51黑料

Peter Antoszyk: Expand on that thought a little bit, Kunal. We've read how AI is the future, it's real, it's here, it's now, it's surging. What impact has that had on digital infrastructure and what does that mean as boots on the ground?

Kunal Dogra: In my view, there are really two mega trends that are powering the current boom in digital infrastructure, in data centers in particular. The first, almost putting a sign on it, is the migration to the cloud. Organizations all over are moving from storing data onsite to moving all of it to the cloud and having it managed often by third parties – think Microsoft Azure, Amazon Web 51黑料 – that manage their cloud operations. So that's been a big driver of the growth in digital infrastructure. I mean, at a very basic level, this podcast is happening courtesy of the cloud. Think about the implications of the cloud and the Internet on retail businesses. It's just fascinating stuff.

So while we're moving into the cloud, and we're probably, call it the second or third innings of the cloud, all of a sudden we have this growth in artificial intelligence which has come in. The major tech companies are obviously in an arms race to develop new AI models, and that's just created such a demand for digital infrastructure assets. AI, at the end of the day, relies on data. Data is the input, and intelligence is the output. And so the digital infrastructure assets, that DigitalBridge and other players in this space own and operate, are key to all of this.

Kyle Colvin: Really well said, Kunal. I think absolutely the biggest tailwind we're seeing right now in the digital infrastructure space is, is gen AI and AI infrastructure. Companies like NVIDIA, AMD and Intel are producing these new specialized AI chips, GPUs to address the gen AI workloads, which are compute-hungry and power-hungry, and Kunal mentioned early innings, second or third inning. You know, something our CEO Marc Ganzi really likes to say: “We're only in the early innings of gen AI-driven demand.” It's too early to tell how big the gen AI adoption is going to be. The early thinking, though, is AI could be as big or bigger than the cloud. You can use the analogy that OpenAI's ChatGPT platform represents the “iPhone moment” for gen AI.

To put into context, in just two months, we saw over 100 million active users for ChatGPT, and that's the fastest adaptation of any consumer technology in history. If you compare it to other global platforms like TikTok or Instagram, it really puts into context how fast the technology is being adopted. ChatGPT reached 100 million users four times faster than TikTok and 15 times faster than Instagram. It's just absolutely incredible to see the pace it’s going at.

Peter Antoszyk: That's amazing. What does that mean for the development of data centers, which DigitalBridge is particularly focused on. What unique attributes are required to develop data centers that can accommodate the needs of the AI development?

Kyle Colvin: Kunal touched on it. The leading logos in the space, the very large tech companies, whether it's Amazon, Microsoft, Apple, Meta, Google, they're increasing their investments in capacity to meet the current and projected demands of AI. That's the important thing to keep in mind when we're talking about the AI revolution – the impact it's going to have on digital infrastructure. Think about a ChatGPT search and how much more power consuming it is than a simple Google search. The chips used to optimize that gen AI consume more power than the previous generation of chips, and with increased data usage comes the need for increased amount of power over time, so it's going to lead over time, to higher megawatts per facility in these data centers. Higher power density per rack in the data centers. And what that means is that it's going to take time to build out that necessary infrastructure, similar to taking a decade to build out the cloud. We think it's likely to take about a decade to build out the actual infrastructure to support and fuel that AI demand.

Kunal Dogra: Kyle, my understanding is that the power capacity needs of an AI data center are something like 3-4x that of a traditional hyperscale cloud data center. It's just incredible.

Kyle Colvin: No, it's a really good point, and there are estimates that going forward, that data center power consumption is going to be dominated by these AI workflows. There's one estimate that I've seen that estimates that in the next 15 years, 80% of all data center power will be consumed by AI. This is incredible, astronomical. We've already seen an incredible amount of growth in this sector, attributed largely to the cloud, and now you're adding to that with the AI demand. And you know, it's going to lead to higher power density as we mentioned, in turn leads to more cooling. That leads to overall increased power consumption, and without creative solutions, that growth and power consumption especially for data centers, it's going to lead to strain on the traditional grid and power systems that won't be adequate traditional power to fuel data centers.

