250. Inside the US Datacenter exuberant buildout - Oct26

0:07 With Laurent Seguelin from London and Gerard Reid from Berlin, this is Redefining Energy.

0:13 Today on Redefining Energy, we're going to talk about data centers. And in particular, we're going to look at the U.S. market and what's going on there.

0:20 Gerard, if you look at the global numbers, it's insane. The worldwide data center CapEx is now expected to surpass $ 3 trillion by 2030, doubling what we have right now. It's the biggest GDP bet in America since the railroad. But first of all, from a partner.

0:39 Right now, it feels like the energy world is one wild ride.

0:43 Changing at dizzying speed.

0:45 Twists, turns, steep climbs, sudden drops. But volatility doesn't have to slow you down. With the industry's largest proprietary data set, expert analysis, interconnected insight, and synoptic AI, Wood Mackenzie helps you quickly react to any sudden change.

1:02 So you can be sure.

1:04 Wood Mack short.

1:05 Wood Mackenzie, Intelligence Connected.

1:08 Back to the show. And it's burning cash. It's insane. So I've got a few numbers. KKR Credit estimate that AI-linked debt for data center, 600 billion. AI infrastructure company, 40% of the S & P 500. AI-related companies, 49% of investment-grade bond insurance. And if you look at the private market, 87% of VC funding goes to AI. Of course, we've seen the entropic filing for IPO. I mean, it's wonderful.

1:40 600 billion of commitment to Amazon, Google, and Microsoft XAI. The amount of cash is crazy.

1:49 And I think it's really important to dig into this team because it's highly complex.

1:54 Yeah, I know.

1:54 And Laurent, as a result, it's really complex. important to dig into this theme. And that's why it was really great to get Eugene McGrain to come on the podcast to have a discussion around this topic.

2:05 Eugene is executive managing director at Cushman Wakefield, which is one of the world's largest real estate broker. They used to do warehouses and they used to do office and somehow they acquired the expertise to do data centers. Now, when somebody say, okay, that data centers is worth a hundred billion. In fact, you get 50% chips, and then you get 20% power, and then the rest is cooling and everything. So it doesn't enter in the chips per se, but you organize everything that's around the world, the permitting, the connection, and so on. So I think it's very important to see that data center wave from the inside.

2:45 Absolutely.

2:45 So let's bring him on the show. Eugene, great to have you on the show.

2:50 Thanks for having me.

2:51 Eugene, a very simple question to start with. You are a real estate expert. How did you become a data center specialist?

3:01 Well, like all things in life, a series of coincidence. I've been a commercial real estate broker for 20 plus years now. I started in California and I was at the beginning just an office broker, you know, working for corporations and dealing with their office leases, which are kind of just complex contract negotiations and market knowledge. Through that time, I worked for companies like PG & E and people like Chevron dealing with their real estate issues and assets.

3:26 And I was very fortunate. I had a dear friend at Chevron, and he recommended that I, you know, I wanted to learn more about their business. He told me to read Daniel Juergen's book. So three books later, I felt like I knew enough about the energy business to not know much. And in 2021, my wife, who's from Rhode Island, needed to get home. And so we moved our family across the country. in that very unsettled time in the world where, you know, people were not signing new office leases or, you know, warehouse leases with much certainty.

3:57 The political climate of the time made a giant push for offshore wind. And there was a commitment to do 35 gigawatts of power and deliverables and contracts were done and investors were attracted. And, you know, with my background, I could see two things. One, that this was real, that the investments were going to be made. And two, that the companies, especially the Danes, were coming over with a completely different understanding of the power grid and how development worked here as to how it worked in Europe.

4:30 And so I started advising those companies on what the kind of different obligations were going to be in order to reach the development. But they had interconnections going through towns that they didn't really understand the permitting process and what the obstacles were going to be. And so as I was doing that, I was doing a lot of infrastructure level work. power because that's what those developments were.

