← The HC Commodities Podcast
The HC Commodities Podcast
Data Centers aren't the bubble with Eugene McGrane
00:55:18
Today we return to the subject of data centers. Are we in the midst of a huge AI bubble? And building over capacity in data centers that are destined to end up as dusty sheds in the middle of nowhere with huge power supplies attached, or are data centers the rate limiting factor and the crucial midstream of the future industrial age? Here to discuss bubbles and data centers is Eugene McGrane, Executive Managing Director at Cushman Wakefield, servicing all manner of clients with respect to real estate needs for the power and data center sectors.
Loading summary
Transcript63 lines
- [00:00]
A
Foreign. Welcome to the HC Commodities Podcast, a podcast dedicated to the commodities sector and the people within it. I'm your host, Paul Chapman. This podcast is produced by HC Group, a global search firm dedicated to the commodities sector. Today we return to the subject of data centers. Are we in the midst of a huge AI bubble and building overcapacity in data centers that are destined to end up as dusty sheds in the middle of nowhere with huge power supplies attached? Or are data centers the rate limiting factor and the crucial midstream of the future industrial age? Here to discuss bubbles and data centers is Eugene McGrain, executive managing director at Cushman Wakefield, servicing all manner of clients with respect to real estate needs for the power and data center sectors. As always, you can really support the show by leaving a positive review on the platform you're listening on and I hope you enjoy the episode. Eugene, welcome back to the show.
- [01:04]
B
Well, thanks for having me. It's always a pleasure.
- [01:06]
A
A Couple of Bits of Housekeeping so firstly this is somewhat of a part two to the episode we just did with Eric Hans Koch, the senior partner at McKinsey on sort of the big CapEx build that's going on across not only the United States but also much of the western economy and actually globally of which data centers are a very big part. So that's where we're talking talking, we're talking data centers themselves in the context of are they a bubble, how challenging they are, what the demand is and how challenging they are to to implement and sort of some of the the economics and the reasons behind them. And I should also say that we are recording this June 9th and it goes out in two weeks time and so so in in prior to SpaceX's IPO. So we'll the bubble might have burst by the time this goes out. Just as a caveat to everyone.
- [01:56]
B
I I just, I can't wait to be featured in the, in the made for TV movie where I'm the guy claiming there is no bubble while the stock prices are are crashing. I thought that was a great part to that that wonderful movie.
- [02:13]
A
We should talk first and set the stage a little bit on both terms but as well as kind of the, the current state of affairs and sort of somewhat how we got there before we do a deep dive into data center demand, data center construction, data center future. And I think we're all even Sam Altman himself, even last year alluded to the fact that there's probably lots of exuberance in the market about AI and in fact you know Stock markets are topping all time records, particularly in the U.S. despite major economics internal turmoil as a result of the straight to full moves. And all this good stuff, you know, because all this terrible stuff I should say because of the amount of money that is flowing into AI and the perceived transformation, transformatory capabilities of AI in the modern economy and the productivity it would give and obviously that is the stock market is now also rewarding those companies that are putting more money into assets, I. E. Data centers. And the commentary is very much this is sort of the dot com bubble of 2001 and we're far from seeing the real productivity upticks in AI. We're very far from a singularity of AGI and in fact we're starting to see a tail off or at least less growth in the usage of AI tools overall and some reluctance to pay more for better latest models when most of us are just sort of, you know, it's, it's basically replaced Google for us in our day to day lives from a search function. I've done a lot of prefacing that but I'd love to get your sort of take on that group of statements.
