Larry Fink and Peng Xiao on Financing AI Infrastructure

So when you look at AI supremacy, which is I guess what we're going to look at, world dominance for AI, it felt like last year, two years ago, it was about talent, it was about the right chips, it was about the right technology, and now it's all about infrastructure. So you announced with Microsoft basically a $30 billion fund. That was just equity. Right. Raised another $100 billion in debt. A hundred of debts. Talk to me, Larry, about the rationale for this. I mean, infrastructure, if you don't have the right infrastructure, you can't be a huge player in AI. Well, hi, everyone. Look, we have great partners with G42 and MGX, but let me just talk about the concept. First of all, it is still about people. We don't have enough electricians in the world to fit the need in building out this infrastructure. Two, it is about national security. It has enormous geopolitical issues all around it. We can talk about what the Biden administration announced last week related to the approvals that will be needed to have advanced technologies in different countries. So it has many geopolitical issues. But as fantastic our hyperscalers are, and the U.S. right now has the greatest technology platform in the world. We have five or six major companies that are in the cutting edge of technology. But as big as they are, as large as they are, some of them over $3 trillion of market cap, some of them have $70 billion of free cash flow, they don't have enough money to build this all out. And I write about the whole evolution of the capital markets, and I believe this is going to be probably the first in our lifetimes where the capital markets, in terms of very strategic growth, is going to play a very large role. And I actually believe how the Trump administration is rolling out now, talking about making private sector capital more easy to invest by rolling off restrictions and regulation and permitting. And so the confluence of the capital markets, all the technology, all the innovation, and then you have to overlay all of that on sourcing power. Because all of this is going to require this intersection of power, capital, and technology. All with the idea that it's going to be ultimately a big leveling for the world, elevating more people in the world, providing more access to information and technology. In the short run, it is going to be about bringing all those three segments together, and it's going to require an enormous amount of financing. And we are very excited about it. I know BlackRock could not do it alone, as large as we are. I don't believe MGX and G42, I'll let Pang talk about that, could do it alone. Also Microsoft and NVIDIA could not do it alone. So it's a combination of bringing it all together. And we were excited to read in the papers about what Masa and Sam and Larry Ellison is doing on a similar project. Pang, can you talk to us about size and scope? When we talk about AI infrastructure, how big will it be? Where will it be? How much money will need to be deployed on this? Sure. First, let me start by saying this. Great partnership here. Global leading investor, and we are a very active builder of AI infrastructure. So globally today, we have about 60 gigawatts of data center capacity, globally. In our conservative joint assessment, the world needs more than 200 gigawatts, potentially over 300 gigawatts. Conservative. Conservatively, yes. Both for training and influencing purposes. Remember, to build a full-blown one gigawatt GP300 class data center, we're talking about $40 to $50 billion. One gigawatt. That's one AI data center. We call this AI factory. By the way, Larry didn't touch on one thing that's also equally important to power and chips. It's connectivity. Connecting the world with all of this. That's another massive investment required. So, as Larry said, no single company or even single government can do this by themselves. This is why we're joining hands between different investors, builders coming together under the umbrella of this global AI fund and to build this infrastructure on a global scale. So, one example I'll give to you is Abu Dhabi alone. Now, UAE is a nation of only 10 million people. But we look at our region, Middle East. We did our assessment. In UAE, to serve the region, we'll need a minimum five gigawatt of AI data center. And guess what? We're building that as we speak. So, if UAE needs five, just imagine other major centers around the world. So, this is a massive undertaking. We have this in partnership. And I would say, as we sit here in Europe with very high energy costs, there is another idea of having the data centers occupy somewhere within the GCC with the sovereignty of an embassy. Digital embassy. Yeah. So, it'll be a digital embassy. And so, that could be a solution for parts of the world that have high cost energy or countries that have to import energy. So, this is, as I'm saying, this represents a lot of geopolitical issues, too. And some very large cooperation, which would be a good thing between governments. Larry, let me ask you, are politicians aware of, you know, are they taking this seriously? Or is it really up to private companies to come up with the funds, deploy it, and get it done? I don't think there's a conversation we're not having with any government that is part of the conversation is about what role can their country play within AI? Some of