
41 segments available
How will we feed the 100s of GWs of extra energy demand that AI will create over the next decade? On this episode, Casey Handmer (Caltech PhD, former NASA JPL, founder & CEO of Terraform Industries) walks me through how we can pull it off, and why he thinks a major part of this energy singularity will be powered by solar. His views are contrarian, but he came armed to defend them. 𝐄𝐏𝐈𝐒𝐎𝐃𝐄 𝐋𝐈𝐍𝐊𝐒 * Transcript: https://www.dwarkesh.com/p/casey-handmer * Apple Podcasts: https://podcasts.apple.com/us/podcast/china-is-killing-the-us-on-energy-does-that-mean/id1516093381?i=1000722101753 * Spotify: https://open.spotify.com/episode/3vV7Vki2qJhO2AyodIaTMo?si=Gv0aEF9mSSu5cmPMksndRg 𝐒𝐏𝐎𝐍𝐒𝐎𝐑𝐒 * Lighthouse helps frontier technology companies like Cursor and Physical Intelligence navigate the U.S. immigration system and hire top talent from around the world. Lighthouse handles everything for you, maximizing the probability of visa approval while minimizing the work you have to do. Learn more at https://www.lighthousehq.com/employers To sponsor a future episode, visit https://www.dwarkesh.com/advertise 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 00:00:00 – Why doesn’t China win by default? 00:08:28 – Why hyperscalers choose natural gas over solar 00:18:01 – Solar's astonishing learning rates 00:27:02 – How to build 50,000 acre solar-powered data centers 00:40:24 – Environmental regulations blocking clean energy 00:44:04 – Batteries replacing the grid 00:49:14 – GDP is broken, AGI's true value must be measured in total energy use 00:58:45 – Silicon wafers in space with one mind each
Casey Handmer discusses the competitive landscape of AI development, emphasizing the importance of solar panel and battery manufacturing. He highlights China's dominance in solar production and questions whether the U.S. can compete effectively. The conversation explores the implications of capital allocation and industrial strategy in the context of energy demands for AI.
"Today I'm interviewing Casey Handmer. Casey has worked on a bunch of cool things. Caltech PhD on some gravitational wave black hole gimmick stuff, then Hyperloop, then the Jet Propulsion Laborato..."
Handmer explains China's energy production challenges, noting that only a third of its energy use comes from electricity. He discusses the potential of synthetic fuels to transform this dynamic, suggesting that advancements in this area could level the playing field between the U.S. and China in the energy sector.
"Do you think that China is better at capital allocation than the United States? Do you think the Chinese business environment is better for business than in the United States? I feel you can make ..."
The discussion shifts to the socio-economic landscape in China, where Handmer argues that despite the presence of poverty, certain regions are wealthy and innovative. He warns against underestimating China's capabilities while also emphasizing the U.S.'s potential to compete effectively in the energy sector.
"Synthetic fuels have been around for 100 years. There are projects in China right now working on synthetic fuels. It would not surprise me if they were thinking pretty seriously about this. Just t..."
Handmer elaborates on the strategic importance of energy in the AI race, comparing U.S. and Chinese capabilities. He discusses the implications of export controls on chips and how similar strategies could affect solar and battery production, highlighting the interconnectedness of the two economies.
"Like the Indian middle class is larger than the US middle class. But also it's nowhere near as wealthy. Whereas there are parts of China which are humongous, which are actually as wealthy as the ..."
The conversation focuses on the U.S.'s potential to ramp up solar manufacturing. Handmer argues that with the right motivation and investment, the U.S. could significantly increase its solar capacity, challenging the notion that China has an insurmountable advantage in this area.
"If we decided we want to produce 100 gigawatts of solar capacity every single year… We’re already on track to do that. Is it going to be as cheap as it is to do in China? My views on this are so..."
Handmer critiques the current trend of hyperscalers opting for natural gas over solar energy for data centers. He questions the rationale behind these decisions and emphasizes the importance of solar energy's learning rates and future potential in powering large-scale data operations.
"It's mostly a case of putting in a phone call to all the different manufacturers here, in Germany, and so on and saying, "We need you to 10x the size of your factory, starting today, blank check, ..."
The discussion delves into the logistical challenges of energy production for data centers. Handmer explains the constraints faced by companies in securing natural gas and the implications for future energy strategies, emphasizing the need for a diversified energy approach.
"That's actually something that Elon and his companies are great at. It's figuring out this mass production, semi-automated mass production. They've got this facility in Texas which is making the ..."