- [03:58]
B
Sure. No, I appreciate it. I think first of all we should recognize that Google as it's currently constituted is an incredibly profitable business and has taken a primary place in all of our lives and our daily work and everything we're doing. It's funny that we talk about the.com bust as a place to start. I grew up in San Francisco in the Bay Area and watched all of that happen. I got out of school in 2002 and moved right back to San Francisco where the employment prospects were not positive at the time. So I lived through kind of both halves of the exuberance.com economy. And you know, it is reminiscent. I talk to a lot of my friends who are contemporaries and we often bring up the company Webvan which very famously had their logos on every cup holder at the new AT&T Stadium which at the time was the Pac Bell stadium in San Francisco for the Giants. So you know, we are at some place in this kind of technological revolution. I think Mr. Altman does kind of try to prick the bubble a bit often in order to kind of level some expectations. But I think the two things that we should be talking about today, and I'd like to separate them are, you know, the stock market valuations about, you know, which companies are now applying AI to their, you know, market reports and how they're going to do things and you know, how AI is going to transform their workforces. And I think we're kind of at the tail end of the layoffs where every company says we're going to lay off 6,000 engineers because we no longer need them. I think rightfully we're starting to come to the end of that because I think that this is my own personal opinion, but I think that the only way that these machines become a productive part of companies is by having people that understand them and understand what the companies are trying to do to work forward. So that let's just leave the AI transforming the economy and either putting us all out of work or making us all work five times harder. Aside where I think it's worth talking about and where I don't think that there's a bubble and I think that there's real information that backs us up and also real momentum is all of the companies that are supplying kind of the industrial backbone to this new level of compute. And whether you call it AI, which a lot of people are saying is a mis termed way to look at this, we're certainly not at agentic AI and we may never be at fully agentic AI or whatever the singularity is going to be. But we're certainly at large language models that are increasingly useful and productive. We've certainly got to find a way to integrate them into our schools so that we don't spend a bunch of time teaching kids how to type prompts where, you know, we kind of refocus on information and the gathering and digesting of information. But if we're going back to, you know, the sticks and bricks or the picks and shovels that has seen, I mean, let me just give you a baseline number. In 2025 we had 12 and a half gigawatts under construction in the world. That's a lot. You know, that was, that's 6x times what it was in 2020. And I'm probably fudging that number. But right now we have 32 gigawatts. 80% of that's in the United States. So 32 gigawatts is a lot more power than we ever looked at structurally for these projects in the past and has become kind of a baseline of a couple gigawatts if we're looking. Meta's got their Hyperion project in Louisiana and that's going to be a 5 to 6 gigawatt project that's under construction. It's going. These companies are building somewhat knowing without knowing the return. But when you look at a company like Meta making that kind of investment they're doing it because they know that in order to operate their business long term they're going to need to be technologically competitive and that competitiveness is going to require this level of infrastructure.
- [08:41]
A
Yeah, I agree with you on that. We're probably at the end of or we're in a maturation point of what AI can and will do. However, to remain competitive you will are going to be using significantly more compute than you ever have. Right. And that I think is, I think we're all seeing that in our businesses. We're not necessarily seeing huge upticks in productivity, but just to maintain a competitive edge. We're all of us using Farm. You know, the Microsoft bill is getting higher and coming more frequently than it ever has in terms of data usage. I think on the valuations. Absolutely. We've got these four or five companies all of whom, it's no coincidence they're running to try an IPO at this point. And again we'll see when this goes out. We might have a clearer picture of this and a lot of that. I, I think there's a dot com analogy there in that we are assuming that there's going to be a winner takes all. Kind of like we saw with Microsoft and Google and you know, Facebook matter at the time. Right. That doesn't, that doesn't to me seem to be necessarily. History doesn't have to repeat itself. Right. There's a chance these LLMs become quite commoditized and no one company has some kind of level of domination. They might have different lanes, you know, an anthropic doing the sort of enterprise wide stuff, chat GPT helping us Google stuff, whatever it might be, but there might be many other names. And so again that comes back to this idea for these valuations in the trillions. They haven't yet figured out how they're going to make sustainable moats for the most part they can't stop losing money and they don't really know what the future is. But this isn't an argument about those
- [10:18]
B
and that's never stopped them before. So let's, let's not use that as a barrier.
- [10:23]
A
Yeah, yeah, okay, okay, but, but, but the point being, the point being computer is expected to continue to grow. Our data usage is going to go up and if you own a data center today, tomorrow or next year, like the economics are wonderful. Right. Because it's a scarce resource. And this is why we're seeing Elon charge anthropic a billion dollars a month to use his mega data center that you Know, as we were just talking, he built before a lot of the current restrictions and controls and sanity was brought to the space. But that's our fundamental argument. It's kind of like it's building a refinery in 1940. You can, you know, there might be a lot of hype about the future of the motorcar, but if you own one, you're pretty good over the course of the next 50 years kind of thing. And even if there were about a stock market crash again, that would suck out a lot of these projects in the future. But currently, right now you're saying they're big projects, they're well funded and they're mature.
