Mastering the Infrastructure Cycle w/ Stonepeak’s Michael Dorrell

[Music] If it's an infrastructure, essential service, you can be very confident in 10 years time. They've got more earnings than they were with more than they are to die. But the path to get there is not likely to be a straight line. That's Mike Dorrell, the CEO and co-founder of Stone Peak, and a pioneer in the world of infrastructure investing. Mike is seeing just about every cycle of this asset class, from the classic port and road deals of the late 90s to the crash of '08. And at a time when many investors are wondering how to get ahead of the AI boom, or worrying if it could all go bust, Mike offers a strikingly level-headed perspective. As long as I've not levied it too highly, as long as I can survive I hit to my earnings for a year or two, give it a year or two, and someone else will want that data set in his face. Today on the show, I'll ask Mike to share his hard-won lessons from previous cycles of infrastructure investing, and how he's navigating the AI cycle today. We'll also explore, to what extent, the rapid build-out of digital infrastructure resembles the dot-com bubble, and more importantly, where that historical analogy breaks down. I'm Hugh McArthur, chairman of Baines Global Private Equity Practice, and this is Dry Punter. Mike, I'm really excited about today's show. Thanks very much for being on. Hi, Hugh. Thanks for having me. A real pleasure. You know, I think infrastructure investing is fascinating, and I've worked around it and followed it for well over 20 years, but you really kind of got in on the ground level and have innovated a lot here. I'd like to start just by asking you to walk us through your personal journey and what drew you to infrastructure investing in the first place. Sure. So I left college. I had no desire or thought of getting into infrastructure at the time. Candley thought it didn't even cross my mind. I went to a firm called McCwory, which was unheard of here in the US when I joined in '98. McCwory had a group that was called Project Instructured Finance. They did a bunch of cross-border leasing transactions back at that time, and they're of a small side show, but it was full of these hard-driving, very entrepreneurial, quite irreverent people that got infrastructure investing going at McCwory. And a little bit of the background just quickly there is that the US leads Australia in most aspects of finance, but we led the US in terms of state governments going bankrupt. So our state government started to go bankrupt in the early to mid-90s, and as a result of that, started to sell off traditionally government-owned assets, which were primarily tall roads and airports. And so this little group within McCwory decided they were going to try and buy these assets. These assets were very, very unique. They're natural monopolies, they're essential services, so very robust demand. You've got to use the airport, you've got to use tall roads. And so this group realized straight away that these assets were unique and really interesting, particularly as a match for long-term insurance and pension fund liabilities. So this group went and IPO'd a cash box in Australia for the sole purpose of bidding on airports and tall roads. And that little team pretty quickly became the biggest, most successful, most profitable group in McCwory by far. I remember in 2000s, Mike, that McCwory was essentially saying the same thing as infrastructure investing. It was the most famous synonymous brand in the world out of Australia. And the question that pops to my mind immediately is why you were doing all that and McCwory was growing so much. How did you wound up finding New York? So I was a young whipper-stepper. I just wanted to get out of Australia. And frankly, McCwory had thought the US was the toughest market to crack for infrastructure, simply because it was the only market in the world with the municipal bond industry. And so relative to Canada and UK and Europe and Australia, the US looked like it was not an obvious place to patrol right at airport privatizations. So as a consequence of that, initially McCwory didn't take it all that seriously. And they sent over a bunch of very young people, almost as total option value to that market. So if you were going to come to New York, I think McCwory would have a few to go and do that. So I came because of the city. And pretty quickly we realized the US was quite a different market to the others, whereas we couldn't do toll roads and airports so much in the US. The US's massive world of already privatized infrastructure that had not here to have been categorized as its own private asset class. So it's got the biggest market for traditional oil and gas infrastructure, just pipelines, processing plants, etc. It's fine away the biggest market for power. It's the biggest market for digital in the world by far. And what was so interesting about the US over and above its size was that because these markets were in private hands, there was a lot more ability to be dynamic and inventive about the way you sourced transactions. So if it's a airport or a toll road that you're trying to buy from a government, governments run these broad based options where everybody's got the same information and it's literally a price battle. And if I'm bidding on the Chicago Skyway toll road that I think, hey, I could get to a much higher price if the city changed the concession agreement in X, Y and Z fashion, then the city would adopt that change. But then they