Financing the AI Infrastructure Boom — Valentina Burger-Jiménez, DigitalBridge

Catherine: I’m Catherine McLean, founder and CEO of Dylan Green. I’m in New York today, and I have Valentina Burger-Jimenez with me. She is the principal at Digital Bridge. Thank you so much for joining me.

Valentina: Thank you for having me.

Catherine: And congratulations on your recent promotion.

Valentina: Thank you.

Catherine: To principal at Digital Bridge, can you introduce yourself and share a bit about your role and what it entails?

Valentina: Yeah, of course. So, I’ve been at Digital Bridge for three and a half years. I made a big career pivot coming into Digital Bridge. I used to practice law before, and now I’m more on the business side, working on the debt capital races for our portfolio companies globally. I tend to focus more on the America’s portfolio companies, and what I do every day is essentially threefold. I work on the structuring and execution of those debt financings for our portcos. It can be acquisition stage, refinancing, growth. There’s another aspect of it that will be more on the lending relationship side, so making sure that we know who does what, well, and where.
And then the other aspect of it is making sure that we know what we have currently in place in the market and where things are advancing. And what I mean by that is, because we have 45 portfolio companies plus, globally, each one usually has more than one financing in place or has had more than one financing in place. Making sure that we know exactly what that has looked like in the different stages of that portfolio company, making sure that we know where the market is headed and what new financings there may be, what novel structures are coming up. And so having that kind of knowledge database is also key for what I’m doing currently.

Catherine: Digital infrastructure is highly capital intensive, as we all know, from a debt markets perspective. What makes these assets attractive to lenders?

Valentina: I think it goes back to the infrastructure aspect of it. So in terms of being highly recurring assets, cash flowing assets, predictable and stable nature of them is what then makes it attractive for lenders that they know they’ll have a safe port in this storm to be able to lend to. I think obviously for digital infrastructure specifically, whoever is listening to this podcast will likely do so by somehow entering into the internet web. And so the physical infrastructure that is happening for that connection is going to be a baseline for making sure that it is recurring and that it’s stable, that it’s something that is not going anywhere, right? So the more we use our phones, the more we are logging on at work, at home, on just our everyday and consuming that data, we see it as kind of fundamental in the way that we’re conducting our lives these days. And there really is, I think for very few people, a scenario where you’re probably going to be reducing your consumption of data in an overall spectrum. You might be saying, I’m not going to look at my phone when I’m having dinner with my kids, but you’re also going to be at work and logging onto your computer using obviously email, but also all these other types of now more advanced AI, for example, that are just using so much more compute power.

Catherine: I mean, I was talking about just today, like how much I use AI. And like a year ago, I didn’t even really use it that much. I mean, only a year ago. It’s just incredible. I have one of the women that works for me. She’s been on maternity leave. She’s been gone for a year. And I’m like, wait till you come back. You’re not going to have to take notes from me anymore.

Valentina: There’s certainly so many aspects that are being facilitated by it. And we always remind ourselves that this current version of AI is the dumbest version that we will see. It’s only in its infancy, really. So the more that these models are going to be developing, the more they’re going to be able to do. That also comes with a usage of compute that might be denser. And so that means that the infrastructure to make sure that this is serviceable then needs to progress as well.

And that has translated right now into making sure we have data centers that are able to take on that level of IT capacity, that you have all the fiber connectivity. And then you also go back to like, OK, you’re using it on your phone more and more probably. And that means using cell phone towers or small cells in venues or in buildings or outside on the street so that you’re densifying those networks and people are able to access as smoothly as possible and have a low latency so that there’s plenty of speed and there’s no lag.

Catherine: Speed is important. And when lenders are underwriting digital infrastructure deals today, what are they most focused on?