- [11:21]
B
So yeah, I mean, actually I'd make it. I'd make a separate argument. I like the oil argument, but there's a better one which is if you look at the dawn of containerization and what it did to shipping. Yeah. Was it better to own an entire railroad line or an entire port system? Or was it better to own a key port that had capacity to expand? And if you look, railroad's a great example. We had booms and busts as that technology got built out across the country. And yet if you look at what has remained incredibly profitable, there are certain switches, certain interconnections within the country and within all countries that rail has been built out or within shipping. There are ports that have continued to remain vital beyond belief. I mean, we're living through this with the straight of Hormuz. But you know, the Panama and the Suez Canal are perfect examples of as we became an interconnected interglobal world. Those shipping lanes and those rail lanes, the ones that drove significant traffic, were always incredibly powerful and beneficial to own. We're at that kind of scale. This isn't just building a refinery and being close to a certain resource. The power, the logistics in getting the type of power that is necessary for these systems and doing it in a way that is sustainable so that you not only have your own kind of control of what your power you're using, but also a guarantee of your price going forward is the moat. Now who becomes the user of the computer? That's a question that we don't know the answer to yet, but we do know that there's going to be compute usage. The great example that recently kind of really drove this home to me was a friend of mine at CDM Smith and he's a senior level engineer there. He just came over from a different firm and we were talking about types of projects that they do and the ability of their ability to chase projects. So if you're a water facility and you've got to do maintenance every year, you need an audit of the whole system. But you can't pay for an audit of the whole system because it costs $2 million and you don't have a budget for that. But with AI, these guys can spend three junior engineers can go and audit the whole system and they can come back and they can do $400,000 worth of work that the system needed and the engineers got paid on. And that's the kind of productivity gains that I think we all agree are going to come. Who controls that compute? I don't think we know the answer to, but I do know that that compute is coming. It used to be the government. The joke was the government is always listening. But how would they ever care or go through everything that's said? Well, they now have everything and they have the compute to measure everything and to collate everything and to understand everything. And I think that's the second leg of this that's really important is as we get into a more security conscious world and we have more unmanned AI instruments, the governments are going to spend a lot of money on making sure they have the compute. And that's a separate, that's a separate discussion. We can keep that or not.
- [15:14]
A
I guess we've sort of done our preface over the course of 15 minutes and you know, caveat emptor, all the rest of it. But I think like what's loud and clear is today owning that computer is very, very valuable. And you know, there's lots of projects going to go on and we're going to go on to that. Just there's one more final bit of setting the scene or at least to raise is of course in the U.S. news at the moment. You cannot move for stories about horror stories about data centers. And it's become, it's becoming a midterm issue. So one of those cultural touchstones where sort of the, the corrupt, whichever, you know, parties in control of that particular municipality. Municipality allowing data centers into small town rural America and electricity prices go up, taps, faucets run out dry and incredibly loud noises keep everyone awake all day. That's a very sort of relevant cultural meme. The moment in the U.S. can you just help us understand a little bit about that public backlash and sort of that losing the narrative quite early on and how that happened?
- [16:21]
B
It's hubris by the industry in seeing projects as they had seen them in the past. And part of that is just the fog of not Knowing what was coming. It's also rearing its head again that it has for both parties of America's inability to build infrastructure. And our inability to build infrastructure is baked in from a very good place. It comes from a reaction to freeways being built through cities. And there has been changes at the Supreme Court level that have made this probably less of an issue going forward. But this still goes through municipalities. I moved to the east coast right as Joe Biden was announcing his 32 gigawatt plan for offshore wind on the East Coast. And it really drove me to learn more about what was possible and what was able. And the big thing that struck me in that build out because a lot of companies invested millions and millions of dollars and billions of dollars in places. A lot of them were from the EU and they came over here with the incentives from the federal government and they tried to deal with that as a throughput through the whole system. And it just doesn't work here like that. You have to go from a county to whatever municipality to the state, to the feds at some point. But it is not just a straightforward proposition to do development. You have to have community buy in. What we've seen. And it's nice because it provides a place for companies like mine who have experience in leading developers through this type of process is the gold rush. Firms have fallen by the wayside. The people that have really known how to do this in the past are the ones that are moving forward with it. But that pushback was real. There's been very public pushback, rightfully so, from a number of well publicized large scale developments that have happened. I think it's important to note that the scale of these data centers should change dramatically over the last five years. And in that change, a lot of municipalities and a lot of people were caught off guard by both the size and the scope of those things. And you know, I mean, we can draw a lot of parallels between the attempts to build out American infrastructure in a lot of different places. You know, on one side there's, there's the parallels to the offshore wind where, you know, we want to put 32 gigawatts of wind up on the, on the East Coast. And you know, every municipality in which the interconnections had to go through basically demanded a toll or just outright said no. And this was power that they needed and certainly now really need. So the structures that we have in place in this country are very effective in stopping development if the people don't want it. So the cases that we're Seeing get so much attention will drive legislative change and have driven legislative change in almost all the districts that people are trying to build these things.
- [20:06]
A
I guess where you see it, where you see, where you see those successful projects today by these either independent developers or whatever it might be, you know, how is engineering like, what's different now to address these concerns that are just so sort of hyper, you know, in the, in the, in the public eye at the moment and as you say, are real as well, right? It's a very significant thing if you're going to have your environment impacted by an enormous data center.