4:51 And then when ChatGPT hit the market and our kind of worlds changed, I just happened to be one of those people that was working with companies that were doing 200, 300 megawatt projects and kind of understood the basic stuff that those people did, which was not typical in the data center world prior to that. You know, a large requirement was five megawatts, 20 megawatts. all of a sudden we were looking at 200 megawatts was, oh, maybe that's not enough. So it was 200 to 500 to gigawatt to now what it is. And so the expertise that I've been able to grow and build is one that's based in real estate and complexities of developing property, the financing aspects of developing property, what has to happen, how

5:36 the steps go, what you have to get through to somebody that kind of understood the PPAs and process that power companies go through. So I just happen to have kind of a unique set of where I am. And my partners, who are very, very good at the data center end of this, just partner with them, and I sit in between, and I think about this stuff all the time. So that's who I am. That's how I got here.

5:58 Yeah, it's a great story. So in a way, you are the right man at the right place. Because I think all the data center guys, they're good in the soft thing, but when it comes to hard, the hard part, which is the walls, the terrain, the the relationship with the authorities and the connections, you have done that before.

6:17 How fast has your practice grown in the past three years?

6:22 The amount of time I spend on this has grown significantly. These projects are so complex and so large that from a percentage-wise, they're like 100% of my practice, but they take so long to get there that if I do two or three of these or four or five of these, That's a lot of them. These are sort of kind of career-sized projects to work on. It's scaled significantly.

6:47 Eugene, can you just talk a little bit about the hype in this whole area and how you navigate through that and what your thoughts are on it?

6:56 Well, it's interesting because it's evolved. When this first happened, every farmer and every brownfield owner of property in the country all of a sudden told everybody that they had 200 megawatts or 500 megawatts or some level of capacity, or they could get interconnection because they knew a guy at the PUC or whatever. And because of the size and complexity of these assets, I mean, we're almost at the other end of that now, in my opinion, where the hype is more narrative-driven than it is on the actual ground of people that are making decisions and people that are writing checks. This is now the part where the really sophisticated developers have taken it to a place where they have to walk through the individual steps and

7:43 the press is against them because the story is against them. So the good comparison is essentially where the fracking business came from. First, it was this energy revolution to an environmental disaster to what it is now, which is really kind of energy independence and changing the map of the world.

8:05 The press and the public sentiment became so strong against these projects that they really had to fine tune what the technology was and they had to fine tune what the message was. So the height now is much more in the financial markets than it is on the ground.

8:23 That makes sense to me. And maybe I'd just follow it on from that then. When you look at it, I suppose I'm trying to get my head around.

8:30 Do you see the.

8:30 Glass as half full or half empty when you look at the potential, right?

8:35 I'm a pessimistic optimist. So one of the questions that we always get and that I've answered in a bunch of places and I think is an ongoing question is, are we building data centers and infrastructure? Or is this the 1870s railroad boom? Or is this some sort of critical resource that we're building and this is going to define the next 50 years? We're somewhere in the middle.

8:58 There is hype in the sector. We talked a bit about some of the players that have emerged as people that can deliver these products. But the vacancy rate, which is, you know, as a real estate guy, the first thing you look at is the vacancy rate. That tells you what kind of delivery there is. The vacancy rate in these large scale, let's call them 200 megawatts and above projects is less than 1%. The overall market of the Americas is 3.4%. And that includes everything, which is a lot of legacy stuff that can't really be repositioned to deal with the kind of higher compute requirements. So if there's a collapse in demand for compute, that would be pretty fatal for the industry. I don't believe we're going

9:41 to see a collapse in demand for compute. We may see a collapse in paying for compute, and that may rattle the people that have done the infrastructure investments. But this is stuff that, no matter what happens, is going to be used. It's equivalent to the fiber build-out of the late 90s. 90s, where fiber got built everywhere because there was a supposedly endless demand for fiber. Well, it sat in the ground in a lot of places for a long time, and now it's being used. So glass half full, glass half empty. I think we have a near insatiable demand for compute. We have a very limited ability to deliver it at the scale that the demand is asking for. And this is essentially the infrastructure build

10:26 out to meet that demand. Will there be carnage in the financials? Certainly. Somebody's going to lose something somewhere, but this is not pets.com.

10:36 Eugene, I like the way you also started there because You also put it in the context of the railroad build-out and all that type of stuff. That was a revolution, a transport revolution. I think that we're going through the exact same thing. There are going to be winners and losers, but the bottom line is you just need to have to be careful.

10:54 Do sensible things.

10:55 Don't do stupid things.