Kunal Dogra: So before we get into some of these challenges, which we should definitely touch upon, Kyle, I think it's helpful to explain to users, how AI data centers can differ from call it traditional hyperscale data centers, right? My understanding is, if the use case of a data center is to train AI, you're obviously using an enormous amount of computing power to process the data, but the location of the center doesn't matter as much. It can be located further away from the end users, but when you start getting into inference models for AI, where you're really solving problems – the goal of the AI is to solve problems, to create productions – you really need to be close to the end user because you can't have delays in transferring information to them. So that's an easy way to differ between cloud data centers and AI data centers.

Kyle Colvin: It's an excellent point. You think about all of these applications and how consumers are going to be using them. What are they going to be using? They’re using their smartphones, laptops, desktops, tablets. The low latency is actually going to be key and so it's not just about the data centers or outside of tier one markets, it's really delivering it through edge infrastructure, bringing it closer to the end user. Really what that means is gen AI is going to be edge-delivered to consumers, and that's why it's going to take time to build out the necessary infrastructure to support the incredible demand we’re seeing.

Peter Antoszyk: Can you just expand on that? When you say edge infrastructure, can you describe a little bit what you mean?

Kyle Colvin: Absolutely. When we’re talking about edge, what we're really talking about is getting closer to the location of the end user. When you're doing with that, the response latency is reduced so bandwidth is going to be optimized. There's also increased data autonomy.

When you're sending data back and forth for long distances, there’s increased lag, especially with gen AIand the amount of power it takes to consume. What you're trying to do is reduce that latency as much as possible, and so you're going to continue to build out the infrastructure in a way that closes that gap and brings it closer to the edge where the data is ultimately processed and used by the end consumer.

Kunal Dogra: In a way, you’re landlord to some of the largest tech companies. Can you talk us through what that means and what your customer contracts look like? Obviously, I've looked at some of those with you guys, but I think it would be great to talk about that because the nature of your contracts are so fundamentally different in this space versus other sectors.

Kyle Colvin: Absolutely. One of the things we touched on at the beginning, and one of the things that underpin digital infrastructure, is really those long-term contracts. So what we're doing when you look at how you characterize “what is digital infrastructure”, it’s the underlying customers that we serve. It's really not retail customers. It's very large tech companies, it's MNOs, those large digital consumers. And the underlying contracts themselves are going to be very long in length. You know, whether it's 10, 15 years plus, and it differs depending on whether it's a data center vs. macro towers vs. when we're in small cell networks. It's going to differ depending on the sub vertical but generally, what we're trying to do is lock-in so you have that stable, predictable cash flows over time, which is not dissimilar to infrastructure, generally.

Kunal Dogra: And your customers, at this point, you're building to suit or are you building on spec?

Kyle Colvin: It's a combination, really. I mean ultimately what we're striving to do is deliver solutions to our clients. There are certain types of digital infrastructure where it's delivered to spec for clients, whereas others you're going to have multiple tenants in the underlying infrastructure itself. And so it really depends on the underlying customer and what those needs are. What we really focus on is being collaborative and showing up and delivering solutions.

Peter Antoszyk: Describe, just to put it in context, a successful data center investment project and, give a sense of the complexities of the build-out that is needed to accommodate the needs of the AI and cloud computing.