Handmer discusses the complexities of energy delivery costs and the challenges faced by utilities in maintaining infrastructure. He highlights the impact of regulatory regimes on the efficiency of energy distribution and the potential for large-scale pruning of grids.
"our ability to go and rent gas turbines?" No, they're not, because there was enough available once, maybe twice. But at a certain point, you realize as you grow, you start to touch all these addi..."
The conversation shifts to the future of energy production for AI data centers. Handmer argues that large industrial operations may need to build their own power plants to ensure reliable energy supply, drawing parallels to historical practices in heavy industry.
"The problem that we see—and the reason that PG&E here in California, for example, is perpetually on the brink of bankruptcy—is that even though the cost of an additional solar panel or additiona..."
Handmer concludes by analyzing the decisions made by hyperscalers regarding energy sources. He discusses the balance between energy availability and production capacity, emphasizing the need for strategic planning in the face of growing energy demands.
"Okay, AI might be a special case. But big picture question. Across different kinds of ISOs, from Texas to Pennsylvania to whatever, people are building data centers which will not be online for ..."
This segment delves into the anticipated demand for solar energy as data centers expand. Handmer argues that the learning rates for solar energy production are significantly more favorable than for natural gas, predicting a shift towards solar as the primary energy source for future data centers.
"At the same time, what is the economic value to you of using Claude or Grok or whatever you use on a monthly basis? A lot. It's obviously much more than the subscription, but is it 10 times more ..."
Handmer discusses the impressive learning rates of solar technology, citing Wright's Law and the rapid cost reductions associated with increased production. He emphasizes that the demand for solar energy is likely to continue growing, countering conventional wisdom that suggests solar adoption will plateau.
"In 2032, we're going to have hundreds of gigawatts of extra demand for data centers and at that point, most of it is coming from solar? Why is that? There aren't enough turbines being manufacture..."
In this segment, Handmer argues that solar adoption is accelerating, contrary to popular belief. He explains how the increasing efficiency and decreasing costs of solar technology will lead to a surge in demand, positioning solar as a dominant energy source for the future.
"So it’s roughly 15-20% per year. Then just as a result of that price reduction, demand skyrockets by probably six times more than that additional marginal production capacity increase. This is on..."
Handmer addresses the challenges faced by natural gas producers in scaling up production to meet future energy demands. He contrasts this with the more favorable outlook for solar energy, suggesting that the uncertainties surrounding natural gas investments may hinder its growth.
"We’re still in the Apple II computer era of solar. Backing up, if the story is that the reason solar is getting cheaper is because there's a lot of demand for more solar, and that demand can sustai..."
This segment explores the transition towards solar-powered data centers, with Handmer predicting that by 2027, the majority of new data centers will rely heavily on solar energy. He discusses the implications of this shift for energy production and consumption in the context of AI.
"you also have to then operate that plant at that capacity for 20 years. If I was looking at the same charts as they're looking at right now, I'd say, "What are the odds that in 25 years' time we..."
Handmer discusses the land requirements for solar energy production, addressing misconceptions about land availability. He emphasizes that there is ample land suitable for solar farms, particularly in regions like Texas, and outlines the logistical challenges of permitting and interconnecting solar installations.
"If a gas plant is still making money, people will keep operating it. But at a certain point… It is the case right now that operating a coal plant costs more than building a new solar plant. So it..."
In this segment, Handmer elaborates on the importance of solar energy for AI infrastructure, highlighting the need for reliable power sources to support data centers. He discusses the potential for solar to provide excess energy that can benefit local utilities, framing solar as a sustainable solution for future energy demands.
"10x more H100 equivalents in the world. About a kilowatt each, something like that. Okay, so that's like 100 gigawatts. That sounds roughly right. You're not the first person to give me a call and..."
Casey Handmer discusses the necessity of achieving four nines of uptime for AI data centers powered by solar energy. He explains that to ensure reliability during winter months, solar installations must be overbuilt, allowing excess energy production that can benefit local utilities. This segment highlights the importance of solar energy in meeting the increasing demands of AI infrastructure.
"No. Doesn't matter. Do trees grow on mountain slopes? So it doesn't matter. For reference, Nevada is something like 80 million acres. Just Nevada, which is like 90% federal land, is 80 million ac..."
In this segment, Handmer envisions a future where vast fields of solar arrays are integrated with data centers and battery storage. He emphasizes the efficiency of this setup and how it can provide a self-sufficient energy solution. The discussion touches on the scalability of solar energy and its potential to revolutionize energy consumption for AI applications.