- [20:33]
B
Well, it's really funny because the perfect example is, you know, the early days of hydraulic fracking, hydraulic fracking was seeing. I mean, they made a Matt Damon movie out of it, the State of California Bandit. And the technology's got so much better, so much more effective at extraction without, you know, continuing to drive those kind of environmental processes. Now, I'm not trying to compare the two things, but the, the type of cooling that's being used, liquid cooling, the impact to the grid, these have all basically become, bring your own power, the ability to do different mixes for the power. There's companies like Emerald AI that are managing the loads on the different data centers so that they're being used more effectively. And then the just basic engineering has gotten much more sophisticated as the time has gone by, both by, you know, a need to meet environmental standards and, you know, development standards, but also to meet development needs. I mean, there is, there are queues teed up for years to get the type of equipment that's necessary to run these things. And so, you know, by just what our general, general accepted engineering rules, we are seeing a lot of the type of engineering innovation that leads to lower noise, lower, you know, lower power consumption, you know, less. There's less water being used. There's, there's just all of these different things that chips are getting smaller. The, the use of GPUs versus CPUs is getting different. There, there's all of these different within the functions that some engineer will hear all of this and make fun of me, but the engineering is, is much, much more sophisticated. It's also driving, much more importantly, it's driving innovation in power generation that has been long made it. I mean, there's a 200 megawatt fusion plant going in into Dominion's grid in Virginia, and that's not good. There's not going to be any pollution
- [23:08]
A
coming out of a fusion plant.
- [23:09]
B
Fusion. Fusion. 2030 Commonwealth fusion systems.
- [23:18]
A
Well, I will let that. Hang there for a moment. I missed that one. Okay.
- [23:26]
B
I mean, yeah, that doesn't happen unless somebody's willing to stroke a couple, to a couple billion dollar check saying, hey, this is important to us and we need the power.
- [23:35]
A
But fusion, not fission.
- [23:37]
B
No, fusion.
- [23:38]
A
Fusion as in, you know, with a
- [23:40]
B
Toka, with a Tokamak. And you know, it's a whole thing. It's a deployable commercial fusion company. And they're building. The Spark is almost complete in Massachusetts and they have a PPA with Dominion for 200 megawatts. And the, the delivery is supposed to be 2030.
- [24:00]
A
Okay, well, I missed that. That's, that's on me. Okay, Just, just from a, from you, from, you know, the real estate sort of the, the development angle, which, if any, are the critical elements. Is it, is it the land? Because there's only, so, you know, it's kind of the Walmart thesis. There's only so many kind of like locations in North America that these things just really work in because of confl. Laws, resources, space, temperature, etc. Is it land? Is it simply access to power? Is it actually that's ubiquitous. It's based basically who's on, who's in the front of the line to get the next chips. When you're talking to these developers, hyperscalers kind of, where is the bottlenecks? What's the heartache?
- [24:48]
B
I mean, all of those things that are absolutely heartaches. I mean, power is the moat. But how you get access to the power, how much of the power is available? And that's really been sussed out in the last two or three years because there's been significant resources put into figuring out. I mean, what happened was when AI became a thing and when data centers went to hyperscalers, we went from looking for, you know, 50 megawatt parcels with, you know, interconnections that were, you know, 18 months out to 500 megawatts to a gigawatt. And that's really where we are. You got to kind of have a baseline of a gigawatt. And then you start getting into all the other pieces and the pushback is from, you know, these early pieces getting decided before people really had a chance to react to them. So when we're guiding people, you know, if they're developers and they have a choice of sites and we're trying to, we're trying to narrow in on sites, we're very specific about what judicial jurisdiction it's in because there's, there's a judicial, just jurisdiction in Texas that's notorious for being a patent troll jurisdiction. You know, what's the regulatory agency look like? Is it something that can stand behind what it says? I mean, we're watching the thing in Utah where there was an end to run kind of done and people don't feel like an unelected board of a couple people should be able to make a decision that they're forced to make. So when you go, you want to have a place that is, you know, legislatively open to and interested in having these sites come along. And you know, there, there is righteous pushback on what these things look like. There's also competition because, you know, we're in the midst of a reshoring of a lot of things and one of them is chip manufacturing. And so if we see chip manufacturers, they have the same power requirements, if not more intense and the same landing requirements and they've got to be close to population because they need people to work there. So you know, we, we guide them through all of that.
- [27:22]
A
Yeah, I guess it's interesting because there's somewhat of a parasitic nature of these things. Right. Because. Well, answer this for me. You know, the benefits accrue to some faraway, you know, plutocrat and the power and the water is sucked out locally. But these aren't great employers. Right. It's not like having a chip manufacturer or refinery turning up in terms of lots of skilled employers. These things sort of hum away and you know, but they aren't creating white and blue collar jobs at scale after they've been built. Is that a fair statement as well compared to other competition? I mean, how do you, what do they give to that local municipality that, you know, over and above whatever they can extract in terms of taxes?
