Please welcome CEO of Crusoe, Chase Lochmiller, CEO of Alo Atomics, Matt Lozac, and Vanity Fair contributor Tom Doten.
All right, good morning everybody. Thanks so much for coming out here. I know it's early for a lot of folks, so we're going to keep it high energy. But I want to introduce you guys to Matt and to Chase. And we're going to talk about the energy that goes into all of the models and data centers and all the things that make all this agentic stuff that you guys have been dealing with for the last couple of days happen. And so I figure, you know, the easiest way to get into this conversation is just give a brief description of the companies you work at and kind of where you fit into the energy equation and then we can dive right in. So if you want to go ahead, Matt, and kick us off.
All right. Hello everyone. My name is Matt Lozac, co-founder and CEO of Alo Atomics. And so what we're doing is we're mass manufacturing nuclear power plants that are purpose-built for powering AI data centers such as Crusoe data centers. And so we got started around two and a half years ago. We were two people back then. Now we are 200 and we are about to move into a million square foot factory in Austin, Texas. Expanding from our current 40,000 square feet to begin to mass produce our hardware. One other cool claim to fame we have is that we turned on our first reactor eight weeks ago and this was the largest reactor that any US nuclear company startup has turned on in history. So it was a 10 megawatt worth of fuel that we achieved criticality on. It was a self-sustaining chain reaction and it was the coolest thing I've definitely done in my life.
And you studied nuclear physics in college too. So, this is like you're finally getting to prove your professors that you're doing it.
Yeah. Exactly. It all paid off.
Yeah. Yeah. Yeah. This isn't just theoretical. And Chase, yeah, just to give us a brief rundown on...
That was an exciting moment for us as well because Crusoe is going to be running a small modular AI factory using that power next summer. So, it'll be the first SMR powered AI factory that we're really excited to partner with Matt and his team to bring to life. But you know I'm the CEO and co-founder of Crusoe. Crusoe is a vertically integrated AI infrastructure platform. We do sort of a whole range of stuff ranging from energy development, both building and operating power plants when we need to but also working with key energy infrastructure partners like Matt and his team that are bringing to life new forms of energy to help power AI. We also build, operate, and very large AI data centers. I know they're like super popular right now. So I was expecting a big... no, just kidding. But we are very thoughtful about when we're building these big campuses where we do try to take a very community-first approach in terms of all the benefits that we're bringing to communities in terms of helping bring new energy to the community to drive down overall power costs, helping bring new jobs. I mean, you know, a lot of these communities that we're building in, we're increasing the overall tax revenues of the community or the county or the city by anywhere from 30 to 100%. So, these are transformative investments that we're making into these communities from a job creation standpoint, from what it means for the schools, for the fire departments, for the police departments. They're very, very large investments that we've been really thoughtful community partners on. And on top of those, we build software that operates these AI factories. And this is really how you actually turn these big investments into value for customers. We operate this large cloud platform called Crusoe Cloud that is really focusing on enabling the key innovators across the AI ecosystem to build and train foundational models across large language models or voice models or world models and robotics, to scaling large AI applications that are using AI to provide services to the economy. So it really is a business that's serving everything from the electron to the token. But today we're very focused on today's panel, we're talking about energy, very excited about the new ways that AI is serving as a catalyst for new energy technologies like Matt's.
Yeah. So I want to actually build our way up the stack, maybe ending with this topic about data centers and the way you guys are addressing communities and the current public reaction of this stuff. We'll get to that. I want to just talk about energy infrastructure and demand because one of the most fascinating things for me as a reporter is learning about the fact that as these data centers are being built up at massive scale, the number one bottleneck that people talk to me about is energy and the American energy grid was not set up essentially for this kind of surge of demand to be able to power the training and inferencing of these AI models. So we can start with you Matt, like what did you sort of see in developing this company that you were like we need to start building nuclear energy generators at a scale so that we can address this problem? Like what was your... how did you address this moment of like we're not able to keep up in terms of demand.
Well so when we started the company it was pretty fortuitous timing. If you look at when we started hiring and you overlay that chart with the Google search trend for AI as a search query, it's like the exact same day that we started the company. So it was always on our radar that we wanted to power data centers as a top priority. But I think no one could have foretold just how much demand there will be from data centers even three years ago which is amazing to say.
