The data center surge: Building the backbone of the digital age

[Music]

Fahmi Enam: Hello and welcome to ‘What's the Deal?,’ our investment banking series here on J.P. Morgan's Making Sense. I'm your host, Fahmi Enam, head of Media Communications and Digital Infrastructure at J.P. Morgan Commercial Banking. Today we're going to talk about all things digital infrastructure from data centers, fiber, tower, as well as the amount of capital that is being raised to support this industry and where J.P. Morgan truly shines in this space. Joining me today in the studio is Scott Wilcoxen, global head of Digital Infrastructure Investment Banking at J.P. Morgan. Scott, how are you?

Scott Wilcoxen: Doing well, thanks for having me.

Fahmi Enam: I'm glad you could join me today, Scott. So now let's dive right into the discussion. First of all, digital infrastructure. We are hearing about these words in print media, in news, all the time, every day. What does it all mean to you and what falls under the umbrella of digital infrastructure?

Scott Wilcoxen: It is an interesting question, Fahmi, and as somebody who spent my entire career covering the space, it's one I don't think about often. But early on when I got started in this business, somebody at Equinix had actually suggested I read a book called Tubes. It's a very basic book. If anybody wants to pick it up, you can get through it in like an hour. But they use an analogy at the beginning of the book and they talk about somebody sitting in their living room in New York typing out an email to a friend in Tokyo. And most of us just assume that happens magically in some ephemeral thing called the cloud or called the internet. But physically what that actually means is but for the Wi-Fi connection in your apartment from laptop to router, there is effectively an unbroken physical connection either over copper, fiber, whatever, all the way from your living room in New York to your friend's living room in Tokyo. That's digital infrastructure. It's all of the physical behind all of what I like to call the magic that we all do on our iPhones on a day-to-day basis. Most people don't realize it, but there's billions of dollars of physical assets supporting virtually every technology experience that you enjoy today.

Fahmi Enam: Got it. And so when I think about these assets, just for our viewers, we're really talking data centers, fiber, subsea cables, cell towers, and similar assets that have been put in place over the years by either private industry or, in some cases, government.

Scott Wilcoxen: That's spot on.

Fahmi Enam: Okay, so first of all, that's actually a great analogy and the next time I send you an email, I'll think about these electrons leaving my phone somehow and getting over to your house, which is an interesting way to think about how we send emails. So when we think about these assets, the digital infrastructure assets, are there like any other infrastructure assets where it's high CapEx, there is a strong competitive moat? Or is there something special about this sector, these assets that are making them such an important point of topic today?

Scott Wilcoxen: Yeah, no, look, I think there's a lot of similarities. And if you actually think about the history, right? 15, 20 years ago from an investment perspective, this really was the domain of kind of classic private equity. And I think as the industry matured and the asset base got larger, more parties started to take a look at this. And I think what they quickly discovered and realized was, from an infrastructure perspective, the downside prediction that you would see in any other kind of classic infrastructure asset class exists. These assets deliver essential mission critical services. There are meaningful competitive moats. Some of that's just the scale of the investment, right? You think about your run-of-the-mill data center as being a billion dollars plus. Some of it's permitting and zoning. You think about cell towers, we've all heard of NIMBYism, right? Nobody wants multiple cell towers in their backyard. So once one's permitted, the likelihood you're going to have a second in a similar geographic region is extraordinarily limited. And then a lot of this stuff is ultimately when you think about sort of the revenue for the asset owners, it's all being delivered under long-term contract generally with very high quality counterparties. Like you, Fahmi, are not paying the cell tower monthly with your cell phone bill. You're paying Verizon or AT&T or T-Mobile, and they're ultimately paying the cell tower. And that tower, of course, is a mission critical piece of their overall infrastructure base. You combine all of that with the fact that data traffic, technology usage, et cetera, really only goes one direction, it goes up. And so the need for incremental infrastructure and the ability to continue to increase the monetization of that infrastructure is really powerful. So while you've got a lot of similar characteristics to classic infrastructure, what you have in digital that you don't have elsewhere is just the growth.