let everybody be aware of that change. And so everybody would get the benefit of my smart thinking. You could get no angle whatsoever on these assets. It was the pure cost of capital shootout. Whereas, if I'm looking at assets owned by private companies, well, it's just a question of what the other company's willing to do. Are they willing to deal with me exclusively? Are they willing to kind of play the structure that I've come up with that's unique? Are they willing to value the relationship I've built and therefore have more trust? And the answer to it a lot of times is yes. And so interestingly over time, this market in the US that was so different to elsewhere in the world that didn't get so much attention, if you look at infrastructure today, the rest of the world much more reflects what's going on in the US than the reverse. Because all the Torres and Airport stuff, that's now so low returning and considered the core of the coin infrastructure that a bunch of pension funds and insurance companies will go on by that directly at six, seven, eight percent leaven returns. It's not so interesting to the GP market. And so what started off in the US is now very, very emblematic of what's going on or around the globe in infrastructure. I want to talk about how that evolves Mike, because you started Stone Peak after the GFC, which I want to ask you about because when we went through the GFC infrastructure was kind of a dirty word for a lot of LPs. So it wouldn't occur to me that the most obvious thing to do soon after the great recession just a few years later would be to go out and open your own infrastructure investing firm like Stone Peak. So what was your your founding vision for Stone Peak and how did your strategy start and evolve over the years coming out of that shock that we went through in 29, 2010? There's a famous expression amongst investors that Warren Buffett says, which is that a string of great returns followed by zero is a zero. Okay, so I do think that these assets that were bought before the global financial crisis that suffered through the global financial crisis ports are a classic example. But toll roads suffered, airports suffered. Same in COVID, airports suffered like massive turn downs in COVID. You've got to set your financing structures up to get through those because if it's a good toll rate or a good airport, whatever happens to be if it's an infrastructure, essential service, you can be very confident in 10 years time, they've got more earnings and they're worth more than they are today. It may have a dip here and there. And so you've got to have a financing structure that survives that dip. And I think that the history of infrastructure is quite different. In fact, just the exact opposite of the history of private equity. So all I mean by that is private equity originated in the US with a bunch of entrepreneurial teams. Contrast that to how infrastructure started, which is literally out of McCory. And McCory was followed by other Australian institutions and then it was followed by Canadian and European institutions. And finally, it was followed by US institutions. So those institutions were mostly investment banks. And by the end of '07, all those Wall Street firms had all started infrastructure funds. And they were coupled together, essentially unknown teams with no investment track record were raising $234 or $5 billion. And you know, '07, which is maybe the most buoyant market we've had, probably probably similar to today. You know, like a lot of money got put out the door very, very quickly by an experienced teams. And so when the global financial crisis hit a year later, there was a big washout in the infrastructure market. And it was reborn after the global financial crisis. It was reborn in a different way. If you must worry the model of private equity, then the investment banking model. And if you talk to my McCory alum, all of us, myself included, went through a real learning experience as to what's the difference between investment banking and private equity investment banking is all about deal velocity. And how do I get this deal done? You're puts on the accelerator versus the private equity model, which is all about, my strong biases not to do the deal. It's about deal quality, not deal volume. It's a very different mindset. a very different culture. And so I just think it took a little off of the infrastructure to appreciate that. Well, let's shift gears then, Mike, and talk a little bit about the AI buildout or what you reference as digital infrastructure, which of course lots of people are talking about. Walk me through the infrastructure that's needed to make the technology work. We hear a lot about data centers and we hear about energy, but how is it working to end? What does that look like? So it is data centers for sure, and data centers spend has gone up by a factor of 10X, something like that pre-AI, and it was already very big to service cloud. And then of course, data centers draw on a lot of power. And so not only do you need the power facilities, which in the US, these days is primarily gas-fired generation. You then of course need the fuel to supply those power generation units. And so it's not purely gas, you can also have green energy, et cetera, but the majority would be gas, powered generation, driving these new data centers. And then I'd probably draw a four quadrant matrix. And the way I've put that matrix is I've either got as a client, so I own the data center, and I'm getting customers into that data