theCUBE + NYSE Wired: Physical AI & Robotics Leaders QA2 | Chase Lochmiller, Crusoe

Clips
More from theCUBE + NYSE Wired: Physical AI & Robotics Leaders QA2

Chase Lochmiller

Co Founder & CEO

Crusoe

play_circle_outline Crusoe's Energy-First AI Data Centers: Building 1.2 Gigawatts Power Capacity in Abilene for Innovative Solutions

play_circle_outline Importance of energy availability highlighted as key factor in data center location choices.

play_circle_outline Crusoe's strategy includes building data centers as a foundational layer for AI development.

Info
Transcript

Chase Lochmiller, Crusoe

Chase Lochmiller

Co Founder & CEO Crusoe

Chase Lochmiller, co-founder and chief executive officer of Crusoe Energy Systems LLC, joins theCUBE’s Dave Vellante and John Furrier during theCUBE + NYSE Wired: Robotics & AI Infrastructure Leaders 2025 event to unpack the intersection of energy access and AI compute. The conversation explores Crusoe’s strategy of deploying high-density data centers in regions rich with underutilized renewable energy.

Lochmiller shares how Crusoe’s vertically integrated approach — from energy sourcing to electrical component manufacturing — positions it to meet soa... Read more

explore Keep Exploring

What has been Crusoe's approach to developing computing infrastructure, particularly in relation to energy and artificial intelligence? add

What is the strategy being employed to identify new locations for building data center infrastructure? add

What unique aspects does Crusoe offer in relation to data centers and their operations? add

bolt Powered by CUBE AI

Chase Lochmiller, Crusoe

search

>> Welcome back everyone. I'm John Furrier here at theCUBE studios for our three days of Robotics and AI Leaders. I'm with my co-host, Dave Vellante. Chase Lochmiller is here, CEO of Crusoe. Chase, great to see you again. I know we have a tight time. You got a call, but your work you're doing right now is, I would say, really historic.

Dave Vellante

>> Not much going on since we last talked.

>> Give us the update. Stargate, you're working on some mega projects, one of the biggest ones being discussed. Give us the update.

Chase Lochmiller

>> Yeah, it's been a really busy chapter. Crusoe has always been building at this convergence of energy and computing infrastructure, and we've always taken an energy-first approach to how we develop the computing infrastructure needed for important innovations like artificial intelligence. And as AI has scaled the bottleneck, for AI's progress has become energy. And this is sort of catapulted Crusoe solutions and a lot of the value propositions that we bring to sort of the center of the conversation. And we've been fortunate enough to have some great partners building very big facilities like what we're doing in Abilene, Texas. We're building 1.2 gigawatts of total power capacity for a very large scale AI data center.

>> I mean, this is the most important global thing happening because it's going to have implications. So the whole discussion, we could probably pick up later, but you gave a talk in D.C., the energy grid challenges and we're bounded by energy. What are some of the things you're doing that's going to, one, bring the scale within the energy equation? What are you guys doing? Is there any innovation there? Can you share specifics without giving away the secret sauce?

Chase Lochmiller

>> Sure. Look, I think the number one thing that we're focused on doing is by taking an energy-first approach, it sort of opens up new geographical locations where you can actually be building this infrastructure. So the center of the world for data centers historically has been Northern Virginia, but Northern Virginia doesn't have the power needed to support this growth-

Dave Vellante

>> Room either....

Chase Lochmiller

>> in demand or space to support the growth and demand for AI, so where else can we go know? Crusoe's really taken an energy-first approach saying, "Okay," that's actually what led us to Abilene, Texas. It was a market that had actually overdeveloped wind and solar and frequently had negatively priced power because there was no marginal demand for that power. Enter AI. AI was this very large demand for incremental energy. And we could sort of help solve this problem of overdevelopment from the IPPs with a very large AI data center and sort of fit on both sides of the equation where there was the energy that actually needed offtake and there was demand from the AI infrastructure side.

>> And solar is a big part there? Are you looking at the solar .

Chase Lochmiller

>> Wind is the bigger part. There's a lot of wind development in that region. Sort of one of the windiest corridors in the United States. You get some of the highest capacity factors for new development of wind farms.

>> So the question I want to ask you is AI factories are hot, we're seeing people talk about it. What's the difference between AI factory and a traditional data center?