Kyle Colvin: We have a lot of portfolio companies, so there, there are a lot of options to choose from. But one that comes to mind that I spent a lot of time working on the last couple of years is Switch, a company we took private in December of 2022. It's a leading 100% renewably powered enterprise data platform. It's headquartered in Las Vegas and has campuses in Nevada, Michigan, Georgia and Texas. One of the things that really attracted us to Switch is that there's a few different aspects: Their proprietary design is really unmatched, they have zero downtime, they also did a great job of securing land and power, and they're focusing on building campuses adjacent to low cost, renewable power. And touching on some of the things that we discussed, especially with the bottleneck of power – and we can talk more about that – it's absolutely key when we're talking about these builds and seeing how demand is actually exploding to focus on a company that actually can show up and deliver for customers. Instead of building a 50 MW facility, it might be a 200 MW facility and you're running out of real estate. There's companies all over the U.S that are scrambling to try to secure land, to secure power and Switch is a great example of one of our companies that really was at the forefront. They were leading, they were an early adopter in the push to use renewable energy in data centers. And you know, where they're set up, whether it's a combination of solar or battery storage, there's different options, it's really quite remarkable and impressive also from an ESG perspective. A lot of their underlying customers have their own ESG goals that they're trying to achieve. When you can deliver to a client 100% renewable energy, it's another way to deliver to what the customer needs.

Peter Antoszyk: What's the competitive landscape look like for digital infrastructure? Who are your competitors, and how do you fit into that ecosystem?

Kyle Colvin: We've gone over our own digital transformation [as a company] as we've divested our legacy real estate and real estate assets to be able to really scale our platform to deliver on a global basis for the digital infrastructure demand that we're seeing. We are an alternative asset manager but we think of ourselves first and foremost as operators. It's the operating DNA of our senior leadership along with our senior advisors and operating partners and what we think really differentiates ourselves from other GP's - we're seeing more GP's come into the space. They want to have a digital story. What we think distinguishes ourselves is our experience and our track record, especially solving customer problems. Just to give you a few examples, we touched on securing and bringing power to fuel data centers; the latest advances in cooling technology; dealing with land acquisition; permitting, zoning issues; developing and connecting fiber, just to name a few. And so we believe we differentiate ourselves through our relationships with our customers. Our competitive advantage is building these great customer-focused companies and what helps with that is having deep roots in the sector, and it really goes to our senior leadership team who have built strong relationships with key industry players over time.

We believe really sets us apart is our ability to actively manage our portfolio of assets as an operator in this space as we collaborate with our customers to deliver solutions.

Kunal Dogra: That's important because the risk profile of digital infrastructure assets really differ from other sectors, right? Putting aside technical due diligence, when you enter into the space, there's a big learning curve. Investors, for example, are focused on drilling in on somewhat esoteric issues like the power supply, the nature of the physical network, the condition of the assets. From an M&A perspective, these are really complex transactions. The covenants are very unique. Purchase price adjustments are bespoke to the industry. When you and I have done deals in this space, the complexity has just been off the charts. U.S. deals for example often involve REITs which, of course, have their own sets of issues to wrap your arms around. That is what makes this space so exciting – there's just so much activity and there's just a lot of complexity in that space.

To Kyle's point, because DigitalBridge and its peers are owner-operators that, that really sets them apart, I'm seeing a lot of investors rotating into this space, but they're also teaming up with established owner operators, who really have the expertise to allow them to invest successfully.

Peter Antoszyk: You mentioned one of your competitive advantages is being an owner-operator, which is key and important, but you also mentioned your capital formation strategy, which is a differentiator. How is it different?

Kyle Colvin: So, I think it might be helpful to get a little bit background on myself. Similar to Kunal, my training background is as an M&A lawyer. I focused my career prior to joining DigitalBridge on advising private equity and financial sponsor clients on a variety of complex cross-border M&A transactions. Today, I spend most of my time at DigitalBridge working with our investment team on our flagship equity strategy, where we invest globally on opportunities in digital infrastructure to generate alpha for our investors. One of the things that we've seen as the demand for this infrastructure continues to grow. We've extended our investment strategy from our flagship equity business to additional verticals, including what we refer to as Core Plus, where we target certain strategic sub sectors that offer predictable and consistent yields such as hyper scale data centers and macro towers and also our digital credit strategy, along with our digital venture strategy, where we look to invest in the software layer at the intersection of infrastructure and technology. This is what we refer to as our full stack approach to digital infrastructure. It enables us to pair capital to the right risk adjusted opportunity for investors as we invest across the digital ecosystem.