them have power and energy, and some of them don't. Some of them need water, some of them do. So, it really depends on each and every country, but it is a conversation we're having in Asia, throughout Asia, throughout Europe, and North America. How does the Trump administration, second time around, change all of this? It feels like, you know, they're putting crypto, they're putting AI really at the forefront of their administration. I hope so. I think under Trump, the majority of the training workload for frontier models will remain on US soil, and it should. It's a very cutting edge advanced capabilities US should protect. But having a grid model, but not deploying it, is no good. That's why we need partners from Abu Dhabi to Hong Kong, to deploy inferencing infrastructure to serve those models to the global population. So, this is why the partnership is so beautifully constructed, and Larry is very focused on investing primarily in the US today, to help hyperscalers to build their next tier of frontier model. And we're much more focused on the global deployment of inferencing fleet of GPUs and infrastructure. But Peng, from a builder operator point of view, how much energy do you need for this? Do you have, are there new partners that you have to bring in? Absolutely. We need a lot of energy partners to be able to make this a viable global undertaking. And hopefully this raises a conversation about what role nuclear play in the energy mix. I mean, I don't believe we have the available energy, if it's going to be 300 gigawatts worldwide. Keep in mind, we have to service, you know, it's very important, if you're going to be building a data center, the data center has to be good for the locality. It can't be drawing power away from the average consumer. So, therefore it can't raise the prices of electricity, or it's not going to work. And so, in every case, you have to be working with the locality and the government, working together. In many cases, it's going to require, if we, let's assume, just round it up, I need a gigawatt of power. We are going to have to source that power. We're not going to be tapping from the grid. And if we do, Larry, if as investors and builders, we do take power off the grid, we have then come in to build additional capacity for the locality. So, that's why this triumvirate of power, data and capital is all, integrated. It's like a triangle, but there's been a lot of talk about using renewable energy. Is now that out of the window because of the Trump administration? I mean, by no means, I think every hyperscaler has long-term aspirations to be utilizing more and more renewables. But in the short run, let's be clear, it's going to be heavily powered by gas. Natural gas. Natural gas in the United States. It will be supplemented by renewables. And as I said, hopefully it raises a whole conversation about the role of nuclear in the future. And it should be a conversation we're having today. Because we're going to need, unless fusion actually works and we have new sources of power, we are going to work. The one, if you think about a Monopoly game, the worst part of the Monopoly game is owning a utility, right? You never want to own a utility. You own a bank. But these utilities, now these power companies are really important now. Absolutely. These utilities, stocks were dragging and dragging and dragging for many years. And now they're all rallying considerably. They have to play a vital role in this conversation. How important energy is, I'll give you one example in the UEE, what we talk about this formula of E equals MC squared. We actually change the formula to E equals to I. Energy equals intelligence. The more you can have that, the more compute you can deploy, the more intelligence you can generate. I don't think in a Monopoly game you still want to own a utility, but. But I mean, this could change in 2025. You might want to own that. We should have the BlackRock Monopoly edition. Right. Like no one's ever thought of that. One more thing I would say is that our friend Elon loves to say this, nuclear fusion. We have it already right now, which is solar. Yeah. Solar energy is a form of derivative of nuclear fusion from the sun. We need to harvest that. We need to do a lot more of that. In what time scale? Because this feels like such a massive project. And it's still quite theoretic, I think, in a lot of people's minds. We had a meeting on this with a bunch of power companies that launched today. And it depends on your time frame. If your time frame's two years, you have one answer. If your time frame's five years, you have another answer. If your time frame's 20 years, you have all different answers. And so it is about the immediacy. And that's some of the issue. Everybody's saying, I don't think a country wants to be left behind right now. I don't think a hyperscaler wants to be left behind, or a technology firm wants to be left behind. So there's going to be a huge amount of capital deployed. Now the question is, how do we ensure that the capital is deployed, that we get the proper return across the board, that we're focusing all the different issues. But in every conversation you have with anybody from the technology side, they're saying the speed in which these AI models are improving is accelerating. And that means the utilization is improving dramatically. Absolutely. I would just say one thing. When I look at different