- [28:07]
B
Well, I mean the tax revenue is not, is, is not inconsequential, especially given that they're not typically going into what was a tax rich development site. You know, typically these are adjacent to some sort of power infrastructure which is not usually in a high population zone. Now there are places where that does happen. The argument I would make is that the type of people that have to work in these data centers and there is a huge backlog of need for the people that go in and work at these data centers to plug and unplug and pipe and do all of the different manual, skilled, specific labor, they don't have enough people to do that. Now people are being master electricians and journeyman electricians are being shuttled around the country from site to site at huge payments. So you know, there is the basic, these Things are not going to just run on their own. They're not going to run on their own with the type of equipment that's inside of them because that equipment is very expensive and can't just be left in a static warehouse. So there are jobs who is building
- [29:21]
A
these things as well. I'm sort of interested in the, you know, are they built, already contracted to a particular and this is consequential for our sort of AI bubble bursting impact. But are they built, already contracted the compute sort of contracted to a particular company? Are they built by those companies or are they built in largely by independent developers who are then, you know, shopping them out to the best offer kind of thing?
- [29:46]
B
The competition is so strong that there is not many that are being built by independent developers and then, you know, gone to lease and then put into a trade where, you know, people can go and buy those things. You know, we did just see Berkshire put in $80 billion to Google to help them build out their, their, their, their infrastructure. And that says more about Google saying, hey, you know, we want a partner that makes the case that what we're building is not just frivolous infrastructure. We're building this with somebody who believe term and as a value proposition, value add proposition. They're being built by all sorts of different people who have expertise along the chain. You know, some are being built like Meta is building their own, AWS is building a lot of their own Oracle's Partic portable partnered in that Stargate deal. I mean there, there, there are, there are various phases of this. Some people are getting them entitled, getting the power brought to the site, doing the, you know, the work to get there independently. There was just a deal that went out for a capital call in Texas where they needed $25 million to retain their space in the ERCOT Q which is a, you know, ERCOT just went to $5,000amegawatt as a, as a hold number to keep your place in the queue for interconnection and then you have to use the power that you sign up for, for five years before you get your deposit back.
- [31:33]
A
Yeah, I mean it basically seems like here, it's another, it's sort of round two of shale, right? The shale revolution is now being coupled to the compute revolution and Texas once again, through happenstance of, of, of geology and, and, and business friendliness and all the rest of it is kind of ground zero for some of this stuff. And you know, ERCOT and myself will no doubt pay the price for that sort of in the, in the, in the short to medium term, but I mean it. And you, you outline the scale of this and it's kind of fascinating that these, you're, you're sort of delineating data centers by their power consumption. And you know, one gigawatt just to put a pin in it, is obviously a decent sized nuke. Right. I mean these things, these things aren't exactly sort of economical in that sense.
- [32:24]
B
They're not just lying around.
- [32:26]
A
Yeah.
- [32:27]
B
You know, if I'm looking at this with my positive hat on, which I always keep on top of my negative hat, my positive hat is that we have over the last 10 years seen that our ability, and this is a thing that's true across all the big economies of the world, our ability to run our own economies through our infrastructure, our access to critical metals and critical minerals, our access to manufacturing expertise has all been eroded very significantly because our focus has not been on doing that. And in a capitalist system, the best way to get something to work is to have somebody that's willing to pay for it. And this is a time when we have these companies that have made tremendous amounts of money, just, just staggering amounts of money. You know, Soft Forza sales is, is a great, is a great business because you know, your, your cost of good
- [33:44]
A
is zero asset light, infinitely scalable. They've got oodles of cash. This is your sort of Steve Ballmer anecdote about kind of the, the pocket change to buy, buy a buyer, a basketball team, the LA Clippers, you know, and, and it was no big deal. Right. And obviously he's gone on to make even more money out of, you know, but it's the amount of money that these companies have is fueling this, the, the, the other, the other sort of just one more positive bit. And then I kind of want to stress test a couple of things with you on kind of why it's not necessarily be so great. But one, another analogy that I throw around is, is that sort of, there's a, there's a fiber optic story here which is sort of, you know, back in, back in the same time as the dot com bubble, everyone was building Enron in particular and you know, my house is, is dated to a remodel at that stage. And the amount of coaxial cable that runs, you know, you could, it must have cost a fortune 30 years ago that's now redundant but essentially lots of fiber optic stuff was built. And then the technology changed to mean that you just didn't need, you know, you get so much more data through one fiber One more argument that why this is a positive story. And again, I'm kind of using the idea of the data centers that are going to get produced in the next five years, right. Is that normally the markets react to this kind of thing with massive over capacity. But this, this isn't building a container, right? This isn't building a railroad switch or the railroads in the 1860s in the US 1870s, 1880s, 1890s and a big collapse that happened. These things because of the scale of them and the uniqueness of the facilities in terms of like location, power demand, the regulatory impact, all this stuff. Is there an argument that also it's quite hard to, you know, it's unlikely that we're going to have massive overcapacity because of the difficulty of actually standing one of these things up. Is that a fair statement, do you think?