Kunal Dogra: One of the ways, Kyle, that I've seen DigitalBridge generate alpha is the flexible and creative approach that it's taken to its capital stack. You and I have collaborated on several complex and unique deals involving your capital stack. You have embraced, for example, GP-led secondaries, preferred equity, convertible notes. There, there are also other things that you do with your capital stack, for example, how you separate your companies. Can you talk listeners through that?

Kyle Colvin: I like to say it's really good for job security because we don't do anything simple or straightforward. We're always thinking, on the cutting edge of what makes sense to where to invest and how to deliver for underlying investors. At the same time, obviously delivering for our customers. So, you touched on a few Kunal. You know, especially in the data center space, it's pretty obvious when you see the growth with the cloud that we've touched on with the growth of AI demand, the very capital-intensive businesses. You know we have businesses that are outperforming the business plans and then having updated business plans that reflect what the AI business case is.

We've been successful in identifying certain assets that might be attracting a different pool of capital. I mentioned our core plus strategy. Vantage is a great example of a company that you and I have worked with, Kunal, one of our great data center companies. And one of the things that they've done over the last few years is separate out their stabilized data center assets – when they're fully leased-up – because it's a different risk profile, a different investment profile when you're going out and securing land, getting a lease for a customer, building the data center, having the customer take space and ultimately having it come online. It’s much different than when that's fully leased up and operational with the cash yields that are then in place. So, what we've been able to do is find another way to return liquidity for investors, return capital to them and also raise additional capital to fuel the growing birth of the business as we continue to focus on developing it. The other things that we continue to do are working with businesses – whether it's edge or MNO's – that are looking to invest more in their own businesses, and they don't want to basically have the cost of their own infrastructure. So, whether it's doing sale lease-backs where we buy the assets and passive infrastructure and then lease it back. Just getting creative in the way that we come up with solutions.

Another good example I think, is the joint venture we did with Liberty Global in Europe. AtlasEdge is a really great project. You're looking at one of the parts of the sector that are really focused on delivering edge, as we talked about. Instead of going out and doing that either on our own or looking to acquire where companies were trading in multiples, we saw an opportunity with a great partner in Europe where they contributed their digital real estate holdings and assets into the joint venture, and we contributed capital along with our operational expertise. There are a lot of other GP's that would have loved to partner with Liberty to do that, but they chose to partner with us because of that operational expertise. It's just thinking outside the box and getting creative with debt financings, with securitizations, and just trying to find a way to provide as much flexibility to our management teams to build the business that they need to deliver to the underlying customers.

Kunal Dogra: Absolutely. Kyle, in your most recent earnings report, Marc [Ganzi] mentioned your CapEx for this year and the CapEx needs are just incredible; Peter, to put this into perspective, I believe they’re spending approximately $15 billion. Is that the right number for budget and CapEx this year?

Kyle Colvin: That sounds right. It’s a big number. We have our work cut out for us, we don’t rest on our laurels. We know what’s ahead of us and excited about the opportunity.

Peter Antoszyk: That is incredible. I’d like to pivot back because this leads into what one of the areas that we touched upon briefly – when you’re deploying all of this capital and building out these data centers and you run into challenges. You highlighted the big one that everyone is focused on is obviously the availability of energy. Can you speak to any other challenges that the industry might face?

Kunal Dogra: [Energy] really is. Where I see this going 5-7 years out, is really going to be an emphasis on building sustainable data centers. I know, Kyle, you guys have started to do things with, for example, solar, you're using different sorts of solar cells in your system. So, I really think the near-term challenge in this space is sorting out the energy constraint issue, and it seems like people have a variety of different approaches to it. I’m pretty confident we'll solve that in five to seven years. But, short term, my sense is that's the biggest issue.

Peter Antoszyk: Kyle, from your perspective, is one of the solutions that you're looking at is building your own power sources?

