Hey everyone, this is Kenny Coins with Marathon Money. I'm with Wes here, CEO of Applied Digital. It's been amazing watching Applied Digital. I know a ton of people that's in the stock and they can tell me a vast amount of information. I have a friend of mine, Peter, he's a big fan of Applied Digital and thinks you guys are going to do great things. And he also gave me, as everybody else did, a list of questions to ask you and things they want to know. But besides that, you know, you've been doing this for a really long time.
Yeah, Kenny, first off, thanks for having me. I appreciate it. We've been doing this — so Applied Digital, you know, we've been in business for about four years, but I've been a tech investor, mostly amongst hardware, digital infrastructure, for oh, 26 years now. So I've seen a lot of different cycles and a lot of different new technologies and big waves and busts. But this one has been by far the fastest and what seems to be, you know, kind of the largest tech revolution that I've seen in my career.
Data centers. Data centers is like big now. They're like really hitting the market because pretty much everybody's going to be in some form of AI in the next 10 years.
Yeah, you know, so on the data center front, just in context for history, so we started doing this, building out Bitcoin data centers. We built out 500 megawatts, which is, you know, a significant number. I think that once we built that, we built that in 24 months. It's a lot of power to bring online in that short time period. But we built those facilities and we never mined Bitcoin ourselves. So Marathon Holdings, Marathon Digital, is our largest customer as a Bitcoin miner, and so we run effectively data center facilities for them. And then in 2022, hired some guys here in Dallas — we're based in the Dallas area — to look at adding high-performance computing alongside of our Bitcoin facilities. And they, you know, they were working for a company that was building data centers for, I believe, Meta at the time, one of the hyperscalers. And we designed a new style, you know, high-performance computing facility. We started building it in '22, and then ChatGPT hits in December of '22. So we were doing smaller-scale builds, and then everything really changed for us in this market in kind of the April time frame, right after Nvidia had introduced their new H100, the next generation of their data center GPUs. There was a massive amount of demand. The facility that we were building was a really good fit for the H100. So that's how we, you know, started in that business. We landed a cloud contract with one of the AI labs that was founded by actually one of the inventors of the Transformer, which is the piece of software that makes generative AI possible. So we started down that path, and I went back to my real estate development team and now data center development team and said, guys, we need — I know we were going to build, you know, 10-megawatt facilities, we need to go back to building hundreds of megawatts because there's a big wave of demand coming.
And there's the data center style versus what we had in the past. So what's been built in data centers, you know, for the past 30 years has primarily been around delivery of video. You know, it's Netflix, it's Instagram, it's YouTube, it's TikTok, and it was all developed around comms. And now we need to build around compute, which is what really, you know, creates the artificial intelligence — is the compute. And when you're building around compute, there's certain things that we can dive into that the data centers need to be different. So we went and designed this, we started building at scale. We have a really large campus going up in North Dakota, and then we have a couple of other campuses that we're developing as well. But now we're building very large-scale data centers for AI workloads.
Like when you say all that with high-performance computing and transitioning, it seems like you made a lot of transitions, more like a lot of pivots, because the market is moving so fast. It's like in 2021 you were here, in 2022 because of the change ChatGPT made, and then you got Nvidia that came out with their GPUs. How, when do you think that technology is going to slow down to where everybody can catch up, or is this just going to keep being a pivot and we're going to have to basically, as an investor, we're going to have to rely on you to kind of say, look, we got to make this pivot because technology is increasing fast. For example, like DeepSeek, that literally rattled the markets on Monday.
