Welcome to the Energizing America podcast. I'm excited to sit down with Wes Cummins this morning and talk a little bit about the Applied Digital story. So without further ado, Wes, I'd like you to go ahead and just give us a little information about who Wes is, where are you from, and why Applied Digital.
Yeah, absolutely. Thanks first, Shane, for having me. I appreciate you taking the time. So, a little bit about myself. I grew up on a potato farm in Idaho. I went to school in the Midwest in St. Louis at Washington University. I played football there. Played football, you know, sports in high school. Small town. Everyone plays sports. That's what you have to do. And then I got into the investment world out of college. And I really ended up being an analyst for a while at an investment bank, and then just a lot of different positions in it. But I basically have been kind of in the tech investing world for the past 26 years now.
And then about five years ago, while I was doing that, I saw an opportunity to build Bitcoin data centers here in the US as Bitcoin became more popular. And then there was a big crackdown in China on crypto in general. And so that's where, you know, 70-plus percent of Bitcoin mining was done in China. And so there was a big opportunity to go build that here in the US. It was really the infrastructure to operate Bitcoin, and we were a data center builder and operator. We weren't a Bitcoin miner ourselves. So we really started there as Bitcoin data centers. And then in 2022, we decided to add high-performance computing to our Bitcoin site. So it was really, you know, GPU-based computing. Now in '22, that was very different than what it is today. So it was pretty minimal use cases, like a small amount. You did small-scale builds, especially versus what we're doing today. But we hired some guys who knew what they were doing in that space, and we designed a facility. But I would say first we tried to retrofit our Bitcoin facility to do GPUs and decided that wasn't a very good path. And then we designed a kind of a slimmed-down, a cheaper version, and we started building that in Jamestown, North Dakota, in '22.
But by the end of '22, we had some GPUs running out of our facility in Jamestown, and we had a few customers that were using the GPUs for deep learning and machine learning at the time. But what happened between there and now is in December '22, you had ChatGPT, and that was cool and novel, and, you know, the fastest app to a million subs or a million users. And so that was really interesting, but it didn't change our world at the time. And then when the Nvidia H100 was introduced in March of '23, we started to see some big changes. So we built this data center capacity that was actually a good fit for H100. It was 50-kilowatt rack densities, cooling, all the things that you would need to run that type of density. And so we were out marketing this, you know, for someone to rent the data center space. And we were introduced into some AI labs and some frontier model companies. And what they wanted was GPU capacity. So we ended up buying the H100s ourselves, setting up that GPU capacity, and providing them that capacity on like a cloud basis, so GPU as a service. And what I saw though in that time frame was the immense amount of demand there was for this. I was pretty surprised by the level of demand. And so I went back to my data center design team and I said, guys, you know, I know we were going to build kind of 10 or 15 megawatt builds on site. We need to start doing hundreds of megawatts again, like we did for Bitcoin. We need to do large scale. I see a big amount of demand for this. And what I really saw was, you know, because of my tech investing background, was that there was a big shortage of H100s at the time, the Nvidia H100, and you couldn't get your hands on it. And so there was a supply shortage of that because of the level of demand. And when I looked at it, I said, look, from my experience historically, like Nvidia is going to fix wafer throughput at TSMC. They're going to fix the CoWoS advanced packaging. They're going to fix transceiver manufacturing. They're going to fix all of these supply chain issues pretty quickly. But what I saw in the market was there's not a lot of data center space that can handle this type of power density. And so I thought to myself, you know, if I fast forward a year or two years, data center space is going to be the real bottleneck in the industry because while you can fix the manufacturing supply chain issues, you know, relatively quickly, finding power, getting land, getting permitting, getting supply chain, building buildings, like that's probably going to take a lot longer than fixing a semiconductor supply chain. So we went, you know, full speed into that, and we started on a 100-megawatt critical IT load building in Ellendale, North Dakota. And we worked with Nvidia's data center team, and we worked with the team at Cray, the unit of HP. So Cray did supercomputing for, you know, two decades, and so they know a lot about this. And we landed on a design. We really started to design to the Nvidia roadmap. We thought they were by far the leader in the space from a hardware perspective, and we were building data centers to house that hardware. So we worked with them, started building the building in October '23, no customer, building it in the middle of nowhere because we believed that this was the way the world was going. And so then if you fast forward to this year, we signed a customer core for that site, that first building, and then two additional buildings. And now we've signed an investment-grade hyperscaler customer for our campus in Harwood, North Dakota, which is up closer to Fargo. And so we see a lot of momentum with additional investment-grade hyperscalers. And when I say investment-grade hyperscalers, we really target five. And that's in no particular order. That's Microsoft, it's Amazon, it's Oracle, it's Google, and it's Meta. And so we have them plus CoreWeave, really our customer base that we're targeting right now in our business. And what we do is we, you know, develop the data center, we design it, we build it, and then we operate it after the fact. And right now, we're signing 15-year leases that are effectively non-cancellable leases. So we get a lot of stability and visibility in our business. And then it turns into a, you know, power, supply chain, people, capital, and customer game, right? Those are kind of the pillars of how we do our business. So we have two campuses going right now, 700 megawatts of critical IT load under construction. We're marketing multiple other campuses to multiple other investment-grade hyperscalers and making a lot of progress there. And so our business has really accelerated here. You know, it was a lot of years of work and stress, and now it's accelerated. And I'm not saying the work of stress is behind us by any means, just getting started from the work perspective. But now we have two customers on board. We have 16 billion of contracted revenue over a 15-year period. And now it's just how do we, we've landed the first few. Now we'll diversify with customers, diversify across locations, and continue to scale our platform, you know, for years to come is the expectation.