Kyle Colvin: Absolutely. We've been doing a lot of work as a firm the last couple of years, looking at the energy transition and seeing the writing on the wall with where the demand is going. Even without the AI revolution, there's already this energy transition going on…seeking out renewable energy and not enough of it to fuel the growth that we see there. And now you add AI to that. So from our perspective, ideally, you would like to be like Switch; you'd like to be 100% renewable. It's probably not very possible for the vast majority of these data centers on a global basis. So, what are you trying to do? You're trying to at least try to carve your own way out to have some energy independence, so you're not 100% reliant on the grid. You know whether you can be 60-70% energy independent and it’s a combination of multiple sources. There's probably not going to be one technology or one strategy that's going to be the silver bullet to solve the problem, but we and the rest of the industry are going to have to be creative and have multiple strategies to solve this. So, we're doing a lot of work right now as a firm on energy transition, and we hope to have some things coming out that we're focusing on to try to solve some of these issues. But it's one that we're acutely aware of.

Peter Antoszyk: Did I read recently — I may not have this right — maybe either of you: did Amazon just acquire their own nuclear power plant?

Kyle Colvin: I didn’t see that, but there are definitely you know, different large tech companies that are looking at all possible solutions. Whether it's nuclear, whether it's deal generated with energy storage, solar, wind, a combination of different things. And when you can be on the grid, when you're off the grid and trying to supplement that. So, you're not 100% dependent. I think that's going to be the key.

Kunal Dogra: You know, we have a massive utility problem that is not just limited to digital infrastructure. It goes to industries in general. Power is just such a constraint, and I do believe we have to look at ways to make nuclear more accessible and to take some of the regulatory components down in that space down to a more reasonable level, so that it can be really used as a solution in the States.

Peter Antoszyk: It's going to make achieving some of the aggressive climate control targets, reduction in fossil fuel, l very, very difficult as the demand for power, not just for digital, data centers, but generally speaking…think about electric cars and building another grid for that and digital infrastructure and you go on and on. It's just the demands are increasing as opposed to decreasing.

Kunal Dogra: We have to be practical about it too, right? You know, if you think about the power needs of, let's use a data center just for example, I can't even imagine if we were focused only on solar or wind turbines. It would take hundreds of wind turbines and god knows how many solar panels. The real estate needs would exponentially. So, there isn't going to be just one solution in this space, it's going to be an amalgamation of these.

Kyle Colvin: So, definitely power consumption is definitely the big one from an ESG perspective that we've touched on. A couple of others come to mind, one of which is regulatory. We've seen, especially in the digital infrastructure space, an increased amount of regulatory oversight which was really led by Europe the last couple of years. It just is increasingly becoming more regulated and you have different governments looking at what is “critical” infrastructure. It's getting to the point, in Europe where basically any data center can be characterized as “critical” infrastructure. The government then takes a very close look at what you're doing, where the underlying customers are, who those customers are. Are there government agencies? Who in the capital stack do you have, including your LPs? So, something we have to continue to navigate, we're seeing that more in the U.S. with CFIUS as well, is changing the goal posts of what critical infrastructure is. So that's something that we continue to navigate.

The other one on the regulatory point that I would flag is really data sovereignty laws; we're seeing that especially in Europe where the transmission of data, especially personal information, is heavily regulated, and, you know, in order to address that, it will even dictate where you're building and transmitting that data as well.

Peter Antoszyk: What about the availability of land and finding the right locations? Are you getting push-back from localities, or how is that influencing your decisions?

Kyle Colvin: No, it’s a great question. I think with respect to going out and building the digital infrastructure, you have to look at different aspects of the different challenges that go into that. One, as you mentioned is land acquisition and whether it's zoned correctly, getting the right permits; is there local opposition? It's not just securing the land, but is it adjacent to low-cost energy that's maybe off the main grid? And, you know, we're seeing, especially in the U.S., in certain markets, land prices just increasing at really incredible rates. Not just the land itself, but because of what it means for what can be built there and what the competing challenges and opportunities are for the use of that land. And really, whether it's adjacent to a low-cost use of power as well.