Yeah, so technology is, you know, always moving faster. And you know, you think back to when the first iPhone came out in, you know, 2007. It's, you know, it's been just, you know, barely what, 15 years since that came out. You know, no one — I mean, old guys like me can remember back to when, you know, we were using T9 to send texts on your 10-digit keypad. So, you know, but you just see it constantly accelerating, and now it feels like it's even more, you know, turbocharged from an acceleration standpoint because of the power of the compute that we have out. So we constantly are evolving inside of our business, and we, you know, we're looking inside, we try to future-proof these facilities as much as we can. Let's see what comes of the, you know, the DeepSeek news. So, you know, the interesting thing about DeepSeek is, I think it was Christmas they put it out, and we went through, you know, the white paper and looked at their V3 model. But it's the logic model, the reasoning model that came out, the R1, that, you know, started getting traction on the app store, and then they, you know, talked about what the costs were and how much more efficient they were in training and inference. Let's see, kind of that story is really evolving. So let's see how that plays out. But the compute has to get less expensive, right, if we're going to get AI to have the adoption that we expect it to have. And I think it's going to go everywhere. You need the cost of compute to come down, just like, you know, when I know my parents bought us an Apple IIc in 1985 that we used to program on, with a monitor like this big, and it was, you know, monochrome green. And I think they spent $5,000 in 1985, which, you know, inflate that for today, for a computer that would do almost nothing except let me play like California Games or a trail on it. And so, you know, to see how compute costs have come down overall, we need to see, you know, that same thing. We need to see more efficiency. And I think this is, you know, a partial step in that. I just, I don't want to jump the gun. I think kind of the lack of skepticism around, you know, DeepSeek — I saw, you know, Microsoft and OpenAI saying today that they thought that they, you know, stole some data through a sister company and are misusing or not using the licensing correctly for the OpenAI, the GPT models. So let's see how that evolves first before we jump into it. But I think the really important thing is compute cost has to come down. It has to come down. But AI is driven by compute, and we need, you know, we're going to need massive amounts, which is what, you know, we're building the infrastructure for. But we're going to see more changes in this market. There's no way that, you know, innovation never stops. And like I said, we've seen it accelerating, whether it was through wireless networks, whether through fiber optics, whether it was through software, and now on compute. And we're going to see it continue to get more efficient and continue to evolve just at a faster rate.
You know, with the whole thing going on with like DeepSeek, and I think you're right, there's still a lot of things to be unpacked and kind of we got to get uncovered. But just the way you laid out how you pivoted over the last, just the last four years, I was thinking, is this going to be another possible pivot that you got to make to make the AI infrastructure, AI digital centers more to fit where DeepSeek took it? Because it took it to another level supposedly, but we don't know.
Yeah, so I think the important part here is we're still using the same compute. We're still going to use the exact same architecture. It's just how it's being used on the compute side for model training, for inference, and that's going to be constantly evolving. But, you know, and there's going to be, I'm sure, changes in compute. I still think, you know, Nvidia has an extraordinarily strong position and very big moats in the market. Even if the compute itself is used more efficiently, it's still going to be the same compute. That's what I'm talking about, that, you know, being able to bring the cost down. So, you know, because in the past when you're using a product, your cost of goods was, you know, typically, especially for tech companies, fairly low, especially if you're a software company, right? Your COGS line would be less than 10% of your revenue, so you'd have like 90% gross margins. And when you're looking at AI, you know, the compute is the primary portion of the cost of goods. And if the compute is really expensive, you need to make the product really expensive to get the type of margins to get a return on your R&D dollars. And so it needs to come down. And so this is a step in the direction. But I think it's important, it's the same compute, same data center architecture, same everything, just used more efficiently.
No, I completely understand that. And the next question I have is coming from one of the people that listens to the show. But your North Dakota hyperscale facility and greenhouse farming, you know, how are you meeting the power demands and the renewable energy that's viable for long term?
So what we have done, and this has been really the way we've built the company, right, just a matter of how we're using it, we've been very good at finding large amounts of low-cost power. And what you need for AI is large amounts of power, even better if it's low-cost. And so every site that we have done so far has been what we call stranded power. And that, you know, in North America, stranded power is power that is being generated but oftentimes cannot be used because there's not the transport grid to get it to a user of the power. So it's created one of two ways typically in North America. Large power-intensive industry goes out of business. So in Rockdale, Texas, which is a big Bitcoin mining site, it's not ours, but that's just an example of this. There was, I think it was an Alcoa aluminum smelter that used to run there, required massive amounts of electricity. They shut that down. All the electrical infrastructure, the generation, everything is still there, but there's nothing, nowhere to use it, no use of it. And so the Bitcoin miners came in and used that. The other way is typically renewable energy, wind or solar. If you know, there's a lot of incentives to build these out, especially in the US, and the incentive is renewable energy credits is how it's typically paid. And so with renewable energy credits, if you're doing a wind farm, you want to generate the maximum amount of credits with the least amount of cost. So you want to go where the wind blows the most and the land is the cheapest. And this is why North Dakota is the sixth largest wind producer in the nation and I think 48th largest in population, right? And so where we're sitting, there's about two gigawatts of wind that feed into a substation, and we sit at, and oftentimes those turbines will have to be curtailed because the grid pricing goes negative when the wind is blowing because there's not enough transmission capacity to take all of that power out when the wind is blowing and take it to an end user. So what we did, take the use case, the power use right next to the generation, which is, you know, the wind turbines, and so they feed in, and then we use that wind. Now we still have base load because you need these to run, you know, 99.999 reliability, 24/7, 365. But it's a way to go and find large amounts of power that typically are renewable energy that are not being used currently, or at least all the time, because you can build generation in 18 to 36 months, but to build new transmission lines, new power lines, you know, can take 10 to 15 years because you need, you know, all this permitting to build for long distances, regulatory. You know, I think that we're going to get looser on that with the new president. But that's what we've been very good at is finding that power, utilizing that power. It's typically very low-cost power, and it's typically renewable.