Do you have actually got top of mind like what the demand, how it spiked, like what the expectation was and where it ended up being to give a sense as to...
I feel like Chase might know the exact number but my perception would be, you know, maybe if you look at the past 25 years, data centers used to be a few megawatts right 25 years ago then it was 10 megawatts...
Massive, that was like a huge campus, something that was in the megawatt range.
Yeah. Yeah. Yeah. And then, you know, went up to 20 megawatts and then it kind of big was then, you know, 50 and now, you know, Crusoe, they're building these multi-hundred megawatt sites, gigawatt sites...
Gigawatts, multi-gigawatt sites, yeah. I mean, if you look at maybe one of the good statistics to share, if you look at the entire data center industry, it really operated in these big hubs. Northern Virginia is like widely considered to be the biggest, most critical hub to operating web applications and sort of the internet. End of 24. So this is like partially through the AI boom, or like early innings of it. It was like 4.5 gigs of total capacity in Northern Virginia.
Putting that in context, Abilene is 2.2 gigawatts. It's half the size.
That's one campus. That's one campus. Our plan is to grow it. Yeah, so it really has been a step function change in demand and I think what you're seeing from hyperscalers is like all these forecasts are continuously getting increased right? It's like, here's our forecast for 27, 28, 29, here's what we're going to spend, oh demand is way in excess of what we thought, we're actually going to need more and energy is actually this really critical bottleneck in terms of how we actually bring that infrastructure to life.
Yeah. And I mean in terms of the... sorry I got a call here. I should have muted my phone. In terms of like, you know, in the old world, in the old school regular data center days when you were building up a data center you essentially weren't responsible for creating new power right? You would build these things and you weren't like, oh yeah and also can you please build the power plant that's going to run it? I mean you would kind of tap into what the utilities had available right?
Yeah, I think that's right. Yeah. You know, like it was, you would work with the utilities. I think there were a couple things that were different in like the old days so to speak. Number one is that the amount of power being demanded by the data centers relative to other utility consumers was a smaller percentage. That's number one. And number two is the urgency, like the planning time frame that we had to actually execute this was like, having a five-year project was not uncommon. Today it's like everybody wants a gigawatt AI factory yesterday and what are the ways that we can accelerate that development and progress. So there's both greater urgency, greater scale in terms of how it's impacting local utilities and so that's really led to a place where you have ultra well capitalized partners that are making these investments and I think people have realized that the way to make this happen is actually we need to invest in the energy infrastructure to upgrade the grids, to upgrade transmission and distribution but also add net new generation that ultimately will lead to having the energy needed to power these data centers and also as a byproduct of that you end up with energy abundance where all local rate payers can benefit from lower energy costs and more energy available which I think is something that we're really excited about and independent of what some of the doomers on social media posts may have you believe, the data shows that where data centers go, typically energy costs come down for local rate payers because of those investments that are being made.
Yeah. Let's talk a bit about Abilene because my understanding is when you guys were building that campus, you guys were providing the energy for it maybe temporarily before there was some grid energy brought online and that was all to be like diesel power generator.
Not diesel, natural gas. Sorry. I'm sorry. Yeah. Diesel is the diesel generators in Memphis. But...
Yeah. Anyway, that's a Memphis issue. But I mean tell me about... can this kind of brings us into nuclear, like why is natural gas... there are limitations with that, it's a more finite resource. I mean is it harder to bring new natural gas online versus nuclear? I mean tell me evolution from one to the other.
Yeah, I mean look, at Crusoe we try to take a very all-of-the-above energy approach to basically activating megawatts for computing and that in certain cases involves natural gas, in certain cases involves solar plus batteries where we, in Nevada, we built the largest micro grid in the United States with a partner Redwood Materials which was co-founded by a guy named JB Straubel who also co-founded a small car company called Tesla alongside Elon. So JB and his team helped us build out this very large micro grid that's just solar plus second life electric vehicle batteries. So basically we took these batteries out of electric vehicles and we're powering an AI factory entirely with the sun essentially. That's like one case. Natural gas, we've brought to life net new gas generation. We're doing stuff with geothermal and hydro in markets that have access to low-cost abundant geothermal resources like Iceland or hydro resources like Norway where we're investing quite a bit. But we think nuclear is a very important piece of that puzzle long term. I actually look back to before the Inflation Reduction Act came out, like even in climate circles nuclear was considered to be like this bad energy source.
