Transcription
Welcome back to Real Vision Pro. I'm Ash Bennington. My guest is a long-time investor in Bitcoin, but his other bets in the stock market have been gathering most of his attention lately. He says that a lot of investors are missing, and I quote, "the deep fundamentals of scarce land, power, and infrastructure." We'll discuss this and much, much more with Mike Alfred, founder and managing partner at Alpine Fox. Mike, welcome back to Real Vision.
>> Thanks, Ash. Always good to be here with you.
>> It's always a pleasure to have you with us. These conversations always interest me personally. Mike, let's start out big picture, 50,000-foot view of what you see happening in markets today. Obviously, there's a lot to talk about with your stock picks. There's a lot to talk about in general with what's been happening in the digital asset / crypto space. 50,000-foot, how do you see what's happening right now?
>> Well, it's a highly interesting environment, but you could say that about pretty much every environment historically. People love to say that, but the reality is it's always interesting. Um, I do think the last few years has been challenging for a lot of investors. Um, people saw the inverted yield curve. Uh, they saw like the banking crisis with, uh, Silicon Valley Bank and, um, they saw potential recessions coming, potential credit crises, the Japanese yen carry trade, uh, more recently geopolitical type of issues, and it created, in my view, like a compressed cycle where basically the only thing that has benefited over the last three years in mass is AI. So, like if you look at the S&P 500, it's large companies with the most exposure to AI, companies like Nvidia. Um, and then when you look at the rest of the S&P, there's largely been a lack of, uh, participation. Like stuff in healthcare and staples and, uh, other other sort of sectors that are not as directly correlated. Uh, and so it's been, it's been kind of tricky, and obviously if you've been in crypto or only in crypto, it's been almost impossible to make money with the exception of just a sort of a straight long Bitcoin position, which is really the only position that I've sort of publicly espoused over that period, and it's, it initially outperformed a lot of equity positions, and now is sort of fallen back in line with sort of a lot of the AI companies. If you look top to bottom or bottom to top from, from kind of Q1 of 2023, it's very similar returns. People forget that Bitcoin actually did pretty well in '23 and '24, and they just remember sort of the recent past. But the reality is like pretty much anything in AI, anything in Bitcoin has done fine over the last three years, but even now people are sort of forecasting sort of a bubble in AI, a potential recession coming, a potential major drawdown in the S&P and the Qs. And remember, we just had one, right? And in sort of Q1, people were worried about Iran. Last year we had a pretty major drawdown into April and then a V recovery. We've had major scares at multiple points over the last three years, and yet everything has more or less made it higher. So I, I, I think the environment is sort of ripe for continued idiosyncratic investing. So if you're focused on very particular themes, you, you get those themes when they're out of favor like I do, and you tend to buy them in a concentrated focus way, and you're willing to hold them through all of the noise, right? Cuz there's also going to be a lot of noise no matter what you own right now. It's people telling you, "Hey, this is a bubble. This is going to blow up. You're going to lose money in this." But the reality is like I've been hearing that for three years, and yet everything that I own at least is higher. So I'm just going to continue kind of focused on what I'm focused on.
>> Yeah, Mike, talking about what you own, I want to walk through this your journey in this space. Specifically, I'm interested in your understanding and your journey into the miners and how that's transitioned into the AI bet for you.
>> Yeah, so they were called Bitcoin miners, but in, but in a sense that was really the wrong nomenclature from the beginning. What, what was really happening is that some people were building infrastructure, like heavy-duty industrial grade physical infrastructure, you know, buying the land, building the buildings, doing the electrical infrastructure, interfacing with the grid. Some of those people didn't think of themselves as Bitcoin miners, they thought of themselves as infrastructure developers, and then other of those folks thought of themselves as asset-light Bitcoin miners, where they didn't actually build anything. They just hosted their machines with someone else cuz they really wanted the Bitcoin. And I focused heavily three, four years ago on the firms that I thought could be world-class infrastructure developers cuz it was my view even then that what was really scarce in this business was, was really high-quality infrastructure that could scale over time. Uh, that could be run with a really high uptime without a lot of failures, right? At a low energy cost. I believed back then that you needed to do that in places like West Texas and the Panhandle in Oklahoma in the middle of the country where there was a lot of excess renewable power. That was not a common view. Uh, the sort of prevailing view three, four years ago was that all data center development would happen in places like Northern Virginia because that's where it was done before, but if you ask people why they did it that way before, they said, "Well, cuz the guy before me did it that way." So, there was no real like first principles thinking about it, but if you really understood the sort of explosion in in compute that was coming, and again, we were looking at the Bitcoin mining space, we were looking at AI. If you remember, Iron had an MOU with Dell specifically on AI in 2019. So, seven years ago. So, this idea that these companies pivoted in is sort of correct for some of them, but not for all of them cuz there are a number of folks that had it in their mind from the beginning as they were developing this infrastructure that it might be used for things like synthetic biology, it might be used for things like high-frequency trading and and hedge funds, it might be used for things like AI. All that has come to pass now. You've got Jane Street and XTX building data centers, right? You've got people trying to do drug discovery with AI building data centers, you've got people mining Bitcoin building data centers, and you've got people trying to do inference. Right? And and so all that has converged and the real choke point is hasn't been chips because actually Nvidia has a lot of great chips. Uh, unfortunately, some of them are sitting in warehouses cuz there's nowhere to to plug them in. So, the gating factor has been can you spin up, uh, heavy-duty infrastructure, right? That that, uh, you can keep online and you can run it at low price, uh, that you could provide to a customer. And you look at, you look at what, uh, XAI and SpaceX just did, right? They, they thought originally