Energy needs of GPU clusters

The following is a conversation with David Kirtley, a nuclear engineer, expert on nuclear fusion, and the CEO of Helion Energy, a company working on building nuclear fusion reactors and have made incredible progress in a short period of time that make it seem possible, like we could actually get there as a civilization. This is exciting because nuclear fusion, if achieved commercially, will solve most of our energy needs in a clean, safe way, providing virtually unlimited clean electricity. The problem is that fusion is incredibly difficult to achieve. You need to heat hydrogen to over 100 million degrees Celsius and contain it long enough for atoms to fuse. That's why the joke in the past has been that fusion is 30 years away, and always will be. Just in case you're not familiar, let me clarify the difference between nuclear fusion and nuclear fission. By the way, I believe according to the excellent Sample Size subreddit post by pmgoodbeer on this, the preferred pronunciation of the latter in the U.S. is nuclear fission, like vision. And in the U.K. and other countries is nuclear fission, like mission. I prefer the nuclear fission pronunciation because America. So today's nuclear power plants use nuclear fission. They split apart heavy uranium atoms to release energy. Fusion does the opposite. It combines light hydrogen atoms together, the same reaction that powers the sun and the stars. The result is that it's clean fuel from water, no long-lived radioactive waste, and inherently safe because a fusion reactor can't melt down.

Let's start with the big picture. What is nuclear fusion, and maybe what is nuclear fission? Let's lay out the basics. So fusion is what powers the universe. Fusion is what happens in stars, and it's where the vast amount of energy that we even use today here on Earth comes from the process of fusion. It also is what powers plants. And those plants become oil, and those become fossil fuels that then power the rest of human civilization for the last 100 years. So fusion really underpins a lot of what has enabled us as humans to go forward. However, ironically, we don't do it actively here on Earth to make electricity yet. And so fundamentally, what fusion is, is taking the most common elements in the universe, hydrogen, and lightweight isotopes of hydrogen and helium, and fusing those together to make heavier elements.

So in fusion, you take these lightweight isotopes like hydrogen and deuterium, and as you combine them and get these molecules closer and closer together, some really interesting fundamental physics happens. First, these atomic nuclei are charged. They have an electric charge, and like charges repel. I think everybody is familiar with that, where you take two positive charges and you try to push them together, and the electromagnetic force between them repels them. So you have a force that's actually pushing against them. In fusion, you work to get your fuel very hot, very, very high temperatures—100 million degree temperatures. And temperature really is kinetic energy; it's motion, it's velocity. So these particles are moving so fast that even though they're coming together and there's this repulsive electromagnetic force, they can still come close enough that another force comes into play, which is the strong force. Once you get within a very close distance, on the order of the scale of those nuclei themselves—of those atomic nuclei, the tiniest thing you could imagine, and probably way smaller than that—these particles then are attracted to each other, and they combine and fuse together. At that point, you create heavier atomic nuclei that have a slightly less mass, slightly less total mass in the system, and that mass equals MC² as energy.

...fundamentally as a source of energy. In fusion, you're taking these lightweight isotopes, you're bringing them together, you're releasing energy, and that energy is in the form of charged particles. It's already in the form of electricity. Fusion itself has electricity built into it without a lot of the steam or thermal system requirements. That's a really nice fundamental benefit of fusion itself. Also, this reaction that's really hard to do turns itself off, so you end up with fusion being fundamentally safe, and that's really a key requirement of any industrial system, is that it turns itself off and is safe. You turn the key off on your car, you know it's going to turn off. I guess the flip side of that, just stating the obvious, but it's nice to lay it out. For nuclear fission, it's a chain reaction, so it's hard to shut off, and it works by boiling water into steam, which spins turbines and produces electricity. Can you talk through this process in a nuclear fission reactor?

Well, I have to ask then, what do you think happened in Chernobyl? What lessons do we learn from the Chernobyl nuclear disaster and maybe also Three Mile Island and Fukushima accidents? I think you're suggesting that it has to do with the humans a bit. So with Chernobyl and Fukushima, I actually put Three Mile Island in a different category. In fact, some of the recent news in the last year is that we're going to be restarting Three Mile Island, because there's such a need for clean base load power. So that's actually a very interesting other topic we should talk about is, is why and how we're doing that. But more than that, going back to the accidents that did happen in both of those systems, you can point to the human failure rather than the engineering failures of those systems. That in Fukushima specifically, there were multiple nuclear fission reactors on the same site that successfully kept running through the tsunami, totally successfully, and were only later shut down for more political reasons.

So, this might be interesting to ask on the geopolitics side of things. I have the chance to interview a few world leaders coming up. By way of advice, what questions should I ask world leaders to figure out the geopolitics of nuclear proliferation? nuclear weapons, nuclear fission power plants, and nuclear fusion power plants? What's the interesting intricate complexity there that you could maybe speak to? The question I would want to ask is, "What would you do?" If we could deliver for you low-cost, clean, industrial scale, tens or hundreds of megawatts of fusion power that's low-cost, clean, base load and doesn't have the geopolitical consequences of uranium and plutonium, of fissile material, what would you do there? How would that change your view of the next 30 years?