Peter Antoszyk: I would also think though, it would be a generate tremendous opportunity for the localities where the growth and development of the areas in which you're developing these data centers.

Kunal Dogra: Part of the reason, Kyle, for that competition is because data centers are clustered, right? The data center clusters are often located in specific locations. In the US, for example, data center clusters are all in the Northern Virginia area and my understanding is that's three to four times the size of the next biggest cluster locally.

Peter Antoszyk: In the development of these centers, what's the role of DigitalBridge in the security of these centers?

Kyle Colvin: I think with respect to data centers in particular. If you go and visit some of these places and Switch, which we touched on, is a great example, you go in and look at the facility and it's basically akin to a military facility with clearance. Obviously delivering for our customers is key and part of that is the security of the underlying data. Typically, as the owner or the landlord, as Kunal put it, we don't actually have access to the underlying data. The customers actually do, but it's very important that we are providing the security for the facility itself.

Kunal Dogra: So, the way to look at this, Kyle, is you provide the facility, you provide the fiber connectivity, you provide the power, you provide the cooling and the security. Is that about it?

Kyle Colvin: That’s right.

Peter Antoszyk: So, what do you see as the trends going forward? Where do you see this industry five to ten years from now?

Kyle Colvin: Well, the easy answer which we've touched on obviously is, is the AI revolution and the demand for that. We're seeing it with our increase for CapEx, that Kunal has mentioned. Really just delivering for our customers. As we mentioned at the top, the process to build out the necessary digital infrastructure for the AI demand is not a one, two, three-year process. We're in the early innings and we think it's probably a decade long process to build that out – not just fitting the data centers for their fit-for-purpose but delivering along the digital ecosystem, if you will, to be able to provide to the end users. Like we talked about, especially going from data centers to fiber to the edge and delivering, it's going to take a collective effort to build that out.

Kunal Dogra: The other big trend that I'm really excited about, you know, almost putting aside public cloud and AI, is private cloud. And Kyle, you spoke about Switch. A lot of organizations rushed into the public cloud. They got into Azure, they got into AWS, and then they realized, “Wait.” They actually needed to bring back the cloud under their own control. So that's what we refer to as private cloud. And so, solutions like Switch which provide that kind of last minute edge technology, I think, are the key to the future.

Peter Antoszyk: Kyle, Kunal, this has been really interesting. This is an exciting area of investment and development. What would you recommend to our listeners to read or follow to stay on top of the development of AI and digital infrastructure, should they have more interest to get deeper into it?

Kunal Dogra: What I'm reading right now, and it's really a great read that I recommend, is Digital Empires. It's a book by Anu Bradford, and it looks at our digital infrastructure and generally tech is going to collide in the future with government. And in contrast, the approach taken in the US to regulating tech versus China and how China intends to regulate tech, as well as how Europe does. So, it's really a fascinating read, and it touches upon a lot of the implications for digital infrastructure.

Peter Antoszyk: Kyle, what about you?

Kyle Colvin: That’s a great recommendation, I think ultimately you know with us, we're really trying to have our ear to the ground and listening to our customers. One recommendation that I would have that touches on some of the power consumption, I thought there was an excellent article in the Washington Post earlier this month by Evan Halper that provides a great overview of how AI is leading to explosive demand. And pushing an aging power grid to the brink, leaving utilities and regulators scrambling for solutions. I'd encourage listeners who haven't had a chance to read that article to check it out. So, I think it provides a great overview for some of these challenges. So, if you don’t have time for the whole book, that Kunal mentioned, I think that's a great article. Fundamentally, we're just seeing, you know, and will see, a power crunch. The power grid, which is the network of power stations and transmission lines that move electricity, it's aging. It's not going to be able to satisfy the demand of increased use going forward.

Peter Antoszyk: Yes, thank you. This has been really fascinating. Appreciate your participating on Private Market Talks. Thank you.

Kunal Dogra: Thank you so much for having us. And Kyle, thank you for joining us today.

Kyle Colvin: Thank you for your time.

Close