That's a phenomenal story, and I love how this is that oftentimes in business or with an idea there's years of idea generation, probably some failed ideas, some struggles along the way, and then bam, it looks like when I drove out to Ellendale, North Dakota last week to check out the project, and it looks like it must have been in planning stages for tens of years based on the army of people out there and the progress. I mean, you have one building that's definitely locked down. There's a tenant in there. You have another building that's pretty well looks like it's complete. You got a third one that walls look like they're getting lifted, and a fourth one that the dirt is being moved. And it's like this little, it brings you back to the origination of America, how it must have been. It's an exciting spot. So what differentiates Applied from other people in this space?
Well, there's a couple of things, and there's kind of two groups that I would say in the space. And the thing that differentiates us from the other people that were building infrastructure for Bitcoin and for crypto is we started down this path early. So we started, I would say, close to two years earlier than everyone else did. We started in '22. Most of the people started doing it in '24 when it became more obvious that the data center capacity was going to be an issue. So that's allowed for us is, you know, what I call stubbing our toe at a smaller scale. So we definitely stubbed our toe. We did the Bitcoin retrofit, walked away from that. We did a scaled-down build of a building, which just has, you know, less redundancy. It's not nearly as big. This was the 10-megawatt build we did in Jamestown. And then, you know, we decided that's not going to work for us. So by the time we went to large scale on our build, we were kind of on our third iteration, and now we've refined that process further to where it's like our fourth-gen data center that we're building now. So there's a lot of learning that goes on through that. And then we've onboarded with almost all of those major hyperscalers I mentioned. We've been through the technical review. Our design's been approved. We're onboarded as a supplier. Like that doesn't sound like a lot, but it's a really big lift to get there. And so that lets us engage pretty openly on new business opportunities with them and move things along in the pipeline faster. So, you know, we worked for a year and a half to contract that first building in Ellendale, and then from there we worked about four months to get the second. And then we'll see, you know, let's see what happens on the third and the fourth and the fifth different campuses. That, you know, you've seen that time frame accelerate significantly because of all the work we've done. And then it helped us refine the buildings too, because what we're building, we want to be able to have the same building work for six different customers and then well beyond that, right? If we went down in the size of companies that we were willing to take in those buildings, you would increase the number significantly. But basically we want a fungible data center space that works for everyone. And so what we've done is created a lot more flexibility in our design, built bigger data halls, bigger air-liquid mix, a lot of those things. So it's just that learning curve of having those extra years on it. And then there's another group, which is a lot of private data center builders, and that's the biggest part of the industry. So most of the data center companies are private, and they've been building for hyperscalers for, you know, over a decade, and they have a refined process, and they're very good. And that's when we looked at this space, we said, look, that's who we want to be, right? We want to be like those companies. They've figured this out. They know how to do business with these large entities, and they know how to build, and they know how to operate. And so what we did again years ago is we went out and we hired people out of those companies, and we created our team, and then we've refined that over time. And so we have a really good team. And then if you compare us to those guys, what we did first relative to them is we took the data center to where the power is, because you need so much power now. Those companies were all used to building in tier one and tier two metro markets, right? That's where data center capacity was built, is in tier one and tier two metro markets. So we took their model, moved it to where the power is, and then figured out how to build in much smaller communities where there was an excess amount of power that we could utilize for AI workloads. And so that's kind of our advantage over those. Now since then, they're doing the same thing. And on the public side with the other guys who are converting from crypto infrastructure, they're doing the same thing as us as well. I think we just have a head start on both of those. And so we're trying to capitalize on that.