10:57 It's always easier to do stupid things. But every day I read the headlines, and then every day I answer my phone. And the headlines on the phone don't match up. in a lot of ways, which is the demand continues to be very strong for the needs for this. And then everybody you're dealing with is using AI. It's incredibly useful.

11:16 We've been generating data for 30 plus years that allows for real productivity. You know, everybody wants better information, and this is the tool to get you there.

11:26 If I go back to power and real estate, I remember, because I'm an old man, 27 years ago, Mark Mills... Mark Mills, who is worshipped every time he goes to Sarawak, said, so end of the 90s, by 2010, half of the electricity will be used by the internet. Okay. And basically that triggered a lot of construction of gas and so on, which then led to the collapse of Calpine and all the bad. He was the big evangelist at the time.

11:59 He's still around. So now, of course, he's saying we need to multiply power by a factor of 10 and Now, I'm interested, if I look from a basic point of view, NVIDIA chips. So you had the Hopper, then the Blackwell, now the Rubin, and we're going next generation to the Fineman. The energy efficiency of the Fineman is going to be 500 times more than the Hopper, which goes to two questions.

12:26 The number one is, what's going to be the value, especially if they have those 20-year contract of a Hopper data center in 20 years? considering the speed of the improvement. And the second one is the guy everybody loves to hate, but is an engineering genius, is Elon Musk. If I look at the development of Colossus, now we've got this new building called Mini Hard. And Mini Hard is about a quarter of the size of the one he built two years ago, and it's even more powerful. It's very difficult to assess, but that's something that the hyperscaler industry has done. The energy efficiency gains that people are going to do which for me is a bigger risk than the demand for compute. The question is, do you have

13:08 this type of thinking or no, as long as the telephone rings, everything's okay? No, no.

13:15 The answer is the projects themselves are the limiting factor on the growth. And because of the size of the projects, because of the complexity, because of the software guys have never had the plan for power before. They just iterate the next level of software. And it's just been easy. So they've never had to do this hard infrastructure level work.

13:38 Meta at one point wanted to build its own Bayside mass transit rail station to run from Palo Alto up to San Francisco. And they hired a whole team. They did all the things that would build an actual rail system. And they got nowhere because it involved things that are, you've got to get approvals, you've got to get Materials, like all of the hard structural costs that are involved with infrastructure, are in themselves a limiting factor in the speed of this sort of build. Now, this feels like it's moving at some level of warp speed, but I would make the argument that it is basically rate limited or expansion limited by the funnel of, you can't get a turbine from GE Vranova or Hitachi for three years.

14:27 You can't get a transformer from those same people for two or three years. There's a queue. So the demand is so strong and the inputs have so much trouble flexing up that I don't look at it as, oh, we're 50 gigawatts ahead of where compute demand is going to be in 10 years. I think where we are is we're probably par or less than par of where we're going to be in five to 10 years especially if the chips are able to get more efficient, because I don't think you're going to see compute demand drop.

14:58 What we've had is it's worth looking at the first internet boom, because that is kind of what's informing a lot of people's prejudices on this. It certainly informs mine. I grew up in San Francisco and watched the first dot-com boom. That was based so much on software.

15:13 It was free.

15:14 Everybody could figure out how to write a new website. And the money that was being thrown at those companies was being thrown at companies that were just basically ideas. They didn't have to go out and really do anything. In order to get this level of compute that's demanded today, you've got to build infrastructure that will be valuable in 20 years. And if it becomes a factor of the chips continue to get more energy efficient, it just allows them to meet. Now, there's going to be some overlap.

15:43 There's going to be a tipping point where we get beyond that. I don't think we're there. And I think that what we'll see is projects get abandoned as opposed to projects being built and then not being used.

15:54 What I just like to add to that, to put it into context, is that Google, just Google alone, Google has had a sevenfold increase in demand for processing power in the last 12 months. 700%. Now, of course, what you don't have is 700% increase in energy demand, but But let's say you have a doubling of energy demand because of energy efficiency and new chips and this, that, and the other thing. The point I'm making is that's going to continue. It mightn't be seven-fold next year, but let's say it doubles. That's what's driving demand. And that is all of us sitting there. And I don't use Google anymore. I use ChatGBT to do my searches. I want them to do an invitation for something.