Chase Lochmiller

>> Sure. So I like to talk about Crusoe as an AI factory company, so we have a hardware and a software component to our business. We build the physical AI factories, these large-scale AI data centers that have big clusters of GPUs all interconnected together. And then we also build software that operates the AI factories. How do we actually make that computing infrastructure useful for innovators, useful for people that want to embed intelligence in their very specific application. But what makes it different from a traditional data center is a couple of different things. Number one is that the density of computing infrastructure is dramatically different from anything we've ever seen. If you look 20 years ago, a data center, a traditional rack might've been two kilowatts, maybe four kilowatts. What we're deploying today for Blackwell and the GB200 is 130 kilowatts per rack that's going to 600 kilowatts per rack and eventually a megawatt per rack just in the next couple of years. So you have super high density configurations, which means you have a lot of heat that you have to exhaust, which means you have to reinvent a lot of the thermal management and the cooling architectures of these data centers. So what we've built is everything is direct to chip, liquid-cooled architecture, and we have a million gallons of water per building that cycle through the building and sort of extract the heat from the chip and then sort of exhaust it through these large rows of coolers that we have, and then it cycles back through to cool the chips.

Dave Vellante

>> And explain, you're not only just standing up the GPUs and everything that goes along with that, you actually build the data centers and that's unique in the industry.

Chase Lochmiller

>> Sure. We're a vertically integrated business, we'll do everything from building the data centers. We actually manufacture, we have factories at Crusoe where we manufacture critical electrical components that go into making these data centers, so they're sort of like data center Lego blocks. And then we also have a fully managed AI cloud stack that virtualizes GPUs that offers a managed Kubernetes offering that has a managed high-performance file system and high-performance block storage solutions as well.

Dave Vellante

>> Well, if I went into a Crusoe data center, how would it be different? I just took a tour, virtual tour of Amazon. I'd never been to an Amazon data center before. I was actually kind of surprised data centers, how pristine they were. What would it look like physically and what's different from the traditional data center?

Chase Lochmiller

>> Again, I think the density of it is quite a bit different. So, in Abilene, for instance, we have eight buildings, each building's about half a million square feet, which sounds like a lot, but if you were to compare that to a traditional data center for this scale of power usage, it's about a third the size.

Dave Vellante

>> Okay.

Chase Lochmiller

>> So it's actually small compared to if you were to do gigawatt scale capacity for a traditional data center, it'd probably be about 3x the total square footage. So it's a lot smaller. You end up with these very high-density compute data halls that are entirely engineered for high-density GPU architecture.

Dave Vellante

>> But traditional hot aisle, cold aisle, or different techniques?

Chase Lochmiller

>> Well, everything is chilled through this chilled water loop. So we have, again, a million gallons of water per building that chill the GPUs, and then it's a closed loop ecosystem, a closed loop architecture, which means that cold water flows through and flows over copper pipes over the chips, and you get the sort of thermal heat transfer from the chip to the water, and then the hot water exhausts and it goes out to a chiller. And when you have a heat exchanger on the outside and you have sort of air blowing over these wound copper coils, and that's how the heat gets out of the system.

>> Chase, are we in an overbuild or an underbuild market? I mean, the demand seems to be high. I mean, to me it seems like there's never enough capacity. And two, as you look at your business, what's the demand curve like? Because I'm sure sovereign cloud's going to be big. What's going on in your business and how would you peg the market? I mean, obviously OpenAI and all these Oracle, there's a local investment angle. What's going on with the build-ups? One, is it overbuilding or underbuilding? How do you think about that? And two, revitalization of the area and impact to your business around the world?

Chase Lochmiller

>> Yeah, we get asked this question a lot in terms of overbuild or underbuild. I think from my perspective, we can't build enough right now. And when we're building, we're not building on a speculative basis. We're building with the full-fledged credit and support and guarantee of companies with the greatest balance sheets in the history of business that are basically saying, "Yes, we want this and we want it ."

Dave Vellante

>> Real faster.

Chase Lochmiller

>> And we want it and we're guaranteeing it. So to me, that should mitigate any sort of financial concerns around like, is there an overbuild in place? And so to me, it's like we literally cannot build enough, Crusoe or others. And the reason for that is that we're reinventing the economy. We are fundamentally transforming the way that people work. The marginal cost of intelligence is eventually going to trend towards the marginal price for electricity. And with that in place, it's like, "Well, okay, if we can 10x the amount of intelligence that we can produce through the workforce through silicon, isn't that going to be a positive thing for the economy in terms of what humans are able to accomplish?" So we're very optimistic that this demand curve is only going to accelerate.

>> What's the pressure on the speed? Not to get into your affairs too much, but I can imagine the pressure to go faster is high.