that Grok was going to be a key demand driver for for AI compute. Turns out they weren't that competitive with Claude and Gemini and some of these other, um, platforms. So, all they did was they just pivoted and now they're actually, uh, a neo cloud effectively. Like the, the kind of quiet thing that's gone on is SpaceX's maybe biggest opportunity in the short term is to build out their neo cloud business. Right? Because they had all this compute that they thought Grok was going to use and they didn't, so they flipped it to Anthropic and Google under what I think are pretty good contractual terms. Uh, so, so that's kind of where the world's going. I think whoever can wield the most compute at scale is, is, is going to win, uh, a lot of battles. It's unclear who, if there's any sort of larger war that gets won, like who who wins it. But right now, you know, obviously companies like Anthropic are clearly out in front and people can see it, um, in terms of the way that they're, uh, getting customers, but also in terms of their financials. I think they're already showing, um, you know, EBITDA positive, uh, numbers for this quarter, which is, which is incredible at the scale that they're growing. Uh, and it's pre-IPO. So, I, I, I really, I really think that this space is, is still in the first or second innings. Um, I think what will happen over say 20 or 30 years is because there'll be macro recessions, people will temporarily believe that this space is over. So, they'll say, "Look, that was a bubble and it popped." Sort of like when the internet bubble popped between 2000 and 2002. A lot of people came out and said, "Oh, see, the internet's dead." And then the internet went, went like a, uh, and so I think AI will be similar where there'll be these periods of expansion like we're in now. Uh, those will be punctuated with periods of of contraction because of usually external factors, macro, liquidity, recessions, et cetera. And then a year or two later you'll go back into a boom cycle again and that'll probably happen three or four more times over the next two or three decades. And so it'll be a mistake in my opinion, um, unless you really have to manage for like very short-term drawdowns and short-term performance, uh, to become too bearish on the sector at any point along that curve. Because again, if this is 1999 internet, yes, you want to be careful the drawdown, but you probably want to be buying that drawdown. Uh, especially if it's large like 50 plus percent, uh, because five or 10 years after that you'll probably be significantly higher again. I think this is a, a, a mega trend that goes on for many decades.
>> All right, so let's start to unpack this. First, I should say as always, not financial advice, not investment advice. All investment involves risk. Uh, and individual risk tolerances and individual circumstances always vary. Uh, let me ask you this, as you frame it, you talk about it as an investment in compute. Talk a little bit about some of the mental models that you apply, that you think about when you make those investments, how you evaluate, understand, and assess those investments.
>> I mean, there's so many dimensions, Ashton, and I'm very tactile, right? So unlike a lot of the other investors in this space, like I'm actually willing to sit on a board. Right? I'm actually willing to to sit with the CEO and and strategize, uh, directly with them inside the board room, you know, inside looking out versus outside looking in. Um, and so a lot of the heuristics that I use are not going to be available to to most investors. Um, but again, there's a price for that cuz when you're on a board, it means you can't trade as much, you know, if, if, if there's a blackout window because of earnings or because if there's any material non-public information in the market at that point or or available to the board, then you can't trade. You got to get general counsel approval. There's an SEC Form 4 if you sell. So, you, as I've stated many times, I've never sold a share of Iron, even though it's gone from 28 to 1 to 75 to 28 to to 60. Um, most investors would have loved to trade that, but, but, you know, you can't. So at a high level though, my, my, my focus from the beginning is on on margin of safety, right? At from the investment side. And and looking for things where there's a high degree of asymmetry. So very limited downside from some level where I'm looking to enter in size, and then and then significant upside. And so when I was looking at Iron and Cypher in particular, uh, you see that you're showing this chart on the screen now, like I was looking at it back in 2023, 2022, when the when the stock price was actually, uh, trading around a dollar. Right? So I think it was late December, Cypher hit 39 cents at the absolute lows, and and Iron hit a dollar and two cents. So that period right there, that V bottom, the end of 2022 is when I was most interested. Now, I was already on the board cuz I joined the board just before the chart started, uh, the month before the IPO, October of 2021. I bought a couple thousand shares in the open market on the way down, but I really, I didn't really get moving, uh, until Q1 of 2023, which is when I started to deploy in more size. Uh, my analysis said that the value of Iron's, uh, existing sites at that time in Q1 of 2023 was maybe two to four X more than the current market cap. So effectively, if the business just failed, and I, and from the board I liquidated all of the underlying sites, all the power, all the land, all the interconnects, all the existing machinery, etc., that I could get some multiple of the current market cap, and my view at that time is that on the upside scenario, if they executed really well in the coming years, it was a 50 to 100 X, uh, type of opportunity. And of course, that, that's what happened. Um, Iron went from a dollar to 75, uh, 76 I think was the high in in November. Um, I don't think that's sort of the end of the story. Obviously, I'm on the board, so I can't tell you give you advice to buy it, and I certainly don't know any specific reason why it would do that, but I just modeling the demand for for AI compute in the coming years, I think anybody who controls and wields a large amount of it is in a very good position. And that's exactly what Jensen Huang and the CEO of Nvidia and others are saying where like if you look at the field right now, there just isn't enough compute for the demand that's available. If there was more compute, it would all get sort of used up almost immediately. Uh, so that, that, that's kind of the, Again, like you can't really understand my thinking on this if you're looking at it today and asking why or is my position size the way it is. Um, because the main decisions that were made for me, like basically all the decision-making happened at the end of 2022 and 2023. And my position size hasn't changed, uh, up or down since then. Um, and I intend to continue holding, right? I can't make any promises to the market, but like it's unlikely I'd be selling anything cuz I don't think we're anywhere near fundamental value yet.