16:37 Guess what?

16:38 I'm using it.

16:39 That's the bottom line. We are driving as consumers demand. massive demand for data because it's useful. And I think it's just going to become more useful. Yeah.

16:50 Eugene, as you said, there is this crazy thirst for more compute, more buildings and more chips, more everything. And of course the turbines, I mean, we know that, but you as a professional, when your telephone rings, And it's a new client or new developer. How do you assess how serious that this deal will go through rather than just pie in the sky? You have a secret sauce here.

17:18 My first boss always answered the phone. You never know when somebody is going to want to buy the Empire State Building. I've been trained with a certain level of optimism in answering my phone. These projects, one, they're so much more interesting for me to work on than a million square foot warehouse for Amazon. It does a great project.

17:35 They're very lucrative. But even at the most technical, it's just high bay warehouse. An office building is just a lease with a subordination, non-disturbance, and atonement act. It's nothing real great shakes. This is a real technical issue. So the way you qualify these things is the level of professionalism on the other side. One of the groups we're working for, the way that I was introduced to them was the former head of BP's trading desk went to go work for them. That, to me, spoke of a level of professionalism and a level of understanding of the market and the economies of scale. That was a real person.

18:11 And I think that's something that we should understand. My friend who I was on the phone with yesterday was a nuclear engineer for Dominion before he went over and worked for Meta as one of their heads of energy. They're not hiring guys like me to run these things. The guys like me are figuring out how to get them delivered. But the quality of the people that are involved in this are really world-class. And if you don't have that group of people, if you don't have the market expert to get into, we have a team. I have a broker in Dallas who has been doing data centers in Dallas for 30 years and worked for HP when HP was building out their first infrastructure. You have to be willing to assemble that kind of group. That's

18:52 the qualification. In the United States, we're broken up into ISOs. And the ISOs we're getting I wouldn't call it artificial demand, but they were getting a lot of, hey, maybe this will happen in their queues because people wanted to just try to get it placed in line. They've all basically come back and said, you need to pay to play and you need to pay real money. So ERCOT's charging quite a bit of money that's non-refundable if you want to place on their interconnection queue. And you start talking about 300 or 500 or 1,000 megawatts, you start talking about having to put real money into the utility system.

19:26 So That's a lot of what drives how serious I take somebody that picks up the phone. These are large projects. You should be real people.

19:34 Eugene, recently we've seen the political landscape, which used to be pretty favorable. And now you've got governors all of a sudden. So the light, they have an epiphany and they think that politically it's easier to go with the grumbling. of the people rather than the developers. So do you think it's short term or there's really something in it that's going to hamper the possible growth?

19:59 I mean, this is one of my kind of favorite observations. So I was in California when Jerry Brown, whose father had been governor, who had been governor in the 70s and then got reelected as governor after Arnold Schwarzenegger was governor, and really pulled California out of the recession. And it was a very, very, very savvy operator and was somebody who had been very pro-development. He turned the city of Oakland around. He banned fracking in California. Now, most people don't think of California as one of the largest oil-producing states.

20:33 But it certainly was. And the reason that Chevron was headquartered there is that was Standard Oil's West Coast branch of their oil production. And the history of oil and energy production in California runs so deep and the expertise is so, so serious and the geologists are so good and the technology is so good. We did not understand what hydraulic fracking was going to end up being at all in maturations. And that governor, who should have probably known better, led an effort to ban fracking in California because hydrocarbons were such a dirty word. And I grew up in California in the 70s, 80s, 90s. And I remember going to Southern California and you could eat the air because it was so bad. So there was certainly

21:20 reason to be anti-development of new energy resources.

21:25 However.

21:26 Here we are now. We still need that energy in order to make the transition into a cleaner energy environment. I look at data centers. I look at this growth of hyperscaler mass data collection center as the major tool to modernize our grid because we haven't had an economic driver that has done this as a business case in my lifetime. There hasn't been anything that could put this kind of capital to work. to upgrade what is a 100-year-old grid. And so I think there's a deep level of short-sightedness for the governors that are reacting to this with what will be long-term bans. I think that there's room for both the technical improvement that the data center developers are doing to address things like noise, water, and

22:17 energy consumption. But the fact that these guys are a 24-7 baseload, standard load operator allows so much work to be done on the grids that need the work. I go to California all the time, but I'm in the Northeast now. The grid is 100 years old. Natural gas pipeline just south of San Francisco exploded, took out a whole neighborhood because it was a 100-year-old pipe. And so the only thing that can drive this level of investment is private industry with a real demand for the power.