"you need to have a lot of solar overbuilt. Is solar overbuilt a bad thing? No. Is the fact that we produce 40% more food than we need a bad thing? No. It's much better than producing 40% less tha..."
Handmer breaks down the land requirements for large-scale solar farms needed to support AI data centers. He compares the land needed for solar installations to historical land allocations for nuclear projects, arguing that the space required is manageable. This segment underscores the feasibility of deploying solar energy at scale to meet future energy demands.
"You're saying solar gets so cheap that it's the way we'll treat hard drive space. We get a bunch of excess. Also the market will be made at the new marginal consumption and production. All the peo..."
This segment focuses on the optimization of energy storage and consumption in solar-powered data centers. Handmer explains how advancements in battery technology can reduce the need for extensive solar panel installations, allowing for more efficient energy use. He discusses the balance between solar energy production and the operational needs of AI systems.
"You could even use microwave links if you really wanted to. You could use Starlink if you really wanted to. I don't know if Starlink would be fast enough. I'm not sure if it's capacity is high enou..."
Handmer highlights the regulatory hurdles that hinder solar energy deployment in the U.S. He critiques the environmental regulations that complicate the installation of solar farms, arguing that they often outweigh the environmental benefits of solar energy. This segment emphasizes the need for regulatory reform to facilitate the growth of renewable energy sources.
"compromising your inference or training. Okay, so you've got, say, a Tesla Megapack, something like four megawatt hours. So one megawatt rack, and then six Tesla Megapacks, each of which is rough..."
In this segment, Handmer discusses the potential role of diesel generators in reducing the solar panel requirements for energy production. He explains how backup generators can provide necessary power during peak demand times, allowing for a more flexible energy strategy. This highlights the interplay between traditional and renewable energy sources in future energy planning.
"he said that if you have diesel generators or something which can take over 10% of the generation during winter, then you can have a 60% reduction in the amount of solar panels you need to insta..."
Handmer compares the capital expenditure of solar energy projects to that of gas turbines, arguing that while solar farms require significant upfront investment, they offer long-term benefits. He discusses the financial implications for tech giants like Meta and Microsoft in choosing between renewable and fossil fuel energy sources. This segment sheds light on the economic considerations driving energy decisions in the tech industry.
"The thing that Meta has realized is that Zuck is running out of time to spend his money to win. The capex is not crazy high, just to be clear. The capex is still dominated by just the GPUs. How muc..."
Handmer addresses the rising electricity prices and their connection to regulatory issues. He argues that current regulations are stifling innovation in renewable energy deployment, particularly in the U.S. This segment emphasizes the need for a shift in policy to harness the full potential of solar energy and meet future demands.
"firehosing energy at a given problem because it rains down from the sky. Between the fact that maybe solar prices will go down and the fact that demand is going to go up. Do you think electricity ..."
In this segment, Handmer critiques the environmental impact review process that complicates solar energy projects. He argues that the regulations are outdated and often counterproductive, hindering the deployment of solar technology. This discussion highlights the need for a more balanced approach to environmental protection and renewable energy development.
"In terms of things that are causing us to lose to China, tariffs are neither here nor there because as we've discussed, we're not sensitive to cost on power. But the environmental regulations tha..."
Handmer stresses the urgency of transitioning to solar energy to avoid economic decline and environmental disaster. He outlines the potential consequences of failing to move away from fossil fuels, including economic instability and climate change impacts. This segment serves as a call to action for embracing renewable energy solutions.
"go through this whole process. If there's one thing that anyone listening to this can do, it would be to have a categorical exemption for solar deployment. Or if I put money in an escrow account t..."
In this final segment, Handmer discusses how batteries can address the challenges faced by the electrical grid. He explains the concept of temporal arbitrage and how batteries can store energy for later use, providing a viable alternative to traditional grid systems. This segment concludes with a vision for a future where renewable energy and battery technology work in harmony.
"We also need to do sulfur injection and a couple of other things. People will point out that transmission line growth has been stuck in a rut for decades. We have all these bottlenecks in terms o..."
Handmer discusses the increasing per capita allocation of batteries and their role in energy storage. He highlights the efficiency of batteries in managing solar energy fluctuations and predicts a future where batteries are ubiquitous, enhancing energy accessibility and reliability.
"had of meaningfully storing energy, storing electricity on the grid, was pumped hydro. That only works in a handful of places and with limited capacity. It doesn't work all that well either. The ..."