Valentina: So I think one of the focus right now is definitely there’s a power aspect of it in terms of the data center, project financing, new data centers being built. There’s making sure that you’re going to have reliable power is key. And so there’s definitely a big focus in that aspect of it. The long term contracts that may be out there that can be from a hyperscaler perspective, but then also if you turn around and think about the co-location or enterprise data centers that may have a lot of different tenants in place, then that’s going to show you a history of renewals of those contracts or those leases. So you’re also being able to take comfort as an underwriter that there is a history of people renewing and it’s not just one single contract that is going to be due for renewal in a decade or more from now.

Catherine: AI driven demand, as we were discussing, has pushed data center projects to unprecedented scale. How has that changed the way you structure debt, whether that’s tenure, covenants or lender mix compared to just a few years ago?

Valentina: There’s a lot of changes in terms of the speed and scale of deployment. So as you were saying, a year ago, maybe you weren’t using AI that much. When it first came out, OpenAI with ChatGPT was one of the ones that hit 100 million users, absolutely the quickest in terms of adoption of technology and prior. And we see that also in terms of the different hyperscalers now and others that are interested in going into AI, making sure that they’re deploying quickly. So there’s definitely much bigger facilities. There’s a lot of project financing that is taking place to build those facilities that can have a level of service like cooling and power that is much larger. So where that translates is there has been a lot of demand or there has been a lot of financing, very large financings in the last couple of years. And now we’re seeing a little bit more of exploring outside the box. And what I mean by box is maybe we were doing more project financing. Lately, we’re also seeing more high yield bonds for construction purposes where the issuer will be able to tap the capital markets, the debt capital markets, and get the funding that they need, the financing for that construction by going into a different pool of capital that what the project finance bank lending for the most part would have been. We’re also not only doing project finance that is strictly with banks. There are certain tranches of that that are more private placements. So going into institutional capital as well that has a pocket for the private placement type of financing. So there’s many different types of pockets of capital that we’re looking at that the industry itself is kind of pushing and exploring. Investment grade bond issuances are also something that we’re seeing a little bit more of a trend as well. Then in addition to all of this, eventually when those projects are now cash flowing and have reached stabilization, they’ll go into a more mature market like the securization market. So there’s the ABS market or CMBS market that data centers have been tapping. At Digital Bridge, some of our portfolio companies were the very first ones in data centers, both hyperscale with Vantage, enterprise collocation with Data Bank. And ever since, we at this point have been issuing dozens of billions of dollars into this market. And we account for about a third of the digital infrastructure ABS market when you take into account digital centers, digital bridge portfolio companies. So we’ve been extremely active.

Catherine:I had this extraordinary figure.

Valentina: We’ve been extremely active across the board. One of our fiber companies more on the enterprise side, Zayo, just priced the latest issuance of ABS due to close tomorrow. And that’s going to be a 2.3 billion dollar issuance overall that is going into both the 144A more public market and privately placed trenches as well. So there’s just a very large volume of paper that we’ve been able to place in the securitization market overall. And we’re continuing to explore different types of structures, both on securitization side, but as I was saying before, like finding a more permanent home for these financings so that we can continue to recycle the capital from the banks and from other lenders, private credit as well. That’s another pool of capital that we also stay pretty close to because our portfolio companies will have different needs at different stages of their lives. And also for different types of needs.

Catherine: When lenders are underwriting a new data center today, what around power are they scrutinizing most, do you think?

Valentina: So that’s really power, as I mentioned before, it’s like one of the main questions that is coming into play. It’s really not a secondary diligence item. It really like there’s going to be, is there sufficient generation? Like ideally, this would already be grid connected, but there are other options in terms of making sure that you have power. So it could be up behind the meter, temporary solution or adjacent to like before you get connected, or there will be a power generation plan that is co-located with the data center. And we take a look at, the lenders will take a look at, and so will we in terms of the underwriting thesis for these projects where you need interconnection. There’s the generation aspect. But when we can make sure that there is also a cleaner energy aspect to it, we’ll try to do that. But obviously like where the wind doesn’t blow or the sun isn’t shining, we need to make sure that there are ways to have power connectivity that is 24-7 and no interruption, because that’s the key of a data center, right? Like that you’re not having any type of interruption on this service because at that point it is mission critical.