22:47 And that's what I think is coming. And that's what I think will be the product of this as the noise of the opposition calms down. And we're also in an election year.

22:58 So Eugene, I like your vision and I agree with it because I think we actually do need to see AI as part of the solution. There's a huge amount of flexibility there.

23:08 They all have backup power.

23:09 They've got batteries in place. They've storage everywhere.

23:12 You can move.

23:13 Demand loads around the place. These guys have got great technology. Google, for example, has tapestry. I'm totally with you on this is that I think it's, and by the way, as you just said, they're willing to invest themselves and they're willing to sign offtake agreements. So use it to see it as an opportunity.

23:31 I can't get past that the grid's 100 years old and we need we need to invest in it and nothing's going to drive it you're not going to go to the pucs and demand repairs pay more money because they can't afford.

23:41 It yeah eugene i don't know if the glass is half full or half empty but there's certainly liquid into it so thank you very much for coming on the show and You know, having this kind of freewheeling conversation because, you know, there's a lot of data, but what's important for us are the people. And you're one of them.

24:03 So thank you very much.

24:04 It's my pleasure. And I wouldn't be a real estate broker at heart if I didn't give you no answer to whether or not the glass was half full or not. Yes, there is water. So I appreciate you having me on.

24:17 Yeah, thanks, Eugene. Great having you.

24:19 So Gerard, I know you are always a bull, always a bull. And you know why?

24:25 Tell me.

24:27 Because you're an equity guy. You love stories. And me, I'm a fixed income guy. So I look at credit. I'm quoting Apollo here. The credit story in hyperscaler rests on the single consensus assumption that the operating cash flow triples from $ 600 billion to $ 2 trillion. If it does not happen, everything's going to go to junk bonds. And it started. SoftBank, when they borrow now, it's junk bonds.

24:58 We've seen the default, sorry, the force measure from Oracle on the Jupiter project in New Mexico. What that means now is if you look at the weakest part of the market, and I talk about it almost a year ago, the Blue Owl, the CoreWeaves, the Neoclouds, now they borrow at 8%, 9%, 10%, 11%. What the market says, not the equity jar, the credit market says, It's less risky to lend to Zambia and Angola than those new clouds.

25:32 That's not going to work.

25:33 Well, actually, I think it's probably worth talking about the equity side as well, because when you look at some of the valuations, obviously, Anthropic and OpenAI have not IPO'd at this point in time, but the numbers floated around are just... off the charts and you sort of wonder, are you ever going to really be able to get a return on an investment as an equity investor in those type of companies?

25:56 I can tell you that all the early investors of their best PR companies going non-stop, non-stop, non-stop on media and pushing it, pushing it, pushing it. There are very, very few people in the equity who's going to say anything. Okay, so you always have the big short, Michael Burry and so on. But the debt, it's the debt. You need to check the debt. If your business model, and we've seen the new cloud, is to borrow at 9% and then lend at 1%, look, you've got the competition from the open weights. You see the price of compute going down.

26:29 So I'm not saying AI is not a revolution. But what I'm saying is that the financial assumptions around the investment wave don't make any sense at the end of 2026.

26:42 Listen, I think the point you made there is a revolution. In a revolution, what actually happens is there's a huge amount of money made and there's also a huge amount of money lost. And I think this revolution is going to be no different.

26:54 But nevertheless, Gerard, the one who managed to build a data center, that's where Eugene was great. So whether it's the whole one or those type of things, when you manage to build a data center, a data center is very profitable right now.

27:09 Oh, absolutely.

27:10 Totally with Gerard.

27:11 But that's all development pipeline was just insane. The reality is great, but the promises, just don't believe them.

27:20 Good, my friend. Well, listen, as ever, good chatting. So I'm looking forward to next week.

27:25 Okay, John. Cheers.

27:28 Thank you for listening to Redefining Energy. Don't forget to rate the show and subscribe on Apple Podcasts, Spotify, or the platform of your choice.