This segment explores the potential energy demands of AI in the coming decades. Handmer envisions a future where AI's energy consumption could lead to significant changes in how we generate and utilize energy, particularly through solar power.
"That trend is only going to continue. We've got batteries that are performing this temporal arbitrage. The sun comes up every day, right? So the power swings from midday—you're otherwise curtailin..."
Handmer argues that the true value of AGI should be measured in total energy use rather than traditional economic metrics like GDP. He discusses the potential for AGI to automate vast amounts of labor, creating unprecedented economic shifts and energy demands.
"in the future pretty radically. It's already decreasing. It's going to continue to decrease. It's especially helpful for solar, but solar is the one that's most intermittent. You can predict the a..."
In this segment, Handmer delves into the industrial challenges of meeting the energy demands of AGI. He emphasizes the need for substantial advancements in solar panel production and energy infrastructure to support the anticipated growth in AI capabilities.
"In that asymptote, I want to get to the crazy nerd sci-fi…. What does our civilization look like? What is happening? Kardashev Level 1. Let's wait to get to turning the entire earth into an AI f..."
Handmer discusses the limitations of GDP as a measure of economic value in the context of AI. He suggests that the true impact of AI on society may not be reflected in traditional economic indicators, as the value generated could far exceed current measurements.
"Human labor generates on the order of $60 trillion of economic value. That's how much is paid out in wages to labor around the world. So that's what AGI can do. Even if you curtail it to just wh..."
This segment highlights the disconnect between energy consumption and GDP. Handmer argues that as AI becomes more integrated into the economy, the focus should shift towards energy use as a more accurate reflection of economic health and productivity.
"Right now, the AI revolution is about routing around cognitive constraints, that in some ways writing, the printing press, computers, the Internet have already allowed us to do to some extent. A..."
Handmer speculates on the future job landscape as AI takes on more roles traditionally held by humans. He discusses the implications for wages and employment, suggesting that AI could lead to both increased productivity and significant economic disruption.
"But if you don't have oil, then you have these oil shocks, which cause double digit decreases in GDP. So the elasticity of demand often matters more than its raw fraction contribution to GDP. Anyw..."
In this thought-provoking segment, Handmer explores the relationship between cognition and energy consumption. He posits that as AI technology advances, the energy required to support cognitive functions will become a critical factor in economic development.
"Again, GDP will see this huge deflation because the variable cost of running AI will just be pretty cheap as compared to paying humans wages. At the point where you've got a mixed economy with an..."
Handmer discusses the potential for silicon-based technologies to revolutionize AI. He envisions a future where silicon plays a central role in both energy generation and computational power, paving the way for advanced AI systems.
"Exactly, if we think that the value of cognition is going to be unbounded, and the way to derive cognition—to the extent you think solar will eventually win—you can derive it from how much land ..."
In this segment, Handmer speculates on the future of AI in space, discussing the possibilities of integrating solar technology with computational systems. He presents a futuristic vision of AI operating in space, powered by solar energy and advanced silicon technologies.
"If you do that calculation, that's 50x1000, so 50,000 AI souls off of one acre? It could easily be much more than that because neurons are much slower than transistors, obviously. Probably 10 ye..."
Handmer concludes with a visionary perspective on the infrastructure needed for future AI systems. He describes a scenario where minimal physical resources are required to support advanced AI, emphasizing the potential for solar energy to drive this transformation.
"You can actually make do with a buck converter or with relays or whatever to match the current output of your solar array with the charge state of your batteries and the power consumption of you..."
Handmer presents a thought-provoking perspective on the energy singularity, where unbounded cognition meets energy supply. He theorizes about the evolution of complexity in organisms and how this might lead to a simplification of energy-to-cognition processes. This segment challenges conventional views on AI development by emphasizing the critical role of energy in achieving advanced cognitive capabilities.
"Making solar arrays, making chips is a multi-stage process. Basically you start off with silicates, which are rocks ideally in a relatively pure form. You chemically reduce them. A couple of dif..."
In this closing segment, Casey Handmer invites listeners to consider career opportunities at Terraform Industries, where innovative solutions for synthetic natural gas and other materials are being developed. He emphasizes the unique environment of the company, which focuses on ambitious projects and the potential for groundbreaking technology. This call to action highlights the importance of skilled individuals in shaping the future of energy and AI.
"We have fusion in stars and the inky blackness of space and that provides our temperature gradient. Then the most efficient way to convert that into usable cognition is silicon. Literally electron..."