Fahmi Enam: That's actually a great way to think about it. And as we think about growth and as we think about investment in digital infrastructure, probably at the forefront of that are data centers. For our viewers' benefit, data centers essentially are large physical buildings that house all the infrastructure, servers, chips, cooling, networking, et cetera, required for processing large amounts of data. So historically, most of these data centers were supporting public cloud, but in recent years, I'm going to say 18 months, 24 months, most of the investments have been around AI workloads. So how would you characterize, Scott, the magnitude of capital required for a data center build out in this cycle compared to other growth waves?

Scott Wilcoxen: The words that come to mind are huge, large, and unprecedented (laughs). I mean the public cloud era, which I really think about is being more or less the early to middle 2010s. First of all, data centers, which capacities typically measured in terms of power draw, I mean, you were building 50 megawatt data centers. And so rough order of magnitude, let's just simplify it and call it $10 million a megawatt. That's a $500 million asset, right? And while public cloud build out ultimately involved hundreds of billions of dollars, it was frequently half a billion dollar chunks, give or take. The AI cycle is just significantly bigger. People are building gigawatt style campuses, which is a 10 plus billion dollar asset versus the smaller assets that were being built previously. And they're just building them in significant quantums. Depending on the research reports you read today, over the next three to five years, the sort of capital requirement behind all of this is not estimated in millions, it's not estimated in billions. It's being talked about in terms of trillions.

Fahmi Enam: And all of this will be driven mostly by the AI workload, or is there legacy, growth in legacy, needs for these data centers that are also driving growth?

Scott Wilcoxen: So look, a significant chunk of this, I would even argue the majority, is obviously AI focused. But this entire AI dynamic is additive to an already extraordinarily large investment base. And so if you think about public cloud services, Amazon Web Services, Microsoft Azure, software as a service type applications like Salesforce, which obviously a lot of folks use, that usage itself is projected to grow at something like 19 to 20% annually over the next five years. And while compute capacity doesn't necessarily grow one-to-one with revenue or usage, right, just given incremental efficiencies and things of that sort, it still grows materially with it. And so there's a report I read last week that suggested, you know, 10% growth in compute capacity would be required to support the growth in cloud spend. And if you think about the existing installed base, I mean, that's again, multiple gigawatts, something upwards of 30 gigawatts over the next three to five years. Again, you use the simple math, 10 million a megawatt. That in and of itself is, now I can't do mental math (laughs), but it's hundreds of billions of dollars just to support cloud. And so then you get... And that's just the data centers. So we're not even talking about the servers, the chips, the networking, the fiber, the power generation, which I think anyone who's picked up a newspaper in the past year realizes power capacity is at a premium these days in the United States. Then you layer on all the AI. And so there's a McKinsey estimate or a report that came out recently, that could be 100 gigawatts by 2030. That's over a trillion dollars of investment on data centers alone on top of the cloud. And then you factor in chips and so on and so forth, and that's how we get up into these 3 to $4 trillion estimates that are being thrown around.

Fahmi Enam: So, Scott, these are large numbers. Given the magnitude of the required capital, just U.S. tech majors alone are spending over $300 billion in 2025. What are the challenges and the opportunities in terms of assembling the scale of capital?