Well, and I think clearly there's the win is where the power is. I think that I was astounded when I drove by the electrical infrastructure to get into your facility in Ellendale. And anyone who hasn't had the pleasure, number one, of just getting out to rural America, get out to Ellendale and check these things out. Especially if you're unaware of what a data center is, they're phenomenal. But as I was driving, I was born in North Dakota. I have business in North Dakota. I love North Dakota. I was driving through these rural farm fields, and I'm headed out there, and I just keep going back to this, Wes, and one of your folks on your team did a presentation at one of the conferences I was at, did a phenomenal job and talked about the power of North Dakota. But I think our listeners need to know what originally attracted you to North Dakota, and how does that compare to the other states that you may have even looked at?
Yeah, so we've done business in other states. In the future, we'll do business in other states again. But the attraction of North Dakota is a couple of things. Abundant power resources. So North Dakota produces eight to ten times the amount of energy that is consumed in the state. And so this is an energy-oriented business. So that's the biggest draw to North Dakota. And then what really worked well for North Dakota is from a political climate. When we started working in North Dakota, the governor was Doug Burgum, and Doug's now Secretary of Interior. But one of his biggest agenda items was using the power generated, using the energy generated in the state of North Dakota for the economic benefit of North Dakota. And so here we are, right? Like that's, we're a perfect fit for that agenda. And that agenda has continued under current Governor Kelly Armstrong. And so political environment, business environment is really welcoming from that perspective for us. It's a fiber-rich state. So you need power and you need fiber to operate data centers. And so a lot of cross-country fiber routes roll through North Dakota. We have at our facility in Ellendale, you know, the town is 1,100 people, but we have four different fiber providers with four diverse routes. We have dark fiber, and the fiber is just getting better there. So the state is really great from that perspective. And so those are some of the key ingredients. And then you have a workforce in North Dakota. You know, the Bakken really pioneered this for us, but North Dakotans go to work like they make it happen. We've done construction there four years in a row through the winter. I mean, you were born there. You know how cold it gets. It's a unique place in the world how cold that place gets. And so you have people there that go out and show up for work every day and make it happen. So it's really good from all of those aspects. And so then when you step back, one of the things that has worked out in our favor and in the favor of North Dakota is that one of the other big elements. So you have power and you have fiber, but when you use a lot of power, you need cooling, right? And guess what? North Dakota has a lot of free cooling. And so what that turns into for us, and really for our customer, which becomes really attractive for our customer at this site. We published a white paper back in June where if you look at North Dakota versus, say, a southern state where you have a lot of data center capacity built on 100 megawatts of critical IT load, we estimate that our customer will save annually 50 to 65 million in operating costs because you use less electricity. The electricity is fairly inexpensive in North Dakota, and you're using a lot of, in the case of Ellendale, you use electricity that would otherwise go unused, right? It's stranded power. And so it's just a lot of different benefits for being in that state in particular. And then if you look at what's happened in data center over just call it the last seven years, even before AI started, you started to see a migration of data center capacity towards the middle of the country, right? Google built a big campus in Omaha, Nebraska, Microsoft in Des Moines, Iowa. Those are not population centers. So that was already a break from what the tradition was of building in big population centers. And if you look at North Dakota, we're just north of that. And it's a natural evolution to move into that state. The only knock that was on the state of North Dakota is data center hadn't really been done there before. Like that was it. It's not a data center zone. It's not a region. And we've already changed that, and we'll continue to change that as we bring more hyperscalers into the state. So now we have two hyperscalers in the state, an investment grade and then CoreWeave. And then, you know, we expect that it's going to be easier to get more, and we're working on that already. And so what that does is, you know, what I tell people is the gravity of data centers historically has been population, and now the gravity of data centers is data. So the more data centers and the more operators you get in a location, in a region, the more you're going to get, right? It's, if you ever watch the movie Gremlins, you know, you're not supposed to get your Mogwai wet. Like it's like getting that once you get one, you know, and then you just pour water on it, and you'll get more in the state. And so it's been a really good setup. It's really good for the state. We work really closely with these small communities. And then the other thing on this, Shane, is that going to these small communities, like just given where I came from, it's really amazing to bring the leading edge of technology into a town of 1,100 people, right? And they get the job creation from that. We train a lot of people locally for jobs that pay, you know, double the amount of what the average household income is in North Dakota. And so it's a lot of benefits. The tax benefit, the property tax benefit locally for the town, for the school, you know, these things that while they're a little bit of a nuisance while we're building it, you know, they're going to operate for decades after that, and there's going to be a few hundred people that come and go every day, and you're not even going to notice that it's there. And you get this massive inflow of tax revenue and benefits and economic benefits for these small communities. And, you know, how North Dakota works, right? It's not just the town of Ellendale, it's the surrounding towns as well, right? Because you don't just have all your workers from Ellendale. It's a culture of I'm going to probably drive 30 or 45 minutes to go to work. I'm going to go one or two towns over. So it's kind of a benefit for the area, not just for the town. The biggest benefits for the town, but for the area overall.