companies in different industries, I always highlight Walmart, who's had an exceptional 18 months. And you talk to leadership of Walmart, they believe so much of their performance has been driven by their utilization of AI right now. They have better information of their clients than most of their competitors. They have better inventory control using AI and intelligence. And it shows up in their financial results. And every company, I believe, is going to start utilizing that in a form. And so I think this is all going to be accelerating over the next ten years. But Peng, who do you think decides this timeline? So is it politicians? Is it the AI companies? I think the market will decide that. I think the demand- Not Elon Musk. Well, he's both the demand and the supply. But on the demand side, seriously, I do believe we'll see significant uptake. At this model, so you heard from Dario earlier how smart this model is getting. And they are not just one, they are multiple frontier models, very capable of pitch-level intelligence already. If you allow me to just slightly adjust topic a little bit, you mentioned crypto earlier. I think it's impossible, great conversation with Larry, we were talking about crypto. What you did with your ETF is incredible, truly incredible. And I actually believe crypto is a currency of AI. We spoke about earlier how soon enough, all of us will have multiple agents, AI agents working for us. They're going to inform us, they're going to support us, they're going to make transactions on our behalf. So worldwide, there will be explosion of financial transactions, 24 by 7. The traditional banks, the SWIFT system cannot support that. So crypto will play a bigger and bigger role in the age of AI. But at the same time, we're going to digitize currency very rapidly. And we've seen that in India, and the great success of leveling in India because of that. And you've seen the great success in Brazil. So there are two countries that are more advanced in these technology, in blockchain, than Europe and the United States right now. And that's India and Brazil, because they had a need because of inflation, because of corruption, they needed to digitize their currency. And I do believe this is going to accelerate. And that's going to be the next frontier, using AI, using this. But crypto is going to play a big role, digitization of currency, digitization of everything we're doing. I'm a big believer, and this is one thing that will change Bloomberg even, I believe we're going to be in the cutting edge of tokenizing bonds and stocks. And we have the technology now, the fact that we are now moving forward in tokenization, every bond and stock is crazy. We should be moving towards that frontier. Obviously, there's winners and losers and all that, but we need to be prepared for the tokenization. And it would democratize more finance if we tokenized bonds and stocks. By the way, that's more compute needed as well. My boss is already expecting me to be at least 20% more productive with AI. So it's going to be only maybe 25%, especially in Davos. Talk to me a little bit about what you're expecting the Trump administration to do for crypto specifically. I think it's already very well articulated. They've fully embraced crypto as a new, not just financial system, reform technology, but a new economic driver. I believe they have a thesis that to extend the dominance of US dollars, they'll be launching very powerful stipple coins to promote broader adoption usage of US dollars. I think this will be a fundamental strategy for the Trump administration. So, I mean, I look at it from the... BlackRock has been built on the entirety of hope. I mean, most of the money that we manage is retirement savings. So you ask yourself, why on earth would anybody invest for a 30-year outcome? You know, or 40-year outcome of retirement, unless they're hopeful that in 30 or 40 years, the world's better. And so the whole foundation of what we have done is the foundation of hope. And we try to promote that concept of hope and get more people to get their money out of a bank account and invest for the long term, right? Okay, and that's what we try to do. As I became a student of crypto, it was very clear to me that crypto is a currency of fear. And that's okay. For some extent. Yeah, but that's okay. Because if you're frightened of your debasement of your currency, or you're frightened of your economic or political stability of your country, you could have an internationally based instrument called Bitcoin that will overcome those local fears. And so I'm a big believer in the utilization of that as an instrument. And so if that becomes true, and you see that it could be a proper hedge against maybe hope, security, or equities, the question is, could you see a 2% or 5% allocation? I was with a sovereign wealth fund during this week, and that was the conversation. Should we have a 2% allocation? Should we have a 5% allocation? If everybody adopted that conversation, it would be 500,000, 600,000, 700,000 per Bitcoin. I'm not promoting that, by the way. So that is not my promotion. Larry, if you allow me, Bitcoin is just one use case of blockchain. Yes. Right? Blockchain is a general purpose technology. It's just like AI. So with blockchain, you can have paperless contracts. You have seamless transactions. Well, that's what you would do in the stocks and bonds. Exactly, tokenization. And one