>> Well, you're always very clear about your disclosures there. You've got those that your investment and the board seats. Uh, so what we can say though for those particularly listening to this as an audio podcast, it's now trading at about 58 bucks a share on my screen here today. Mike, let me ask you this. Talk a little bit about your fundamental view of the business model. Talk a little bit about what it is they do and how you see it.
>> Well, they, they develop the, the physical assets that you need, uh, in order to run AI effectively now cuz they're, they're pretty all in AI at this moment. Um, but they, so they find the land, right? They find the possible interconnect agreement to connect to the grid. They develop that, right? They secure everything they need in order to to take advantage of that opportunity. Uh, they, they once they buy the land, they, they acquire all the long lead, uh, assets, right? So the, the, the generators and the, the substation equipment, right? And the electrical infrastructure, the transformers, etc. Those all get shipped to the site. At some point, you may have thousands of workers on site developing some of these sites to get them to to scale. And then a lot of people stop right there. So the big divide in, in this sector is is folks that are doing what's called colocation, which is basically where you're like a landlord, right? So you build out the whole, the whole site and you lease it to someone else and then that person plugs in the, the chips that they acquire in the data center. And usually monetizes that with another external customer. Right? And so if, if you're doing that model, right? It's, it's, it's sort of less capex heavy because you don't have to buy the GPUs, the GPUs being the most expensive component of this stack if you're thinking about delivering compute from top to bottom. Like you need the land, you need the power, you need the electrical infrastructure in the buildings, everything. But eventually, once the buildings are constructed and plugged in to the grid, you actually need to put the racks and the servers and the chips in the building. And so that's where a lot of people stop. And I think that some of those businesses, the Cypher and TeraWolf and a few of the others, will be quite successful because they're going to get really good at at doing that. Iron's decided that they actually want to wield the compute on top themselves. So they, they want to actually do what, you know, SpaceX xAI is doing. They want to do what CoreWeave and and Nebius and others are doing. The, the main difference though between these, these companies is that not all of them are vertically integrated. You know, SpaceX built their own via xAI, right? They built their own data center in Memphis called Colossus. They developed it from scratch, like they did it very quickly. And it's been pretty successful according to most people in terms of like how well they were able to to do that. And and I think Nvidia gave them a lot of accolades for for that. Some of the other neo clouds are have been more asset light historically. So somebody like Coreweave from what I understand has like 20 plus external providers that they use that are physical operators. So they're, they're, they own the compute. They own the chips. They're monetizing it directly with the customer, but they don't actually own the rest of the physical infrastructure stack. And I think the lesson from sort of the previous Bitcoin mining industry is that if you do it like that, there's a chance at some point that your economics don't quite work. Because you're paying a hosting provider. There's also a risk that you don't actually deliver the service the way the customer wants because you can't control every aspect of it because you don't control the full verticalized stack. And so my bet personally is that if Iron could pull this off at scale, it'll be very hard to compete with them. What'll happen is people are going to try to converge on both ends. Right? So some people who are sort of asset light now are trying to acquire infrastructure. And then some people who are all in on infrastructure now at some point are going to build their own cloud too cuz they're going to realize they don't want to give up that piece of the economic stack. We'll see. But over five, 10, 15 years, I think there's still quite a bit of opportunity here because the market largely still doesn't understand these economic models.
>> Before you go, that was just the preview. The full conversation goes much deeper. What's really driving markets, where the risks are, and how the best investors are positioning. That's what we do at Real Vision. We connect the dots before they become obvious. So don't stop here. Watch the full episode now and more on Real Vision.