What about mini reactors, like nuclear power?
So for me, when I think about power and, you know, the amount of power we need for data centers, the amount of power we need for, you know, transitioning to electric vehicles, the amount of power that needs to be generated versus what's generated now, I think you walk through a couple of ways. So we've gone out and found a lot of the stranded power where we can take the use case right to the point of generation. That's great. There's only so much of that that exists now. There will be more that will be built. So we start using a lot of that power. You know, there's more wind farms that are permitted. But then the next logical step here is gas. And so you use gas, and there's some stranded gas in Western North Dakota and Western Texas and Pennsylvania. Those are the key three places here in the US that have it. And so you move to using stranded gas, and then further out you look at nuclear. I think it's just a long window. So you got to see how do we bridge from here to there. But, you know, nuclear, it's, you know, you probably have a guess that's as good as mine as far as when those can be permitted, built. You know, on SMRs, the small modular reactors, the, you know, I don't know what the cost level is. I think I saw some, the NuScale, I believe, you know, they had something going in Idaho or Utah that was, I think, around nine or 10 cents, something like that, which is an attractive power price. So if they're there on the technology now, the next phase is permitting. So maybe you can pull that ahead, especially with the new administration. You might be able to pull that ahead. But, you know, the way it goes in my mind is, okay, we get stranded power, then we go, you know, you got to build gen with gas, and then you build generation with nuclear.
Now you keep talking about with the new administration. Do you think there's going to be a lot of ways for you to capitalize on what's going to happen for the next four years, or do you feel real confident that your company can capitalize on what's going to happen the next four years?
Yeah, you know, the most important part is the market that we're serving, number one, and that we're doing the right things in that market. But, you know, having a friendlier regulatory environment, and it was already starting, by the way, even before the new administration started. So Biden in the last few weeks, you know, they did this push on dropping some regs on data center development. Because I think if this week has taught us anything, right, there's real competition in the world for AI. And I think this is a race that the US has to win. And so we're building, you know, the infrastructure to help win that race. You know, there going to be the new arms race. This is going to be, you know, what the Cold War was in the 20th century, right? This is of utmost importance that we win.
Yeah, no, I see. Like you're kind of equipping yourself. You know, you got a $5 billion dollar partnership and with a company growth strategy. You know, can you like break down that, like what that strategy importance means for Applied Digital?
So we announced a deal two weeks ago, I believe, on January 14th, Macquarie Asset Management, I believe the largest infrastructure investor in the world. They've done a lot of data center investing. What that does is it puts the capital pipeline in place. So we have our power pipeline, we have our supply chain in place, we have all the pieces in place to go and build this, build out, you know, what we want to build for the next five years. Now we have the capital partner to do it. And what that does is it takes the, you know, we've had to do a lot of dilution at the public company over the past year building this facility in North Dakota. What this does is it removes it and says, okay, Macquarie is going to be our equity capital. And this is important. So the equity capital, $5 billion, really unlocks closer to $25 billion of total capital because you bring project finance. Which think if you were, you know, building anything, you're building it, you go to build a house, you get a construction loan, and then you flip that construction loan into a mortgage, right? You typically get 80% of the capital is mortgage on your new house. The 20% is your equity that goes in. So that's the way you should think of ours is that $5 billion is the 20% of equity that goes in, and then we can get mortgage capital for real estate development that comes on top. That's how you get to the $25 billion number. It's really a powerful number for us. And for that, you know, we're giving up 15% of a subsidiary company, not of our public company, a subsidiary company that is developing these facilities. And so anything that Macquarie puts equity capital into goes into that subsidiary, and we own it 85% and they own it 15%. But now everyone can know looking forward, hey, we can go and build, you know, all of these campuses that we're talking about building. We have a really large amount of capital available to us to go and build it. And, you know, what the dilution is, it's 15% of that subsidiary. And so we don't have to just, you know, last year we were constantly doing capital raises to keep up with the build cost because this is billions and billions of dollars to build these facilities. So but to have Macquarie, which is one of the, you know, like I said, I think the largest infrastructure investor in the world, big name, a lot of expertise in data center, came in, looked at what we're doing, wanted to partner with us and provide that capital. And I think they'll be a real partner, not just a capital source for us, which was the thing that attracted me to them specifically.
