MARA Q2 Earnings | Fred Thiel CEO Interview | Top AI Stocks to Watch | MARA Stock

Great to be here.

Yeah, I think you know revenues which obviously are a function of Bitcoin price were down year-over-year. You know, the story nowadays is much more about our transition to building data centers and converting our powered land assets. And I think the real headline was more about how we've grown our capacity in the pipeline. Once we close Long Ridge and the Madagorta site, we'll have over right around 4.8 gigawatts of power and land, which means that, you know, we're amongst the largest players in the industry with that portfolio. And we're very excited about getting ready to sign our first leases and, you know, see that transition actually go through a full cycle of, you know, convert land to data center. And we're excited about what the future holds. And this is a world where power is in very short supply while the frontier model guys and the open source guys are duking it out to see who's going to dominate. The one thing that is not at all in contention or that anybody's disputing is you need a lot of power to run a lot of compute to run all of that AI regardless of what model is going to win. You need a lot of power and we think we sit on the right place and that's why it's fun to be on the Power Analysis program.

Yeah, great question. I think I've said in the past that we operate under the rule of the three C's, right? The three constraints. Capital, compute, capacity. Capacity is think of it as a built data center. Compute being what you plug into the data center and capital being how you build it all. And that was the concept in the Bitcoin mining days when we went from, you know, asset light where there was lots of capacity everywhere but compute was scarce and capital was available. So we raised a lot of capital, bought a lot of compute, grew and that made us you know at one point the largest in the world in what we were doing. Then we vertically integrated, acquired power for less than replacement cost by buying the capacity from the people that owned it during the downturn. And now we're even buying actual power generation. The Long Ridge deal is, you know, a 505 megawatt power plant with long-term commitments, positive EBIT dollar cash flow and a lot of land to build a lot of data center capacity. So very excited about that. The capital side is always going to be a challenge. But I think for one thing, data centers which have long-term contractual leases, you can get funding for those because of the long-term contractual nature of the agreement. You couldn't do that in the Bitcoin mining world. I mean, right? It's like Bitcoin miner. No, no, no, no, no, no. I'm going to run away from you.

Yeah. I mean, that's how NYDIG ended up becoming such a big miner.

So yeah, so I think you know capital is a challenge but it's a very different market. You're talking about you know the likes of Blackstone, Apollo and others who are looking to invest in these assets. It's very much a real estate play. Very different than the Bitcoin mining world which was all about being cowboys and kind of like in the oil industry they call them wildcatters. I think now as you look at traditional data center business especially around AI it's all about a capacity constraint and I think the you know if you look at the challenges people are going to be fighting about in the future whether it's model, the variety of model you're using, your cost per token etc. Energy today is such a small percentage of the cost of a data center that over time like it was in Bitcoin it will grow to be a bigger and bigger piece as the rest of it commoditizes and you know silicon will eventually commoditize software will eventually commoditize and at the end of the day the infrastructure can't. So, and even with the space stuff, while I'm not a big believer in putting data centers in space in the short term because it's a very expensive and a very hostile environment, I think longer term a lot of the larger players are viewing it as a way there's no NIMBYism, which is this American expression about not in my backyard. Government regulations are much less heavy. You just have to get the stuff up there in space and you have infinite power though you don't have a very good way to get rid of your heat but I think this is going to be a very exciting market and the latest analyst forecasts from whether it's McKinsey to whoever are showing great growth in both power demand and data center capacity demand through 2035.

And that's just the big four if you look at total the quantum this year is over a trillion dollars.

Yeah, I think part of the reasoning behind Governor Abbott's decision is there are a lot of phantom applications in the queue process, right? It's a speculative process. If you submit an application, there's no real skin in the game. And if you get an allocation, then you can turn that around and sell it to somebody. I think part of the process here is let's really flush out how many real applications there are that have real substance and credibility behind them. And so that's the reason they're doing an audit. So this is a pause so you can audit the queue. And so we're very supportive of that. We think it's the right thing to do. There are some people estimate anywhere from 20 to 40% of the queue is phantom capacity requirements. So what you don't want to do is all of a sudden you know drive people to go build a bunch of capacity that's not needed and you know the biggest issue comes then down to things like infrastructure around transmission capacity things like that which are very expensive and long-term in nature. What's great about the Madagorta site for us is its location is adjacent to power generation. It has over 10 supply lines of power running right into the facility. So it's ready to go from a perspective of the grid infrastructure. There's the grid utilities have to do nothing to turn us on. You know, we're very supportive of what Governor Abbott's doing. We think that Texas is taking the right approach by essentially trying to flush out the phantom demand out of the queue and see what's really there. And you're starting to see that they are already releasing some I think QTS was there was some mention yesterday I think about QTS maybe being told that they were getting an allocation. So I think this will all flush itself through. You know the original date was meant to be August 7th. I think we're going to start seeing some applications be approved over the next few months and it'll flow through.