Well, and I think you made an important point there is what it attracts to North Dakota beyond Applied Digital and what a win for them. You know, I'm hearing of other data centers that are in the works in the state that are really exciting. And one thing is, you know, as I was, I went to a couple coffee shops in Ellendale. There's not more than a couple. And I asked people in there, how are you feeling about this data center? Where did you come from? Do you live here? Have you lived here your whole life? I got exactly zero negativity about Applied Digital's presence in that town. And so, while it can be easy for folks from a town over to say, what's going on over there? Get your boots on and go visit with the people that are positively impacted by it or negatively and get their thoughts on it. And you don't find much negativity in Ellendale. And the other thing that really surprised me about that was you mentioned that the whole area benefits from it. You know, Ellendale is such a small town, but the Applied Digital site is just outside of the main hub of the town. So it's a couple blocks over, and it gives a nice little barrier. It's just an absolute beautiful setup. And Governor Burgum, when he was in office, he always told the oil industry, innovation over regulation. I want you guys to innovate over me regulating. And I think that kind of a business mindset has really served North Dakota well. And it's awesome that you guys are here and doing incredible things. When we think about AI and the future of AI, how are you designing your facilities to make sure you're going to be able to keep these tenants forever?
Maybe not forever, but for a while, for a long, long decades, right, is the goal here. So future-proofing is something that we think about a lot, and so that's what we've done. You know, when I mentioned we added a lot of flexibility to the site, to the building as we move forward through it, this is around future-proofing, right? And there's a couple of things you need to think of. Power densities going higher or power densities going lower, right? Right now we're in a trend of power densities are going higher, and that looks like that's in place for the foreseeable future, and that's great. We're fine. We would deal with that. We need to do electrical distribution up to, you know, we need to move from 415 volts up to 800 volts. So we plan for that. You're really planning to go up to like one megawatt rack densities, which is pretty insane. So if you look back at how data centers operated previously, if you went back, you know, still even the vast majority of data center capacity, you're operating at like five to ten kilowatts in a rack. And now our first deployment in Ellendale is 135 kilowatts a rack. So massive jump in what we're doing from a rack density, but then you're thinking of going from 135 kilowatts to a megawatt a rack. Like you're dealing with those kind of things. So we need to be able to do that. But what if there's a big breakthrough four years from now and power densities go down? Well, we've built a lot of data hall floor space so that we can accommodate that, and it's easier to do in North Dakota because we're able to get more land, and it's not really that much more expensive to build a bigger footprint. So we're doing a lot of those things to future-proof what we're doing, what we have in the building for our customer in the future. Now, I think it is important to note that when we sign these contracts, and it's a 15-year contract, that our customer accepted the building as is, and so we have that protection through 15 years of if we need to make changes to it, our customer needs to pay us to make those changes, right? It was accepted. But I also need to make sure our customer is successful. So we've built maximum flexibility into these. You can't cover everything. We both know that, but we're trying to cover as much as we can.
Are you concerned about the explosion of electrical need? Or like, how are you, where are your thoughts about right now? You said North Dakota produces eight to ten times the power that it uses, but what are you, where's your thoughts in the future here? Are we going to be able to keep up?
Yeah, so if you're thinking further out, so what we are doing is we're using, you know, power that's not used at a certain location, and then we go to those locations because they have a lot of distribution capacity, electrical distribution capacity. But there will need to be additional generation built in the state. And North Dakota, once again, fantastic for this. You have a lot of trapped gas in the Bakken that can be used in the state, and so we're already working through that process so that when we use the power that's available now, but we can get more power to these locations. You should expect that in, you know, late '27, '28, '29, that, you know, we're already working through the solutions with our utility partners there to bring more generation online to feed what we think is the growth trajectory for this business and this industry.