- [35:43]
B
Yeah, no. And that's kind of my base case is, you know, three years ago when ChatGPT became the hottest thing and everybody wanted to put up 100 megawatt data center, the wheat has been separated from the chaff, the pushback from municipalities has been real, it's cut into projects. And so the focus has really become on large scale projects that can be delivered. And so you know what, and you can, you can link in the show. Link links with, you know, Cushman just put out a big international data center report. And, and this is pretty much what we're saying is, you know, vacancy rates are not nil, but they're, they're pretty darn close. And pre leasing on construction is, projects that are in construction is very healthy to almost complete just with very little kind of slack in the line. So it would be hard to say that we're going to over build. There might be some of it. But I think that given the scale of where compute needs are today, it would be hard to say that five years into this cycle from now we're looking back and we're like, oh man. Yeah, that was, they were, they were building. You know there's an old saying, an office building when, when you bit. When you build your office building so far out that you can see cows, you've done, you've done built out too far. Yeah, that was, that was, that was the guy from Equitable Life. But I don't think we're going to be seeing cows from data centers in the sense that we're going to not have gone so far out in order to build these things that will have built useless infrastructure. I think that there's going to be Great equity in the physical infrastructure that's being built. It does go with the fiber argument, is that there was an overbuilding fiber, but now we're using that fiber.
- [38:02]
A
And I, and I, and I kind of, I mean again, this is sort of separating out the evaluation of anthropic or SpaceX to the underlying, you know, bet on a data center and the direction of compute and the, the scarcity of these. And there's a, the one argument we haven't really touched on. But I think, you know, you and I agree and probably every listener agrees in a world of, of security over efficiency, which we're definitely in, owning your own data centers is, is high on the list for every national, for every country. Right. So we're also probably in a world where we're not in a just in time world and we're not on a, oh, it's cool. We'll use, you know, again, the cloud is just someone else's computer. So we will just use that idea that sovereignty of data is also to play here. If we were to say a couple of reasons why this might be shaky. And again, I'm trying to sort of pass the idea of we have a 30% stock market crash all feeling, you know, in September because everything's caught with us. That'll certainly put the wind up a lot of stuff. But I imagine that if you own a data center this time next year in that scenario, you're still feeling pretty good and you're probably still going ahead and building, you know, the one that's, you know, at FID or whatever it is, because it still makes sense. Right? Because we're still seeing those directional trends. I think a kind of keen to stress test with you how this narrative might break down. One is definitely technology and there's sort of, you know, we're like, like I'm sort of astonished to hear about fusion actually being bought and put in somewhere, you know, which to me is heralds the fact that UFOs are going to land next week. The other is of on that line is quantum, quantum computing. It seems even in a world of quantum computing that data is, you know, it's still going to require data centers. But there is that kind of like actually someone comes up with a chip that uses vastly less, less power, has the same output, vastly less, makes far less heat as a result, all this kind of stuff and suddenly. So there's the obsolescence argument. Right. Can you tackle that for us?
- [40:09]
B
Yeah, I, I, yeah, I mean, I think even in a world of fusion and, and and aliens, you know, there'll still be a transitionary period. This, this gets into more of a philosophical answer. My, my. We are just now coming to terms with the fact that we've digitized every piece of human information ever and are now all have been generating our entire. I mean like look at our kids, right? I have pictures of, I have thousands of pictures on my phone. Hundreds of thousands of pictures on my phone from my, my child's, you know, time. All my children growing up. We haven't stopped collecting more data on the human experience. There is a deep desire, and I am cynical about it, to monetize that collection. So there's going to be a business interest in doing it. I also think that there's a deep desire to improve the human condition that's possible now in a way that, you know, the people that are farsighted during this time of technological change and who see the value of human intelligence and creativity and you know, just basic humanity are going to be the winners. The people that take this as cold hard information and it'll spit me out the right answer are going to be the losers. So let's just say that there's a 30% drop in the stock market because the IPOs unbalanced the stock market and bring forth economic Armageddon in a lesser form. I'm betting that's probably a lower interest rate environment. What is nice about this time through is these things are being built in a reasonably high historical.