Scott Wilcoxen: Look, I think it depends. I mean, you mentioned the tech majors., roughly half of the dollars that we've talked about will ultimately be funded by the, quote unquote, "Tech majors." Hyperscaler CapEx in 2025 is estimated around $325 billion, which is obviously nothing to sneeze at. But against a combined market cap across that universe that's measured in trillions and cash on balance sheet that's measured in the hundreds of billions, it's a big bet, but it's not, you know, a balance sheet breaker. I think when you think about third party developer, operators, et cetera, who are doing a lot of the data center construction in the United States, that's where things start to get a little bit more nuanced. Starting on the debt side of the capital stack, if this is a 20 year contract with an A rated counterparty that's ultimately coming into that facility, that's pretty easy to debt finance. I think as you move away from that scenario, it starts to change. Whether that's an upstart AI company that's looking to take down a large amount of data center capacity to support their ambitions or what have you, the structuring starts to get a little bit more important but is still ultimately achievable. And we've seen many examples in the market of deals of that sort get done and we've seen examples as well where various ecosystem participants that have bigger balance sheets, that have better credit, you know, characteristics have actually stepped into support some of that investment, ultimately to continue to foster a thriving marketplace for their own product and services. When you move into the equity side of the capital stack, it starts to get a little tricky. Despite what the press talks about with respect to hundreds and hundreds of billions of dollars are being committed, I would actually argue there's a fair amount of equity saturation across the space and the dollars that need to be raised still are quite large. Just think about a $25 billion infrastructure fund. I don't know for sure, but I suspect that places them in the top 20, if not the top 10 in terms of largest infrastructure funds on the planet. At 10 to 15% kind of concentration limits, that suggests individual investment sizes of 2 to $3 billion. And most of them have already made investments. And so that's where you kind of start to run into challenges. And so forming this capital is increasingly requiring going to multiple pockets, across multiple regions, and in many instances, looking across multiple layers of the capital stack and incorporating both structured as well as kind of straight common equity to ultimately fund these investments. That, to me, is the backdrop where J.P. Morgan really shines. We're built to assemble capital across the entire cap stack. Senior debt, junior debt, structured, preferred equity, straight common, etc. We're active across multiple markets, whether it's bank lending, institutional and private credit, securitization markets. We have relationships with sovereign wealth funds, insurance funds, pension capital, so on and so forth. And we're really able to craft tailored capital solutions that ultimately enable the formation of capital in these quantums for these market participants.

Fahmi Enam: And that's a great point, Scott. And away from raising capital, when you think about all the other needs of these businesses. It could be operational, such as treasury, payments, FX, hedging around the power costs, setting up ESOPs, or it could be governance. Or it could be finding board members, or even getting introduced to future, strategic partners. I think these are all the places where J.P. Morgan can provide holistic, end-to-end solutions across the different pockets of the back that cover these companies.

Scott Wilcoxen: Absolutely.

Fahmi Enam: So let's talk about these large numbers and the pace of deployment being incredibly rapid. Whenever you open The Journal, you are looking at some more data centers being developed somewhere in the U.S. or globally. What are the risks here. And is over-building in your mind?

Scott Wilcoxen: I think over-building is at the top of everybody's list, maybe not in the way that you might immediately think. I think, to me, what that really comes down to from more of a root cause perspective is ultimately the pace and scale of AI investment monetization. I think it's very safe to say at this moment in time, investment has meaningfully outpaced revenue generation. It's not to say there's not monetization, but I think that the sort of scale doesn't quite balance. I don't particularly see that, in isolation, as cause for alarm, but I think the uncertainty around how, when, and in what magnitude end-user demand materializes in a monetizable way does make it very difficult to plan out over a multiple-year cycle. In terms of overbuild, kind of boom-bust type dynamics like we saw with fiber in the '90s for example, I don't really see that as a material risk. I could see a situation where monetization doesn't catch up to investment and development. I could see a scenario where development pauses and capacity is absorbed over a period of time. But I could also just as easily see a situation where uptake is so much greater than expected, we actually remain materially supply-constrained for years. I go back to a point I made earlier. Specifically on the data center and the physical infrastructure side, everything's being built with a signed contract in hand. And if you think about the large-scale tenants that are taking the space, they have multiple businesses, multiple revenue streams, and multiple uses for a lot of that space. And so again, like I think you can manage a situation where your needs change dynamically and the downside, to some extent, becomes development or new development pauses for a period of time versus assets are left abandoned the way they were with fiber in the '90s. I think the other thing that's helpful for people to think about, and this is one of those, it's always hard to quantify in front of credit committee (laughs), but um, just think back to 2007. That was the year the iPhone came out. The things that you do today on a daily basis and you don't even think about, whether that's online banking, person-to-person payments, hailing Ubers, calling Lyfts, GPS navigation with real-time traffic data, streaming entertainment, literally anywhere. All of those things have required meaningful physical digital infrastructure investment. And when you think about all the different things we're going to be able to do with AI over the next five years, 10 years, 20 years, it's really hard to comprehend what that's going to ultimately require. I just know it's going to require more.