Chase Lochmiller

>> Sure, so speed is king right now. Speed is very difficult to replicate as well. So that's actually one of the unique aspects that Crusoe really brings a lot of value to the table. Compared to the big companies, they are structurally designed in a way that structurally disadvantages them to move as quickly as we can move. And a lot of that has to do with we're a company that has a very simple decision-making architecture. It's like I can just make a decision and then we do it, versus having to go through nine levels of the chain of command to actually move forward with something. So speed is critical at this moment in time, and a lot of it comes from the fact that everybody's feeling the pressure. There's a lot of competition in this foundational model space, there's a lot of competition in AI to drive improvements. And the fundamental way to drive improvements is scale, speed and scale.

>> You mentioned density. I know we don't have a lot of time. I want to get this out there because Jensen always puts his roadmap out there. He brags about it and I think it's rightfully so, but five years out.

Chase Lochmiller

>> It's a good roadmap.

>> It's a good roadmap. Yeah, go faster too. As you deal with legacy, I mean, basically Grace Blackwell will be superseded by the next one, so that's clearly in the roadmap. How do you design that in to stay a few moves ahead, or is that not a factor?

Chase Lochmiller

>> I think the new generations of hardware, they don't automatically depreciate the last generation to zero. We're seeing Blackwell come online right now, demand for Hopper is still sky-high. It's a sold out SKU.

>> It doesn't affect your system though. You're building a system in these data centers, but on the design side, that does not impact you guys at all?

Chase Lochmiller

>> The design side, it does change a bit just in terms of everything is trending to higher density, more compute power in a smaller location, which you have to engineer around. But I think we're just trying to stay in lockstep with NVIDIA in terms of how we manage a lot of the cooling architectures for next generation.

Dave Vellante

>> And on your website says, "Reserve your GB200 today," so I'm inferring from that it's a supply constraint and you just had to deal with AMD, so I wonder if you could talk to that in the whole training versus inference discussion.

Chase Lochmiller

>> Yeah. I mean, know Crusoe ultimately is a AI cloud platform that serves its customers and we want to help support the industry. We kind of view ourselves as a infrastructure layer that helps serve the innovators and builders to go change the world. And we've definitely seen demand for multiple hardware suppliers in that space and we ultimately want to just help support the industry by supporting both different SKUs across NVIDIA, like their high-end products like GB200, as well as some of the more cost-effective ones like L40S, as well as other players like AMD.

>> Chase, great to have you. I know you got a hard stop and we got to go to the event. Definitely want to do a follow-up. You are a leader in obviously physical AI, which is robotics, part of our theme, but AI infrastructure. And again, congratulations and we're so thankful to you coming on because you're setting the table, this is a generational move. Final words to share just where you are at mentally, how you feel, your view on the historic moment we're in?

Chase Lochmiller

>> Yeah, look, ultimately, I'm just grateful to be in the position to help society and help humans prosper in the future through technology. The technological wave of innovation that we're seeing coming led by AI is unlike I think anything we've ever witnessed in the history of humanity. And I think there's massive human prosperity gains to be had from us all benefiting from implementing and utilizing AI in a really meaningful way.

>> And we think as you build these systems, we're going to start to see the geopolitical, kind of maybe alliances and changing of how people organize and apply some of that AI. It's really interesting.

Chase Lochmiller

>> Yeah, that's actually interesting. So a lot of people often ask me like, "Oh, well, can you build in certain locations or do you have to build in certain locations because of latency concerns?" And latency is kind of a concern, but it's way less of a concern than actually data sovereignty. So actually the reason to build in a lot of different countries is because folks are... There's a lot of different geopolitical-

>> There's alliances going on around data center footprint....

Chase Lochmiller

>> angling. Exactly.

>> I mean, this is an opportunity.

Chase Lochmiller

>> Where the data is and it's like every country needs its own AI gigafactory and-

>> Malaysia can't have it. They're going to partner with somebody else. I mean, I'm serious. This is an opportunity. So what you're doing is amazing. And again, we're so grateful for coming on theCUBE and sharing and looking forward to doing more.

Chase Lochmiller

>> Absolutely. I'm happy to come back some other time. I'm sure we'll have some great news to share-

>> In New York....

Chase Lochmiller

>> in the coming months. Yeah, yeah, coming to New York.

>> Okay. This is theCUBE here is part of NYSE forming the Wired community again, bringing experts. The key is this wave of innovation is going to change the world we live in, not just the apps we use, but also change the makeup of how we organize ourselves and ultimately prosper. This is theCUBE doing our best to bring you the action for Dave Vellante and me, John Furrier, thanks for watching.