- [42:21]
A
Well in an average cost of capital. Right. They're standing on the economics, not on the unicorn status, you know, which we had Edward Chancellor on talking about. Right. Yeah, they actually, they make sense in a standard economic, you know, in, yeah,
- [42:36]
B
NPV people, people are buying and making money and selling at market in order to make this happen. It is certainly lining a lot of Chisholm Huang's pockets, but he seems to be very smart about deploying that. But you know, this is not a, this is not the type of environment it took to get, to get solar panels and wind farms to work initially. I'd argue that they now work on their own. But you know, you're, you're the, you're the energy guy. You know, I think we had to, we had to bend the cost curve enough to get there, but we're definitely there. This is, we're, we're operating at, you know, very expensive capital for, for speculative
- [43:24]
A
projects which, which the other bit is kind of like. And again this is kind of the, not everyone's going to do amazingly out of this. There's always going to be the one that builds a data center that can sees the cows, you know, as a really good, good phrase. You also talk about sort of the capital crap and that actually this is a bit of a sea change for many of these companies. And you know, a lot of these assets, these refinery, all this stuff, I mean they're not, they're not wonderful things to own over the long term. If you know, if you're sort of. Well, you tell us, tell us about your, you know, how that plays into this.
- [44:00]
B
So I mean the one theory that I've had that, that makes that warms my, my, my dark, my dark times cockles. Warren Buffett bought Berkshire Hathaway and it was a textile mill of kind of great renown. But what they really had was a good balance sheet. And he used the balance sheet to build this financial empire that is today. What he realized was and this was a textile mill up the road from me in New Bedford. And every time they had to make a capital improvement to build faster weavers or you know, put in a new loom or do some sort of automation, it didn't increase the margin, it didn't, it did increase the productivity but it didn't increase the profit. So it was just running on a treadmill just to stay current. And AI the way it is being adopted has that same in the term that he uses, capital investment trap. And so the question is, are these companies, the big ones that are building these bits of infra, these huge monuments of infrastructure doing this just to keep up and do they generate higher, higher returns by doing that or are they just doing it to stay current? And I mean Google is the one that's easiest to point to and just say, well you know, you control the world through search for 20 plus years and it was the best money printer that was ever created. Now you're building Gemini. Does that really do more for you than search ever did or are you just paying to stay current? And it's just that, is that just now your cost of doing business?
- [46:03]
A
And it clearly is right. Like we all now use a chat of various forms I, you know, as a data synthesizer, a search function with synthesis, you know, and so they can't not afford to do it. But again it's destroying their, the old advantage they had in search, which kind of goes back as well to this. And this will be the fascinating thing I think to play out at the moment because of the massive constraint in let's say the refining bit, right? The, the midstream, everyone's having to build their own data centers or lock up independently developed data centers, you know, the most efficient model you would assume. And this is why you see the oil industry going through sort of periods of integration and d. D integration, whatever the, whatever the right phrase is there
- [46:51]
B
it said that's disintegration.
- [46:53]
A
Yeah. Is, you know, is. Or unbundling, you know, is precisely those sort of different forces at play. But generally speaking, you know, an upstream producer is very good at producing oil. A refiner is very good at refining and a. And a retailer is very good at retailing. And if you have a great trading platform on top of all of those and you can kind of get the three plus, you know, one plus one plus one equals four of those bits. Bits. But typically you can't, you can see, you know, be just interesting to see does Google really want to be in the business of owning lots of hardware and all of the headaches that that is. And the answer is probably no at the moment. They have to, otherwise they'd be locked out. So how this matures in terms of, you know, I quite like to be a data center owner and have lots of different, you know, companies compete to use my compute on a second by second basis. Right. To optimize my returns. So how that all plays out I think is going to be quite interesting.
- [47:49]
B
I agree. I mean, I think it's fascinating to watch the people that have won every one of these since Steve Jobs came back has been Apple. And Apple is just like, yeah, we'll see what happens. And right now they're going to use Gemini and rebuild Siri as a specific AI using Gemini as an overlay.
- [48:14]
A
But ironically, Apple is very much a. We want to own the hardware and they're the one company that owns the hardware and the software because they saw that as the key control over the user experience. Right.
- [48:28]
B
And they were right.
- [48:29]
A
And they were completely right. Yeah, I find it interesting and I think again, actually if you want to, I think Microsoft again I somewhat using it the same way. Right. They have obviously been investing, but Microsoft is like actually where it's part of our suite of products. You know, obviously having Copilot there as opposed to it being a, you know, again we come back full circle to these valuations that are coming up. Ultimately many of them are sort of for the most part single tool companies at the moment. They've got to figure out how, what, how their, their business model gets them across an entire range of products that either locks you in as a, as a, you know, as a SaaS or whatever it might be. But it, yeah, yes, the.
- [49:12]
B
No, it's, I, it's, it's, it's fascinating. Right, because we don't know who's going to.