more thing I'll mention, if you allow me, Larry, is, again, the convergence of blockchain and AI. Now with AI, you can expect within this year, we're going to see deep fake, so real, you cannot tell the difference. People watching this on live stream, they will know we're not here. We're not here. We're having dinner somewhere else. So to counter that, you need some form of digital watermark. That is actually a cryptographic evidence that this is real. It's my signature, his signature. Who puts that, the tech companies? By the way, they're so excited at the back, they're trading crypto on the back of Larry's comments. It's all your fault. No, we control our private key. It's going to be a private public key. It'll be personal. We'll have control. It's going to be like your irises, your fingerprints. It'll be a digital key. But it's stored on blockchain, witnessed by the global population. Everybody knows. It cannot be faked. But does this change everything for just capital markets? It will. Yeah, I mean, my gosh, if we could tokenize bonds and stocks tomorrow. One of the great things that BlackRock is being accused of is how much vote we have. I wouldn't vote anymore because we would be able, through tokenization, the owner of record instantaneous. And each owner of record will do their proxy voting. I would love that outcome. No, but it would democratize investing in all ways that we can't imagine. Even political systems, how you vote in your country can be changed by that. And do you think that, again, the Trump administration will accelerate that? I think they are moving in that direction. How far they actually will go remains to be seen. But if you do have a digital validator, like your fingerprints or your irises, I mean, you could tokenize voting. You could do everything. And so it will change how we think and how we live and what we do. We've been talking to some South Asian governments, including Malaysia, Thailand, Philippines, Indonesia. All of them historically have the populations who are trading black market bitcoins and what have you. But now the governments are realizing the U.S. is driving a global policy shift. So they're not doing something very clever. They're saying, you know what, we're going to issue a new form of digital ID to every single citizen. Along with that is a digital wallet. Which Brazil and India has already done. You're both so fired up about this and you're so excited about it. I'm going to ask you what you worry about in the next 12 months. Is it that something goes wrong with inflation or is this that, you know, some of this gets into the wrong hands? Well, let's say cyber risk is ever present. And the amount of money that we're going to all have to make sure we're protecting. And we give us surety. And we give strength around the value of your digital identity. So all of that, that's the foundation of any order. And so let's be clear, the amount of time and effort and money, the author, you know, authenticating everything is real. And so if there is something that breaks down that authentication, then we have a timeout. The biggest risk we have worldwide today is the world believes we are past the high point in inflation. And I could see real realities of elevated inflation. I know that is not the market forces are saying at the moment, but I could truly see a scenario where we are going to have elevated inflation. We would have a higher 10-year treasury. And obviously the bond market, that's why I love the capital market. We're going to find that out. The bond market is the ultimate tester every moment in the market of what's happening globally. And so we'll see. But there is not a low probability. I mean, below 50%. But there is a probability that we could see elevated inflation again, which would be creating an elevated 10-year treasury, which changes a lot of the conversations. And this is what, because of an overheating US economy? That's a possibility. Look, I'm here in Europe, and I've never seen such pessimism. So I'm going to leave. I wasn't one who's normally pessimistic about it. And I'm actually relatively optimistic now because it's so bad. So I would always go against Davos, and you'll make a lot of money. So the pessimism is so tangible about Europe, it almost feels like it's a nice counter trade. A good entry point. What about US restrictions on semiconductor technology? Do you expect that to change? This has been an issue, frankly. Yeah, I think the Biden administration left a very difficult exact order on the desk of Trump where, really, we discussed this already, speaking very frankly here, was a gift for another country, to China. That is, if you do not supply the necessary baseline technology to all our nations, to your friends, guess who they're going to call? They're going to call somebody else, because people cannot live without AI, without infrastructure. So I think there has to be a balance, what to restrict and what to share, and ultimately work with folks like us to build this global coalition. And that's the geopolitical issues that I referenced in my opening. It's larger than that's in the ether at the moment. Are there active conversations that you know of, of either trying to restrict? And you're optimistic about that? I'm very optimistic that the next four years will bring a lot of positive changes. Okay, well, thank you both for joining us. And for a good conversation around the world.