So how, what can you give me a percentage, how this expands your growth, like the speed of it? Because I know when you have the capital raise, because I've been a part of that world, when you have to raise capital for a company, it slows down expansion because now you got to, every time you build something and you start running out of money, you kind of, oh, now we have to go back and raise capital. So we got to slow, not really slow, like put a pause, but like you got to slow down what you're doing to go raise more money. With this $5 billion from Macquarie, what is the percentage of speed you guys can see happening for Applied Digital?
So you're completely correct, Kenny. Like that, it's been a big time consumer for our manpower at the company, you know, constantly raising capital and building this and going onto the, you know, finding the next site and, you know, contracting with customers and all of those things that we need to do. So this frees up a lot more of my time and a lot of my team's time to go and focus on the building and expanding portion, so on the execution rather than constantly working on the capital side, which we, you know, we had to do because you, you know, you need to pay for these. So it just, it gives us all of the firepower financially to go and build, you know, a massive amount of data center capacity. It does it with, again, and now everyone gets to know this is the dilution, right? It's not a question because, so, you know, we're working on signing the lease for our facility in North Dakota. Soon as that lease signs, what do you think happens? I know what I think happens is people are going to be like, they want me to sign that lease, they want the company to sign that lease, and then it'll be signed, and then it'll be great, and then they'll go, okay, what's next, right? And then when you're going on to what's next, it's, you know, okay, they got a new campus, new customer, but then there would always be a question mark of how much does that cost me in dilution? Like what do they get? How much capital are they going to have to raise? And now we have this really easy formula of how we raise that capital. So not only does it free up my time, but for shareholders of our company, it lets them know specifically what the formula is and what the dilution is instead of trying to guess, you know, two, three years out of the future, like how much are we going to have to do that. Well, we set the formula. We get, you know, like I said, it really adds up to about $25 billion of capital here if you do the 80-20 piece. And it just, it provides a ton of firepower for us.
You know, with everything you said, which frees up your time, and I'm going to give you my little opinion piece, and you tell me if you think it's wrong. To me, when you get this kind of partnership and you got $5 billion coming your way, which is trying to say is really more like $25 billion, you know, I think what ends up happening to a company like yours, you can almost get more market share. Am I correct for saying that?
Well, this does two things for us. So it allows us to go faster, and if you can go faster, you can get more market share. But two, the kind of, let's call it, we're still a newer company. We're doing, you know, what I think are really big things, but we're newer. And this is such a credibility stamp, in my opinion, for us that lets us go and win more of that business, lets us, you know, sign leases with, you know, I've been really clear, like we're focused on building for hyperscalers, right? These are massive companies. And now we have a massive financial backer, and that just helps us from closing business, not just building, but closing business perspective. And so I think it does add that for us, and so that we can do a lot more than we would have been able to do by ourselves.
No, 100%. And, you know, I'm going to take it one step back. We were talking about in the beginning of like the agreement of the 85%. So that agreement gives Applied Digital an 85% ownership stake in the future APC assets. The other 15% belongs to the partnership you just made. Am I correct?
So yeah, we get 85, Macquarie gets 15.
So like 15, we're taking 15% dilution for all of that capital. And then how do you see that benefiting your shareholders, long-term investors?
The most important, like I said, is now we know what the dilution is. And there's a, you know, in year five, there's an opportunity for us to buy that 15% back to own 100% again. But like I said, it really sets the framework of how we can get the capital for our growth without constantly diluting in the public market. So I think that's just a better framework for myself. You know, I'm the largest shareholder of the company. I think it's a great framework for the company, and especially for the amount of capital that it brings in.
You know, AI is kind of moving fast. So Nvidia, those are like leading AI infrastructures. Are you partnering with like other similar companies?