center. My customers can either be the very well capitalized big tech companies. So the obvious ones, the Matter, Google, AWS, which obviously very strong counterparts on the one hand, or they can be the newer, call it cash flow-negative startups, which would be things like open AI or anthropic or XAI. And that obviously, a bet on a open AI as my customer is a very different bet to one where Google is my customer. So that's one axis of that quadrant. The other one would be, where is my data center located? So if my data center is located in and around, say Northern Virginia, which is, I'll use that because it's the biggest data center market in the world. If I have my data center located in Northern Virginia, I feel pretty good about the long-term demand for that data center. Whereas if I like my data center in a middle of a farm somewhere because we haven't had cheap power, and I'm training AI, because if I'm training AI, I don't need to be near a population center. That's a very different bet. So depending on where you sit in those four quadrants, take the most risky, but you're going to a data center where open AI is your customer and you're sitting in a farm in the middle of nowhere. That's a very different bet to Google as my customer. I have a 20-year contract, and I'm sitting in Northern Virginia. And so that's how I categorize, if you like, the different data center bets in and around AI. And you mentioned, Mike, that the investment in data centers is 10X, what it was prior to AI. How do your positions don't peak to keep up with that level of demand? It differs by market. So the study point is what price, what price do I need to pay to enter a market, or just inversely what return am I getting to enter that market? And so what we found is that in Asia, we could buy existing data centers and we can build new data centers at pretty interesting returns and pretty interesting cap rates. And over time, the multiple of earnings at which Asian data centers are priced went up a lot. And so today, I'm very happy building new data centers in Asia. I'm not so happy buying existing data centers in Asia. In the US, I've never been happy buying existing data centers because the cap rates are so low or the earnings multiples are so high. Now, anyone who bought existing data centers, pre-AI, did very well. Like, we got that wrong. We didn't see AI, it's far enough to forecast that. And so anyone earning a data center, or buying a data center pre-AI, I did very well on that data center. But I would say that even the build cap rates even building a new data center pre-AI look pretty skinny in the US. Whereas today, building a new data center looks really interesting in the US. So summarizing all of that, we are building a lot of data center capacity at the moment. In fact, I'd say everything we're doing in data centers the last couple of years is finding land, it's connecting power to that land and getting good customer contracts on the back of that. And so in Asia, you can do that at 12, 13, 14% cap rates. In the US, you can do that at 9, 10% cap rates. With good 15, 20 year contracts and annual step ups on those contracts, they're very interesting returns. But on the other hand, if you want to buy an existing data center platform, you're probably looking at two to three percent cap rates, which is tougher. Like, you need a lot to go right to make that work. So we'd rather build, which takes longer to deploy the capital in its trickier. Just because to us, it seems like you return ought to be higher and your risk ought to be lower. So that's how we thought about it. Seems like there's a lot of momentum here and it feels like you have a lot of confidence that there's going to be sustained momentum in the data center and infrastructure space for digital giving the growth in AI. And it strikes me that people are worried that we're in some sort of a bubble and there may be elements of AI that are in a bubble. I'm not sure. But what strikes me as different between what's going on now with the digital infrastructure build out and what happened with, say, the broadband fiber build out 25 years ago is that that was largely unutilized when a dark fiber was put into the ground, whereas it seems like every single data center, as soon as it switched on, is 100 percent utilized. And can you just comment on how you see this period of time being different and what gives you the confidence that the build is going to continue to be strong for years into the future? The other thing I'd say about fiber is that fiber capacity is almost limitless. So it's very easy to increase the capacity of fiber just with a few software changes, whereas for a data center, you can't just physically increase the data center space, of course. But we come back to that little matrix I mentioned earlier. So firstly, if I can build a data center in and around Phoenix or Northern Virginia or Chicago, but one of these big markets, I have a long-term contract. And by long term, 15, 20 years from Google or Apple or Microsoft or one of these folks, I'll do that all day long. Like, I don't really care if there's an internet bubble or an AI bubble or not. I'm really betting on Google surviving. And I've even got a backup, like Google's going to survive. But you'll play further. I've got a backup in that I'm very confident that between cloud and AI demand for data center capacity is going to grow over time. Now, can it go through a flip? Yeah, of course. It grows at a big rate. So I'm not even sure the growth goes negative. But