Fahmi Enam: No, that makes a lot of sense. And obviously as this industry develops further, you know, the key would be to differentiate perhaps new entrants that are chasing a trend from developers or players that have done this successfully over previous years and have the wherewithal, the technical wherewithal, to develop at scale. And in particular, I think you and I both are seeing a lot of players that historically focused on the edge or co-location data center market now pivot and uh, focus a lot more on hyperscale development side of things. So that's something that we have seen in the market and I believe we're going to continue to see more of that. So we spent a lot of time on data centers. Beyond data centers, fiber and towers are also key pillars of the digital infrared revolution and we are seeing significant investments in these spaces from both private sector as well as the government through the BEAD program. So give me your 30 seconds on this.

Scott Wilcoxen: They're obviously super important. The dollars are a little bit smaller on a relative basis and so I think the press just remains a lot lighter. Fiber's ultimately the connective tissue. Go back to my example at the start of our discussion, you have to have connectivity. Data doesn't move in a vacuum, it moves over a medium, whether that's fiber or wireless. We've got a number of players that are pivoting more of their businesses towards building out connectivity for large data center campuses and things of that sort on more of the enterprise or the commercial side. And then obviously COVID accelerated the need to enhance connectivity and bandwidth for the home user as well. And we continue to see that at pace, and that's where the BEAD programs and some of the things that you mentioned from a government subsidy perspective have come into play. But there's a ton of activity going on, bringing the requisite connectivity across both enterprise and consumer. On the tower side, look, I mean, how many of us do the bulk of our stuff anymore on a, on a fixed desktop computer, right? The world's gone mobile. All of that requires the same types of bandwidth as they're needed in a, in a wired environment or a fixed environment. And so we continue to see investment there from the carriers as they build out networks, they overlay incremental capacity to keep pace with data growth. That sort of same build out is needed to bring that bandwidth to bear. Again, it's just smaller numbers so it gets a lot less press but it's critical and it's vital.

Fahmi Enam: Great. Makes sense, Scott. Obviously this is a complex and quickly evolving topic, and we at J.P. Morgan are focused on staying cutting edge here as it relates to serving our clients with all the advisory, financing, and operational banking needs in digital infrastructure. That brings us to the end of today's episode. Thank you, Scott, for sharing your valuable insights and expertise and thank you for tuning into another episode of ‘What's the Deal?.’

Scott Wilcoxen: Thank you, Fahmi, for having me and thanks to everyone for listening.

Fahmi Enam: We hope you found this conversation insightful and be sure to tune into our upcoming episodes as we stay up to speed with the markets. I'm your host, Fahmi Enam. Until next time, goodbye.

Voiceover: Thanks for listening to ‘What's the Deal?’ If you've enjoyed this conversation, we hope you'll review rate and subscribe to J.P. Morgan's Making Sense to stay on top of the latest industry news and trends, available on Apple Podcasts, Spotify, and YouTube.

To stay ahead of the curve, sign up for JPMorgan's In Context newsletter, packed full of market views and expert insights delivered straight to you. To subscribe, just visit jpmorgan.com/In-Context.

This material was prepared by the investment banking group of JPMorgan Securities LLC and not the firm's research department. It is for informational purposes only and is not intended as an offer or solicitation for the purchase, sale, or tender of any financial instrument.

[End of episode]