- [49:15]
A
Yeah. Or if indeed there's going to be. It doesn't have to be the same, you know, game plan. Right. There might not be like again, there's lots of different, you know, okay. There's lots of consolidation that has happened as a result of other forces. But you know, there are, you know, there's not just one pencil maker in the United States. Right. There's not just one. Unlike Google where they basically dominate everything with a couple of also rans. It doesn't necessarily, you know, there might be a world where the LLM is just a utility. As Sam Altman said, you are not willing to, you know, you and I don't need on a daily basis the latest model. We're not willing to pay for the latest model. And you know, yeah, there's a bunch of, you know, pirated ones. You can get open source from China if you really want to lose all your data. But you know, like, I don't know, past is not necessarily history, is not necessarily future. In this case, it's uncharted waters. But there are some real parallels to the energy industry in terms of vertical integration. But there's also obviously very consequential in terms of, you know, you, you've got to, you gotta, if you're making a gigawatt facility, a power plant for someone, you've got to, you've got to hope that they're still going to be around in three years. Otherwise you can have a lot of stranded power out in the middle of nowhere.
- [50:25]
B
Yeah, that's, that's the crux of it. Right. I mean, you really got to believe. But you know, the answer is the companies that are, the answer to that specific challenge, right, is the companies that are doing this right now are, you know, building their own power infrastructure. And to build and finance power infrastructure of that size, you got to have at least a 15 year commit, right? At least a 15 year PPA. And so, you know, you're betting on the balance books of companies that are some of the biggest in the world, you know, saying that they're going to be around for 15 years. And as a real estate guy, I take that bet all day long. You know, that's something, that's something you could commoditize and sell pretty happily, you know, without, without having any trouble at all.
- [51:16]
A
You can imagine a scenario where the, the current People with data centers today or data centers in three years would be quite happy for the bubble to, to burst in inverted commas, you know, to take some froth out of it. Yeah. And just again, in 10 years time, you're, you're once again, you'll be making money hand over fist because the data centers weren't built in, you know, 2028 or 2029, or weren't starting development in those years, you know, and you have another gap. Right. It's a classic capital cycle.
- [51:43]
B
A very smart guy. I was involved in a project outside of Nevada, outside of Reno, and, you know, there's a ton of water there. We, we didn't have all that much access to power, but he'd been involved in a bunch of infrastructure deals in California and kind of across the country. And, you know, we kind of looked at it. A bunch of different, different groups looked at it. He said, you know what, I've worked with these software guys, I've worked with these infrastructure guys. I can tell you the software guys are used to just saying, we want it tomorrow and we want it today and getting their way, because it's a much more fungible business. He says the infrastructure just doesn't work that way. He says, I'm going to take my time. I'm going to get all of my decks in a row. I'm going to have all the power I need here. I'm going to have all the infrastructure I need. It's probably going to take me two or three years more. And I'm willing to bet that I will outlast the other people. And so far, he's, he's proven pretty prophetic to me.
- [52:42]
A
Yeah. Well, I'll leave you with one thing. Given it is it's summer and, you know, and you and I are old friends, which is just talking about the amount of data that we're all accumulating. Well, there's an ongoing theory at the moment that they can't find dark matter.
- [52:57]
B
Right.
- [52:57]
A
No one's actually been able to find it, despite what sort of all of the, you know, it's sort of the God of the gaps in many physics models. Well, there's a theory at the moment that actually dark matter is simply data, and that data itself has a weight. And that if we had sufficiently sharp enough measuring tools, we could weigh the difference between a full hard drive and an empty one. And that actually, if we carry on creating the data at the rate we are at the moment, you know, we start to have some profound impacts on the Earth's spin and all the rest of it. So yeah, one, one.
- [53:30]
B
So when, so when are you and I going to talk? Noetics? Because that's.
- [53:34]
A
There we go. Yeah, the Akashic records. I'm on it.
- [53:38]
B
I mean, I mean, I mean thought has weight, you know, the human soul has weight and a hard drive. I'm ready. I'm ready. I don't want, I don't want you to cut this part off of the podcast because I want people to know how sort of the deep down there is there's a strain how woo woo
- [53:53]
A
we get, you know, when the mic's off. But anyway, but on on very real world topics of thinking about building a data center, building the energy supply data centers or getting a market take on what exactly is going on and I know there's loads of questions I didn't ask that people are probably screaming at me that actually sort of very basic things people want to know but they should reach out to you. Eugene. I'll put your contact details in the show notes and also alongside the Cushman Wakefield report that you mentioned and look forward to having you back on in a year. And we'll see where the story is and let's you know, either fingers crossed either way which way you lean on these things. Whether by the time this episode comes out the the SpaceX IPO has has rocketed off or plunged back down to earth to use a corny pan.
- [54:45]
B
It's either trees never grow, don't, don't grow all the way to the sky or this is different this time. So we'll see.
- [54:52]
A
All right.
- [54:54]
B
Hey, thank you very much.
- [55:01]
A
Thank you for listening. To find out more about HC Group, our global offices and our expertise in search within the commodities sector, please visit ww.hcgroup.global.