even if it does, if I've got that long-term contract, I'm good. But I'm also in these markets where even if Google went away, which it's not going to go away. But even if Google went away, someone's going to need this data center at some point in the future because I'm in these markets with going demand. So that's the safest bet I can make. It doesn't always work out that way. Like, you can't always get all the stars alive. But I'm not so worried about internet bubble. On the other hand, if I build a data center in those same markets-- so again, I'm talking the big markets-- LA, Chicago, or at least big markets-- if I build one for anthropic, say-- and I'm just picking anthropic as one of the startups. That's a different bet. That is, I think, got some exposure to a possible AI bubble because we all know that you can have the best business plan, the best business idea, the best business in the world. But if you rely on outside capital to fund that business, capital markets have a habit of shutting down when you least want them to shut down. And so you look at the capital spend for those big AI startup businesses that spend, didn't they, tens of billions a year, about so funding from debt or from equity investors. So if that shuts down, I think that those startup companies are going to go through a pretty tough time for a period. And so I would be concerned that you are exposed to a turnaround of the exuberance if I'm investing in that sort of data center. Now, just to run that out a little bit, I want to say, anthropic in that example does go bust. Well, at least I still have a data center in these great markets. And as long as I have not levered it too highly, as long as I can survive a hit to my earnings for a year or two, give it a year or two, and someone else who want that data center space. So there is such build up and demand for these markets that even if you have a hit for a couple of years, I'm confident over time, someone will in that data center. So that's how I think about that example. But the last example, which is I build a data center for anthropic or XAI or open AI in the middle of a field somewhere. And I do it because power there is cheap, construction costs are cheap. It's training AI. So I don't need to be around population centers. And I apologize. I probably throw a term out there because they're in common usage in my world. But just to explain that for a second, AI data centers can either be for training, which can be anywhere, or they'll be for providing a direct service to customers, which are called inference data centers. And so when you're on your iPhone or computer, and you're communicating with the AI, that data center that how the servers that you're connecting with, it needs to be near by you, so it doesn't take so long. - Right. - That's called inference. So if I'm training, I can be in the middle of nowhere, so I've got an anthropic training data center in the middle of nowhere, oh my goodness. Like, we just would never do that, because if the funding stops, I'm not sure what I got. - Right. - You know? - Oh, makes a lot of sense. - We read all the time that these data centers use an enormous amount of power. Is there a tension mic between these large power demands of the digital buildout and the desire to transition to more renewable energy sources? - Yes, yes, absolutely. - The follow-up is, and is there anything to be done about it? - Build a lot of power. (laughs) - A lot of power. - You know, so just to put a quick stat around it, US power was staggered for like 30 years. We had no net US power demand in place for 30 years. We became more efficient with how we use our power. We exported a lot of power on hungry industries to places like China. That power demand didn't increase. And then at the moment, it's increasing like 5% a year or something. - Right. - That's an unbelievable turnaround. And so you think about all of the industry that needs to be set up to cater to that massive increase in supply. We need to build to meet that demand. So the key component of a power generator is the generation unit, just practically speaking. They were set up for this no growth world for decades. And now they've got to ramp up all their factories and their suppliers need to ramp up all that takes a while. So I order a new power generation unit from Siemens or GE today. You know, in 2026, I won't get that unit to 2029 to give you a little bit of an idea. And then as you say, on top of that, you know, this transition is going on. It's going on at different rates in different countries. And I would say that maybe Europe as an aside, but certainly in Asia, I think it's going on as much for energy independence as for environmental. My division, literally all on me, but that is if you're China or Korea or Japan or one of these prices, you're an importer of energy. And so if you can build great energy and become a more independent energy country, then that's more security for you and your economy. Yeah. So on the next episode of "DriPowder," I'll ask Mike how he's zeroing in on opportunities across the energy sector and how de-globalization has opened up some surprisingly compelling opportunities in shipping and logistics. That's by definition making tri-less sufficient. Instead of coming directly from China, he is going through secure route. And so if you've got a given volume of triad and you're making it less efficient, by definition, you need more infrastructure to support him. I'm Himakarather. Thank you for listening. [MUSIC PLAYING]