Valentina: Right. Hospitals, like it could be.

Catherine: Hospitals, financial institutions, all of it, right? Doing traffic lights, right? I mean, could be. Exactly. Yeah, quite a mess. So that kind of like resiliency is also like super important to make sure that it stays on the entire time.
Catherine: Are you seeing more data center projects integrate on-site power from the outset? And you had mentioned a couple, but what forms of generation are most common?

Valentina: Yes. So we definitely are seeing more of that kind of dual facilities where from the very beginning, there’s going to be the building of a data center and power generation. We’re even seeing now where the financing is being done together. So that is giving people a little bit more to think through and making sure that that structure is able to service both concerns and opportunities. And in that sense, like what kind of generation, there’s definitely a lot of gas power and battery storage as well. I think that’s the one that we’re seeing more. I know that there’s also talk about nuclear and where we are there. I think not just yet. For the most part, we think it’s a little bit like decade far out right now.

Catherine: As AI demand is increasingly driven by hyperscalers, how are lenders thinking about tenant concentration risk and data center financings?

Valentina: So I think for this point, there are kind of like the top hyperscalers right now. And in terms of concentration, I think investors constantly look through to see like what are then these hyperscalers providing what kind of service? Because you will not necessarily have the same type of application. Now you can have also hyperscalers that are doing AI. You can have a hyperscaler that is more cloud based. And those are really two different business models. You’re going to also think through the kind of positive credit qualities, long term commitments that they’re making are really important key for making sure that the underwriting is being done to very long term predictable cash flows. And so we have these conversations with investors, pretty often debt investors. And sometimes they are able to deduce because not always will the data center operator be able to disclose the specifics of the underlying lease with certain hyperscalers. So they’ll take a look at making sure, OK, we think this might be this hyperscaler, but it’s servicing this kind of end customer as well or this kind of product. And so that way they kind of keep tabs on the ultimate tenant diversification. But they also layer in a diversification in terms of the operator and the sponsor. So when you add all those different layers, then there’s more comfort in the underwriting that they’ll do, even though it’s a relatively small pool of hyperscale operators.

Catherine: Goodness, it’s very complex. There’s a lot that goes into it. As states and local communities push back on new data centers, how are lenders pricing that regulatory risk?

Valentina: So I think that we’re seeing more of the community pushback. I think some of the data centers are trying to make sure that they are active partners of those communities. There are a wide variety of regulations depending on the different states. So there will be different markets in terms of what we consider Tier 1 or expanding in Tier 1, Tier 2 data center markets. And there are also markets where you just don’t see that many data centers. And it might be because of, let’s say, it’s either scarcely populated or there is less availability of land. For the most part, a data center will try to be located where there is access to power and that there is also the space itself that you need to be able to build them. You’re also looking for, on the regulatory aspect of it, incentives from the governments to make sure that these projects are being built there. So in terms of underwriting them, it will depend on whether, like there will be different spreads of where you are if you’re in a market where there isn’t much of a demand for data centers versus where there is. So the classical example is more Northern Virginia, right, where you have so many data center companies. And so if you’re underwriting a financing for a data center in Northern Virginia, then you’re going to be pretty sure that if the tenant were to leave for whichever reason, which is a very difficult, I think, threshold to meet because it’s just so expensive and so difficult to be creating such a potential disruption for the business, for the tenant to move out. No one’s really going to move across the street just because they found something slightly 10% cheaper, just because you might be able to almost destroy your entire business. Too much risk, right.

Catherine: Exactly.

Valentina: But in terms of the underwriting of it, if something were to happen and the risk of being able to release that data center is much lower, and so you’re pretty sure your spreads can tighten significantly.

Catherine: And I would assume that, I mean, do you see more applications for Loudoun County and Texas, I know is jumping on the data center bandwagon. Are you seeing that certain states where there’s more appetite for the data centers, or do you see that it’s becoming more popular in general in other places that you wouldn’t necessarily have thought of?

