Bitcoin Investor Week | Fred Thiel (Chairman and CEO at MARA Holdings) with @AnthonyPompliano

Yeah, I think if you look at the big demand right now for energy is from obviously the HPC AI sector and they need energy that is on now and not on in 2029. And if you're trying to get access to 500 megawatts of power and get in a queue, you're not going to get it anytime soon. So the hyperscalers are now very focused as are the neoclouds on looking at okay I don't need a one gigawatt data center I need anything north of 100 megawatts which has the ability to grow and I need the power to be on already because there are two issues that impact a hyperscaler's ability to bring capacity online and realize the timing of when they bring their site online is more important than when it costs them to bring it online because it's all about a capacity war. Who has GPUs running? It's not about necessarily the specific type of GPU, but it's who has stuff running because the more stuff you have running, the smarter your model is, the more you can do, right? So permitting is the second problem. So you may have the ability to take powered land and build a hyperscaler site on it, but if that county has a moratorium won't allow you to build a big hyperscaler center, it's an issue. And I think what you're going to see now is a huge amount of innovation around what defines a data center and how they're built that is going to make it much more attractive for hyperscalers to talk to people like us who have over a gigawatt of power that's turned on and ready to go.

So we have 1.1 gigawatts of power that's operating as Bitcoin mining today. We have another 700 megawatts of power that is unenergized if you would, that is expansion capacity for that and we have another couple of gigawatts of power in the pipeline behind that. A lot of our growth for Bitcoin mining is actually targeted internationally. And the reason why is in the US to get 100 megawatts of power I have to fight a hyperscaler or somebody who's trying to get it to a hyperscaler. But internationally there is much less demand for AI and hyperscaler capacity. You can go in lots of countries in the Middle East, you can go in countries in Europe and a big site for them for AI is 40 megawatts. And so you have power dislocation economics happening around the world where you have places like Saudi Arabia where they have built huge amounts of power infrastructure specifically to generate hydrogen and the hydrogen market isn't developing. And so you have all this renewable energy that's available but isn't connected to a grid to be sold and there isn't a city anywhere nearby that is available that they now want to monetize. You have countries like France where 66% of the energy is nuclear, the rest is mostly renewable and the capacity of the nuclear power plants is running below 70%. And so they have way excess energy, but there it's a permanent issue. But so I think you look at the business that we or our peers are in and I think you're going to see most of the US miners will do some portion if not all conversion towards AI HPC domestically and those of us who operate internationally will continue to grow Bitcoin mining internationally where it's much more economically viable today than in the US.

No, I think listen the Bitcoin blockchain is oversecured, right? I mean, we have way too many people mining Bitcoin. You don't have enough transaction volume today to support the industry. You go to the halving in 2032 and a lot of miners are going to be worrying about what the business they should be in.

We operate in a market where today the cost of a terahash of Bitcoin mining capacity is at its lowest point most probably ever. You know you can buy mining rigs today for $89 a terahash. When you go back to the peak in 2021 it was $80 a terahash for a much less efficient machine. What is driving that? That's tech innovation driving it down. It's a combination of oversupply and no demand. Think about it. 40% of miners in the US are the public miners. The vast majority of them have all said, 'We're going to do HPC or AI.' They're not buying any more miners. Bitmain has shut down what was going to be a US assembly facility. They're trying to place machines anywhere they can to put them to work. Nobody is buying Bitcoin miners because right now unless Bitcoin price moves up, there's no profit margin in adding capacity. You're literally just taking money off the table from yourself. And so it's a perfect time to expand if you are wanting to expand somewhere. And it's a marketplace where really energy rules and you have to have energy and that drives everything.

No, I mean data centers in space, I don't know if any of you have ever built electronics that have to operate in space, but your average GPU will die a quick death from gamma radiation in space and there's no way to service it. So I think it's a great Elon thing for him to focus on. Build data centers on the moon. It's a moonshot type of project, which is perfect for him. Data center innovation is, you know, if you look at Bitcoin mining, it costs you a little over a million dollars a megawatt for infrastructure and miners. Infrastructure alone in the data center world is 12 million a megawatt. Add to that another 24 to 30 million a megawatt for GPUs. That's a crazy cost. Now granted, great revenues, but if you can at least take the infrastructure down to a million dollars a megawatt and solve that problem through modularizing a data center. And in the world of inference, you don't need to have all these huge clusters that are interconnected. I think you start seeing all of a sudden an ability to take advantage of power in all sorts of very interesting places. And just to be fair, you know, revenue generation in the AI world is only done through inference. That's what you pay for, right? You don't pay for training unless you're building models.