Fireside Chat with Crusoe at The Montgomery Summit

So Crusoe is an AI infrastructure business and we are a vertically integrated AI infrastructure business. So what that means to us is like four key aspects of the infrastructure stack. Starting with kind of like the physical infrastructure of land and power development. You know there's a lot of talk in the market around you know power constraints to energize AI. We also, the second layer of the stack is really you know data center development. You know the scale of the data centers being built to support these large clusters of GPUs are unlike you know anything we've ever witnessed before. Just you know it's kind of an order of magnitude step up in terms of the scale and the complexity and the cooling requirements and the networking requirements. And then the third layer of the stack is really you know deploying large clusters of GPUs like the actual silicon and then operating the infrastructure software that manages those GPUs to produce useful results and enables you know AI innovators to you know train foundational models or run inference and AI applications on those chips. And then the final layer of the stack that you know we're really focused on is actually creating a utility from those chips with more of a serverless application. So you know today the main application we deliver there is managed inference where folks will host models with us either open source or private fine-tuned or custom models and then we give them an API and they feed tokens in and we give them tokens out. So it is a we cover a number of layers in the stack but given the speed at which the AI space is moving that vertical integration has been critical to our success. You know to your question on the crypto side, you know I think one of the philosophies of Crusoe is you know we've always taken this very energy first approach to developing computing infrastructure and a lot of that was frankly inspired by you know the digital currency mining space which where you know power costs were really the key profit driver in terms of the overall Bitcoin mining ecosystem. And that led us to you know taking this energy first approach where we would seek out low-cost abundant energy resources and then bring demand for that energy with computing to those sources and that led us to markets like Abilene, Texas.

Yeah so you know a bit of background about me. You know I spent like the first 10 years of my career as a quant building like AI and machine learning models to forecast you know stock returns. And in that role we were you know early days engineering a lot of the features and then running classical machine learning techniques like logistic regression, linear regression but a lot of the magic was really in the feature engineering and feature design. When deep learning became popular and like we had sort of this you know aha moment with the publishing of the AlexNet paper we started experimenting with deep learning models to actually discover features. We started shifting our training workloads from CPUs to GPUs and using these deep learning models to actually produce those alpha forecasts. And simultaneously I was like experimenting with a lot of like NLP side projects and computer vision side projects. And I had this like strong conviction that like AI and intelligence was going to be embedded in every aspect of sort of the economy. And if that happened like energy would actually become the bottleneck because these GPUs actually require a lot more energy than traditional CPU workloads. So my co-founder and I actually went to high school together. I didn't really know much about the energy space but I knew it was like an important aspect to computing infrastructure and my co-founder sort of comes from this background of like having lived in an energy family. You know multiple generations of oil and gas entrepreneurs. You know he studied geology as an undergrad and you know a Thomas Watson fellow studying global energy resources really had sort of spent the early part of his career building in the energy ecosystem. So you know the Crusoe business is really about the convergence of energy and computing and really activating megawatts for intelligence. And really, it's representative in me and my co-founder's partnership between energy and computing and AI.

Yeah, so Crusoe is named after Robinson Crusoe who was stranded on a desert island and had to be really innovative with his resources in order to survive. And we try to think of ourselves as channeling that same sense of innovation with stranded energy. That was like the original philosophy is like where is energy, you know, not being used for something else that we could actually use it to do something innovative with computing. Today we're sort of focused on any energy we can get. You know, that's kind of like the market we're in. But you know, that was sort of the genesis of the name.

Yeah. No, I think it's always powerful to see pictures and images, but you know, this is some drone footage from our data center campus in Abilene, Texas that we are building to support OpenAI and Oracle. You know this project was kicked off in June of 2024 and you know we have kind of the first two buildings which are in the foreground that are operational today running large workloads for OpenAI and then we signed an expansion agreement with Oracle and OpenAI to build six incremental buildings that are coming online as we speak and you know this I think will be an instrumental piece in sort of driving a lot of innovation for OpenAI. The total power capacity of the campus is 1.2 gigawatts. For those like you know outside of the energy space in terms of like how do you rationalize how much power that is, I grew up in Denver so 1.2 gigawatts is about the same power usage as the city of Denver. So you know it's Denver in a data center kind of it's like it's a lot of power.

I think investors should think about Crusoe as a you know AI infrastructure business. We can do multiple things across the stack and you know we think vertical integration is critical to our success in terms of being able to deliver things at scale and at speed that's being demanded by the marketplace today. You know when I first got the opportunity on this facility there was an RFP that sort of went out and they wanted a 100 megawatt facility in one year and you know there were 32 data center providers that responded to the RFP. I was like the last one like after everybody had said no like there's no chance the closest bid was two and a half years and you know I was like you know one that was stupid enough or brave enough depending on how you look at it to say oh yeah we could definitely do that. And part of the reason we were able to execute with that speed and be able to deliver those first facilities in 11 months was that we actually had vertically integrated critical electrical manufacturing for key components that would have otherwise been massive supply chain risks to delivering the data center on time. So you know I think you know this space is evolving so rapidly that being vertically integrated enables you to unblock unique challenges that ultimately crop up when you start hitting the scale of the facilities and demand that's in place.

You know, I can't comment whether or not we're planning to or not, but you know, we're always sort of evaluating the public markets. We think, you know, the public markets are ultimately probably the best place for a business like Crusoe to live just given the capital requirements of what we're doing. You know that first eight buildings of Abilene was about a 15 billion dollar project that we were able to fund with JP Morgan on the construction debt side and Blue as a project equity partner. But you know our ambitions are you know many many many times bigger than you know just this single campus. Today we have about 45 gigawatts in our development pipeline. And you know we're helping build some of the infrastructure for a lot of the biggest companies in the world and that just requires a lot of capital. So you know we think the public markets would be really receptive to you know this scaled AI infrastructure business that's delivering for these leading companies in the AI arms race.

I begged him. No. Michael, you know Michael and I kind of had a really good chemistry kind of the first time we met and you know this was a process for us in terms of you know trying to find the right fit for CFO at Crusoe. You know Crusoe is not a super straightforward business. It's not like just all equity and you know we're an enterprise SaaS company or something like that. You know we have a lot of capex. There's a lot of unique ways of funding things. So, you know, Michael was someone that I was just incredibly impressed with from our first conversation and you know, we had a very good dynamic between us and I felt like he was kind of up to the challenge of meeting the financial needs of Crusoe as a business and you know, as soon as I did reference checks with folks like you, I got a lot of glowing reviews about Michael as a public CFO and having run MongoDB joining at I think 40 million in revenue and growing it to over two billion in revenue and having been there for about 10 years. He was just a perfect fit and he was also a perfect cultural fit which was a big deal for us.