Yeah. So you have two things you have to look at. One is the timeline of how long does it take for you to develop the data center which is 18 to 24 months typically. And then how long before the power is actually turned on and energized. So in the case of Madagorta, the first tranche of power comes on next year, second half of next year. And so for us what's very convenient is we still have just under 30% of our capacity is hosted with third parties. Those agreements start falling off next year. So we'll just shift that capacity away from those hosted to a site like Madagorta. Of course, the economic cost of the power there is still very attractive because whenever we acquire power on land, we make sure it's fit for use for Bitcoin mining or data center use. So, that eliminates what would be a future concern from people is what are we going to do to get replace the capacity we're losing from our hosting agreements. That allows us to take power immediately as it's available, monetize it while we're building the site. 2 gigawatts is a very big site, but all of that power won't come on in year one. It'll be over a two-year process, possibly three years. What's exciting about the site is it lets us do multiple things. It's big enough so that we can do one or more hyperscalers. We can add potentially some neoclouds. We can do collocation and we can do even enterprise critical IT load. So you can similar to how a real estate developer develops a strip mall. You have an anchor tenant or two and then you have other and that lets you optimize the profitability of the location. Because obviously you know hyperscalers are hard negotiators and other people are not. And you know one of the benefits of the Exion investment that we made earlier this year is that they are experts at operating this critical IT load private cloud infrastructure which is a high margin business. And so we have this opportunity to take the power and allocate it according to its best use which lives up to our whole philosophy of evaluating every electron and seeing how we maximize the value of it. So very excited about that part.

Yeah. So, we've been very quiet about that because a we don't want to let's just say negatively impact deals that are in process by signaling too much. We also don't want to put undue pressure on ourselves than that. But right now, we're very focused on the quality of the tenants. You know if you were to go dig up Barry Sternlicht from Starwood did an earnings call for the public debt I believe that they Starwood carries and he commented on the high number of people that were interested in the sites that we're developing together and so I'll let you listen to that comment but essentially we've seen a lot of interest there are a lot of people vying for capacity and we're really trying to cherry-pick the best people to have the right level of brand name and good credit so that we're not putting ourselves in a position where we're suffering from concentration risk or you know being under the thumb of a single potential tenant.

You know, a typical deal is going to be the way we look at them is a yield on cost, right? And so, you're trying to maximize that yield. But you also want to be wary of credit risk. And, you know, if I allocate capacity to a particular tenant and they're not able to complete their side of it, then that's a risk for me because I've now, you know, it's opportunity loss, right? At the same time, I need to think about, you know, the 15-year creditworthiness of somebody during the life cycle of that lease. So, it's not just I get a tenant and I'm done. I've got to get a tenant and they have to still be in business in 15 years. So, we're being very picky about who we sign up because it's a long obligation from the tenant's side and we want to make sure that they're going to be able to do it. Now that being said, when you have a big site, you have the ability to really, you know, play with the puzzle pieces and build a portfolio such that you may have some things that take multiple years to build, some things that can happen very quickly. You can scale the power as it's made available. You can do things like build capacity for colo in the kind of bits and pieces between the big blocks and really optimize your profit mix. And I think this is where being very disciplined in how one does things and not just signing one big lease with one big tenant and you're done. It becomes very important because even with the hyperscalers you know there are questions as to who's going to win the races here and who's going to be left you know there is no guarantee necessarily that Google Amazon Microsoft or Meta is going to be in 20 years the leading cloud provider who knows right so you really have to be very careful and I think it's all about being very prudent and looking at the long-term aspect of this because that's certainly how our shareholders look at it and you know we're obviously trying to maximize the benefits here but what I think the great thing about this transition for us and for our peers is that you're taking an asset that before was ignored which was you know curtailable power and power in the hinterlands of Texas and places like that and now all of a sudden you're generating great value with it and you know my only desire is that the people in the communities who are afraid of data centers think a little bit about the fact that while data centers don't create huge numbers of jobs, many of these communities benefit hugely from the property taxes that these data centers play and they will be long-term property taxpayers. And for some of these smaller communities, the property taxes are of such a quantum that, you know, you're talking about high schools can be funded, first responders can be funded, all sorts of resources can be funded just by the taxes alone. And that, you know, to look beyond just the fear of data centers. You know, these aren't nuclear power plants people are building in the backyards. They are huge contributors to the communities and we certainly pride ourselves in being a very big contributor to the communities where we operate and we hope that everybody else is as well.

Well, I think here's an analogy, right? If you are a startup and you're going into a market, if your timing in going into the market is wrong, you can spend all your resources launching a product and the only thing you're doing is educating your competitors. Because the market's not ready for it, right? And you know, go back to the internet 1.0. You had all these companies build all these internet companies that were fully vertically integrated, Webvan, etc. and they all failed because the market wasn't ready. Now look at how Apple enters a market. Apple is never first into a market, right? They will innovate with something new, but they know very well when they go into a market that there is enough demand to make that product successful. And when we started looking at AIHPC originally back kind of three years ago, if you look at the economics of AIHPC, inference is where the money is made by the operator. You don't make money necessarily in training. So training was just a real estate rental model where inference is where you really make the money. And so we felt that inference was the place that as inference became hit its inflection point that would be the trigger for us to come in because there would be enough people with enough business models earning money that there would be enough demand so we could maximize the value of the amount of power and land that we had. Had we just had one site, yeah, it would have made sense convert it to the first attractive person. But we had enough capacity that it made sense that we should wait to ensure that there would be smart operators which there are now were before there weren't. Where there are people who have capital, people who understand how to build these things, understand what it is they want built so that you can actually execute, right? And to be fair, and I've said this publicly many times, three years ago, if I'd gone to Google and said, 'Hey, my name is Fred from MARA. We've got all this land and power. We want to build a data center for you.' They would say, 'You guys don't know how to build tier 4. You have no idea how to build to our specs. And why in the world would I ever want to work with you?' Right? That market's changed. And so it's the timing is, you know, all about market maturity, customer maturity, process maturity, and technology maturity, right? I mean, today, if you look at and if you talk with Wes at Applied, you know, he's been at this a long time. Look at some of the first design stuff they were building to originally with their first sites. I mean it was a moving target right now it's a lot easier. So I you know sometimes not being first is an advantage. And in the technology industry there's an old saying about you know pioneers get arrows in their backs. Fast followers are the ones that become successful.