Valentina: I think it has. I think it’s both, honestly. There is definitely more data centers being built in some of the Texas or Atlanta area, Ohio as well. But there are others that are kind of popping up in addition to them, where maybe you didn’t think that there was a data center market, but it makes sense because you have, again, the availability of power. You also have fiber that needs to make sure that it’s connected. All these data centers need fiber. Otherwise, they would be essentially an island. And then there is also, going out there, you need to build it. You will also, the tenant will need to have it somewhat accessible to be able to run the business and service their equipment in that data center. So it can’t be too isolated. I feel like maybe at that point is when people say, oh, what about data centers that are orbiting the earth?

Catherine: Oh, I didn’t know that was a thing.

Valentina: Not really. I mean, that’s like an idea that is being thrown out there. And it’s just like, OK, how do you build that? How do you then put all the equipment? How do you service that? How about making sure that you’re, sometimes you will need to exchange those chips that are inside the racks of the data center. It’s expensive. It’s expensive. And sometimes, well, it depends on data centers orbiting the earth. It’s going to be currently maybe prohibitive in terms of getting operations to do that.

Catherine: Well, one thing I will say is living on the earth, especially in planet America, nothing would surprise me at this point. We’ve been through a lot. So nothing will surprise me if we had data centers orbiting the earth. One of the things I just want to talk about, I know that’s not covered in the notes, but I’m just curious. Being a recruiter, obviously, I’m very interested in jobs. And one of the arguments in some cities and towns around data centers is that they actually don’t create that many jobs. They do initially, obviously, when they’re building it. But then afterwards, it doesn’t really mean there aren’t a lot of people in the data center. Do you have a thought on that? You don’t have to answer that because I know we’ve not put it in here. But if you have a thought on that, I’d love to hear it.

Valentina: So I think there’s definitely the job creation in terms of the construction. There will be jobs that are going to be servicing the data center itself. One of the things that a data center operator will also provide in addition to the power and the space, the cooling, it will be the physical security. That doesn’t mean you will have armies coming in to protect this in terms of like that job creation. But there is a physical security force, right? So you will have people guarding it. And in addition to that, I think it’s just you need to take a look at holistically, right? What kind of jobs are going to be created? And there might be a shift in the types of jobs that were done by workers before AI and post AI. And we’re really living through that. And so we’re really seeing that shift right now. I think there definitely will be different types of jobs.
I think some of them will also like I always think that where you are experiencing something, maybe that’s going to be more valuable in a world where everything can be more automated. And what I mean by that is like, you know, you going somewhere, you going to the park, you going to a concert, going to see a live performance, it’s maybe the value of that is going to increase or be different. That’s how I see it, at least in terms of, you know, it’s not like going to a music concert.
It’s very different from having AI created music. So in terms of those experiences, like there’s definitely a shift in terms of what job creation there will be.

Catherine: Yes. It reminds me a bit. My son came home one day and he was like, we did he say they had gone? He was like, we went to Greece today. And I was like, you went to Greece today, honey? And he was like, Yep, they went to Greece. They had Google Earth. And they went and they took them right to the ruins. And they had the TV. And there they were, like, literally in real time, like, you know, it’s just so astonishing. Like, I mean, he was in Greece today, literally on Google Earth, but it’s obviously not the same as being in Greece.

Valentina: Yes, it’s close. But that reminds me, there’s a place in Rome, I think it’s called Palazzo Valentini. And it’s Roman ruins. You go inside and it is like the actual ruins. But then they take lasers and recreate what it looked like with color inside. So it’s a very, like, it’s very much of an immersion experience. But with a very physical touch. It’s not the same as being in one of those, like, you know, four white walls that they just flood with color. It’s really recreating and being in ancient ruins from the Roman times.

Catherine: And you’re in that space.

Valentina: And I feel like that’s going to be so much more impactful for anyone to live through and be there, even though you might have some component of technology, versus me looking like, you know, just going on, like seeing a video of it.

Catherine: Well, thank you so much for taking the time to talk to us today.

Valentina: Thank you for having me.