Okay, so let's return to the basic question. We already mentioned it a little bit, but is nuclear fusion safe? So, are the fusion power plants we're talking about safe? Yes, fusion power is fundamentally safe. The physics and the reactions of the fusion system itself mean you don't have runaways. We've talked about some of the human factors around power plants, power systems, and industrial scale systems, and that's something that we build into the design of these from today. We look at how these systems might fail. In fact, some of the analysis we do, we did this analysis for the Nuclear Regulatory Commission over the last few years, looking at how to regulate fusion power. As we're building the first fusion power plant, we need to make sure we're regulated safely. So we spent a lot of time doing the technical and political case in the United States of how to regulate fusion.

So high level, what are the different ways to build a nuclear fusion power plant? Can you explain what a tokamak is, what a stellarator is, and what's the linear approach that Helion is using? There are a number of ways to do fusion. Fundamentally, in all fusion approaches, you're trying to do the same physical process, which is take these lightweight isotopes, heat them up so that they can move at high velocity—over 100 million degrees—bring enough of them together. We call it density. Enough of them together in a certain volume so that you have reactions happening at a higher rate, and keep them together long enough that they are able to collide into each other and do fusion and release energy. That's the fundamental core. Now, how you do that, how you bring those particles together, how you hold them together long enough, there's a wide range of technologies that, as humans, we've been exploring since the 1950s.

So you're trying to reach 100 million degrees. How do you get to that temperature fast? And by the way, what can you say to help somebody like me understand what 100 million degrees is like? It seems insane. What does that world look like? I guess just everything is moving really fast. Like you said, you can't put anything mechanical in there. Yeah, so a couple of key things happened. When gas is that hot, there's... We talk about the states of matter. You have solids, where ice, it's cold. The atoms are now bound in a lattice structure together. They're held together. And then liquid, you've broken a lot of that lattice structure. They can move around. They have some kinetic energy, but they're still pretty contained, they stay in the bowl. Keep heating it, now you're in gas. And now these particles are free to move around. They're moving around, they're bouncing off of each other all the time, and you can keep heating it from there, and that's where we talk about some more phases of matter. We can add a little bit more physics here. We talk about rarefied gasses.

And there's a computer managing all this. How do you even program these kinds of systems to do the switching? Is there some innovation required there? I'm continuously amazed by what the pioneers in fusion were able to do before the computer existed, because they had to control things at this scale. But maybe it was pretty hard, and why we've been able to take what they did and build on it, is because now we use modern gigahertz-scale computing to be able to do this. Even when I started my career, we talked about megahertz processors. Megahertz is microseconds. That's great. You're kind of at the border of fast enough, but you can't do computation at that speed if all it can do is respond in one microsecond. But now gigahertz means I can do a thousand operations in that one microsecond, so I can do more useful things. So we use mostly...

All right. So there's a million questions there. So first of all, how much understanding do we have about how many collisions happen? Can we go to the fusion? How many collisions are there and how does that map to the electricity? And maybe can you just even speak to the directly mapping to the electricity, which is one of the differences between this approach and the tokamak approach? So how much fusion do you get out from these systems? And that's really the key question. So we already talked about beta, that B squared, the magnetic pressure is equal to NKT, N being the density, T being temperature. And then we talked about fusion, where your goal for fusion is to get particles hot, high temperature, get enough of them together, density. And then you want to get them together long enough. We call that tau. So N, T, and tau, long enough that fusion happens and a lot of fusion happens, more than any of the loss rates that are happening, NTT. And in beta with B squared, you know already two of those parameters, NNT, are equal. And so that tells you right away the goal is to maximize magnetic field, absolutely maximize magnetic field.

Yeah, this is really exciting and really inspiring. So I have to ask then, what timeline do you think, like first working, out there, nuclear fusion power plant? When do you think? Yeah, so what we've been able to do is rapidly build, every few years, bring a new fusion system online. In 2023, we signed a deal with Microsoft to build a power plant for Microsoft for one of their data centers. This is a power plant that is plugged into the grid, generating electricity from fusion, with a very tough, ambitious timeline of 2028 for the first electrons from that power plant.

So the Kardashev scale: a Kardashev Type One civilization is when humans are either catching or generating as much power as what's incident on the Earth from the sun. Type Two is the next big one, where you're catching as much energy from all the way around the sun, so massive amounts of energy. A lot of times, people talk about it as incident, as in you had solar panels the size of the entire planet blocking all of the sun. But I think really, you should be thinking about it as what can we generate? What can we make here on Earth? And what we know is that, you know, we're only a fraction right now of Kardashev Type One, and we've got some work to do. And there are not a lot of technologies that can get there, just from the point of view of the fuel. Right.

When you look out there at the stars, I'm really confused by what's going on, because I think there is for sure thousands, if not millions, of advanced alien civilizations out there. I'm really confused why we have not, in a definitive way, met any of them. So again, continuing the pothead questions, what energy source do you think they're using? If what I'm saying is true, that there is alien civilizations out there, do you think it's, like, pretty certain that they, in order to expand out into the cosmos, they would be using nuclear fusion? It's hard to imagine anything else. Right now, where does energy in the universe come from? It comes from fusion. It comes from stars, and we know that that's the process. So whether they're harnessing the star itself, Kardashev Type II, or are they bringing fusion along because they want to go somewhere and they're bringing it with them to go visit, I think that's pretty likely. You bring up the Fermi paradox.