metal in that ton is about $1300. So almost 100x increase. Once you have a terraform sized hammer, a lot of problems start to look like nails.

Casey explains that the challenge lies in sorting the rock at the atomic level. The goal is to grind up the basalt and sort it by atomic number to create separate piles of silicon, iron, aluminum, potassium, and titanium. This process is inherently energy intensive. However, using solar power to drive this extraction makes it a viable path for primary material production. This approach becomes especially critical for space exploration. On Mars, high quality ores may be scarce, but basalt is everywhere. Mastering the ability to unshuffle these minerals would allow a Martian base to produce whatever materials it needs from the ground beneath it.

People often worry that data centers and AI will break the power grid or destroy the world. However, human progress has made life significantly safer and better. Casey notes that in the 1800s, it was normal for people to die on ship voyages to Australia. Today, a single death on a flight would be a tragedy. This progress continues as we find new ways to power our technology.

The future of data centers will likely mirror the history of aluminum production. Large industrial facilities used to build captive power plants, like hydroelectric or coal plants, right next door. Casey expects data center developers to buy large plots of land and cover them with solar panels and batteries. To ensure the facility runs 99.9% of the time, they must overbuild these power systems. This creates a massive surplus of energy that sits unused for most of the year.

You could have a data center that has effectively 99.9% utilization with solar and batteries. What that means is that 99% of the time it has excess power and that power is essentially just not being used at all. It's just sitting there.

This excess power changes the relationship between industrial projects and the utility grid. Instead of being a drain on the system, these sites become net sellers of green energy. This shift can transform the economy of rural areas. In places like western Ohio, solar power can generate 100 to 1,000 times more revenue per acre than traditional crops like corn or soy. Local communities benefit from being next door to these self-funding, non-polluting power sources.

Building this future requires a dedicated team willing to work on projects that others think are impossible. Casey focuses on leveling up his staff at Terraform Industries to create a Terraform mafia. He wants employees to gain intense, frontline responsibility. This ensures that even if a startup fails, the workers have gained valuable skills and experience that stay with them.

We actually need to be extremely intentional here about making sure that we're leveling people up. We're building the PayPal mafia or the Terraform mafia in real time.