Right, and it never made sense to me. I was always...
We all saw Chernobyl on HBO. I mean that probably...
Yeah. But you know, it's like Germany shut down all these nuclear plants. I think like widely environmentalist, anti-nuclear in spite of the data telling a very different story in terms of the deaths per megawatt created or the critical injuries per megawatt generated. It's actually like the safest energy resource we have out there. And I remember this moment where Exelon was trying to dump nuclear assets to sunset them because they were so worried about the long-term liabilities of them. We looked at buying one for a dollar and I was like this is how I... so yes, and it's like one of my great mistakes of the company is like I should have just done it. So I was like this is crazy for a dollar. Everybody was like no you have no idea the liabilities that you're going to take on. You got to shut this down. And that was a moment where Diablo Canyon was going to get shut down here in California until MIT and Stanford collaborated on writing a paper that said, you know what, California set express climate goals. If they shut down Diablo Canyon, it's going to set those goals back a decade or longer. And like so it was only this moment, I remember it's like a critical inflection point for nuclear where the Inflation Reduction Act was published and they added this notion of having a nuclear production tax credit and it sort of shifted minds of like oh actually nuclear is a good energy source, it is base load, it does produce power cleanly 24/7 with zero carbon emissions, like hey that's actually that's kind of what we want right? And I think it really opened up and laid the groundwork for a lot of innovators like Matt to come in and try to bring these next generation nuclear technologies in smaller form factors that could be safely deployed, could be scalable through high volume manufacturing which is the approach he's taking and we're just really excited about just the innovation taking place across the ecosystem both in the small side as well as like I would like to have some of these bigger AP1000s deployed. I think our biggest concern with them is like both timeline of how long they take to get deployed. It's a 10-year construction process and it's 10 billion per gigawatt, right? Which works well if you amortize it over a hundred years, which is like it's a 100-year asset, but like you know it's hard to find investors that will give me a loan for 100 years. So it's kind of like a different...
And Matt, I mean, we were talking before we got on stage that, you know, studied nuclear physics in college and wanted to build this company and then Fukushima happens and that was probably, that's got to be maybe aside from Chernobyl, one of the lowest moments for people's opinions of the energy and its safety. So, how did you kind of go from or how did nuclear go from that moment to this, you know, now one of the most sought-after energy sources for the AI boom?
Yeah. Well, I think the past five years have been pretty incredible. I mean, if you said nuclear, you're doing a nuclear startup maybe 10 years ago, you would get a dirty look or laughed out of the room. And even three years ago when we told investors we'd be turning a reactor on in two and a half years, many of them literally laughed. And so it's pretty amazing how fast this perception has shifted. And I think the interesting thing is the product, the ideal nuclear product that the customers are demanding today is very different from the ideal nuclear product from 50 years ago. So if you look at it there's 400 reactors around the world today and they're all basically bespoke. They're one-off gigascale projects and there is the AP-1000, but the reality is all the hundred or so reactors in the country today were one-offs for these large utilities where different people built them. They had different designs and that's really the challenge. I mean that is why it does take 10 years plus to build one of these large plants and they cost so much money.
I mean before Vogtle what was the last reactor that was built? So there was a failed one right, VC Summer, that was also happening and that was a whole scandal but before that it was decades before, like in the 70s or something like that. I mean it's insane.
Yeah. Was there a specific... I don't want to go too much into the history here but just entertain me. Was there a moment that you saw in America where like it was at its lowest where politically it was non-viable? I mean, was there something that happened that Three Mile probably didn't help, but was there...
So Three Mile, some people view it as the nail in the coffin, but it wasn't really the main turning point. So, yeah, nuclear cost, there was a time in the 70s when nuclear was as cheap or cheaper than coal.
So like on a normalized basis to today's dollars, it was actually around 5 cents a kilowatt hour. But then regulations started to increase. This could be a whole other panel, but basically there's a lot of people who were trying to throw the baby out with the bath water, people who wanted to get rid of the weapon, and they didn't understand that the energy source is very different. It's like the difference between a stick of TNT and a stick of incense. It's a different product. And so one is not the other. And it's really unfortunate because like Chase said, nuclear is actually probably one of the best energy sources we have available, right? Clean base load, works anywhere anytime, tiny footprint, and contrary to a lot of popular public perception, is as safe as solar and wind.
