Odd Lots - This Is What It Takes to Get a Data Center Financed

Episode Date: December 11, 2025

Data centers are weird things. They're partly real estate assets. They're partly extremely advanced technological products. And they have to find a way to consume a tremendous amount of electricity from the grid -- or they increasingly have their own power plants on site. And beyond that, they've become extremely controversial, with more and more communities pushing back on their development. So how do you get all your ducks in a row when a new project is proposed? Who provides the financing at which stage of the agreement? What are the legal complications that arise? On this episode, we speak with Travis Wofford, a partner at the law firm Baker Botts, who works in the firm's AI practice. We discuss all the intricacies of these projects, the challenges that arise, and how things have changed in this space just since the beginning of the year. Read more:Oracle Earnings May Not Be Enough to Assuage Debt, AI Deal FearsNextEra Shares Fall Amid Push to Move Into Data Centers and Gas Only Bloomberg - Business News, Stock Markets, Finance, Breaking & World News subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at  bloomberg.com/subscriptions/oddlotsSee omnystudio.com/listener for privacy information.

The news doesn't stop on the weekends. Context changes constantly. And now Bloomberg is the place to stay on top of it all. Hi, I'm David Gurra. Join us every Saturday and Sunday for the new Bloomberg this weekend. I'm Christina Rafini. We'll bring you the latest headlines, in-depth analysis, and big interviews. All the stories that hit home on your days off.

And I'm Lisa Mateo. Watch and listen to Bloomberg this weekend for thoughtful, enlightening conversations about business, lifestyle, people, and culture. On Saturday mornings, we put the past week's events into context, examining what happened in the markets and the world. That on Sundays, we speak with journalists, columnists, and key political figures to prepare you for the week ahead. Join us as soon as you wake up and bring us with you wherever your weekend plans take you. Watch us on Bloomberg Television. Listen on Bloomberg Radio, stream the show live on the Bloomberg business app, or listen to the podcast. That's Bloomberg this weekend.

Saturdays and Sundays starting at 7 a.m. Eastern. Make us part of your weekend routine on Bloomberg television, radio, and wherever you get your podcasts. Bloomberg Audio Studios, Podcasts, Radio News. Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Wisenthall. And I'm Tracy Allaway. Tracy, you know, we've obviously been doing a lot about data centers, but one thing that's sort of becoming interesting is they just seem so complicated.

There are so many different moving parts, and I don't mean like technically complicated, although that's part of it. But they really are this sort of weird thing where they're a high-tech thing. They have the most advanced chips, or at least some of them have the most advanced chips in the world. Their real estate plays, and we talk about securitization and so forth. And then they have to figure out all the stuff of how they're going to connect to the grid and maybe they're going to build their own power plant inside the data center, et cetera. There's a lot of moving parts.

Yeah. There's a couple of things here. So, number one, it does seem to be a complex space. And whenever I think about how some of these things are being financed, I get that. What's the meme with the guy standing in front of the board with like all the pay and string connections. It all feels a little circular sometimes, that whole ecosystem. But the other thing is just the sheer scale of the like financing requirement for doing this. So I was reading a Morgan Stanley report. I think it came out over the summer. They were forecasting 2.9 trillion

of global data center spend through 2028. And just for context, total KAPEX spending by all companies in the S&P 500 in 2024 was $950 billion. It's crazy. Yeah. And by the way. The numbers are mind-boggling. Well, the other big question is if you actually stack that up against existing revenue from generative AI, existing revenue is like $16 billion. Yeah.

Right. So there's like there's a bit of a gap there that needs to be filled. Right. Completely correct. The other thing that strikes me is very interesting and tricky is you have these projects and put out this big capital outlay, capital commitments, et cetera. And then the timelines. Like, oh, when do you get connected to the grid? When does that turbine for your insight gas?

generator yet actually get delivered, maybe 2030, et cetera. So it feels like I get the impression you're dealing with very uncertain timelines. And so how you establish that in the language, so everyone feels protected, like what if you could ever get connected? What if anything can happen, right? What are the expectations for when this actually starts to pay off? When does it pay off? What are the obligations of the tenant, et cetera? We don't really know. Anyway, I'm sure it's different. Number one. So many more questions. We're just scratching the surface. Yeah, absolutely. And, you know, there have been some interesting things happening. So a lot of the activity has been done in the private market.

Yeah. But obviously, AI players and companies have a lot of different options. So they could go to securitized markets. And I think we've been seeing some CMBS and ABS deals there. Private credit, I mentioned private equity, bank lending, or they can go the public route into corporate bonds. And so I'm really curious how a lot of these players are choosing between those options. Well, I'm really excited to say we have the perfect guest, someone who is really really at the intersection of sort of every one of the things that these data centers are also at the intersection.

We're going to be speaking to Travis Wofford. He's a partner at the law firm Baker-Bodz. He is the chair of the corporate department based in Houston. He works on all of these things. He can tell us what's in the details of all these agreements. So, Travis, thank you so much for coming on Adelots. Great to have you here in studio. Thanks for having me. What do you give us just a quick overview of what's your role at Baker-Bodz? What do you focus on?

What do you call it a practice area? Is that a late? Is that illegal for a legal term? Yeah. So I'm a deal lawyer. I help people build by finance and sell projects and companies. And so we take multidisciplinary teams and put them together in order to actually make people's dreams of reality is the way I like to explain it.

The idea is somebody comes to us with a very complex project and they need an integrated solution, legal solution to that problem. So you want to build a data center. You've got power. You need people that actually understand the air. the water, the interconnection with the fiber, the interconnection with power, the land, everything else, and then the finance layer on top of that, we provide a one-stop shop for that. Okay, so my question is, how busy are you right now? Are you just getting, you know,

requests flying at you constantly? Yes, we are very busy, which is a good thing and it's the bad thing, right, because your capacity constrained just like your clients are. My husband is a former lawyer, so I know how bad it can be. Yes, but the good thing is they're interesting deals. It's interesting stuff to work on. It's in the news on Bloomberg as well. And that makes it fun. It seems like at Baker-Bots with like the long history based in Houston and all the energy stuff.

I mean, this is all sort of novel to everyone, right? Especially the generative AI data center. Like that aspect is just a couple of years old. Can you talk about when there is a new thing that everyone gets excited about that maybe even a couple years ago no one was talking about? Now, granted, I know data center is a lot longer than a couple years old, but the explosion of interest, how do you build that sort of team that understands all of the dimensions so that you could provide that one-stop shop service? Yeah, so it's a really good question. And, you know, I always go back to the expression. There's nothing new under the sun.

When you actually think about what we're doing, you're building a power project, you're building out a data center or the powered shell or the powered land or the like. People have been doing that before. When you're financing these, we're actually using a lot of the same financing structures. It may not be exactly the same, but it rhymes. So 10, 20 years ago, we were doing whole business securitizations of cell towers. That same securitization technology, you move forward to residential solar. And now you're moving that same thing forward to CMBS and ABS of the data center. So they're very similar.

The rating agencies are very familiar with these things. You have to look at a few different variables and understand them slightly. differently, but particularly energy infrastructure and telecom infrastructure, those are two things that we've been doing for decades. Oh, walk us through the differences then between, say, you know, a cell tower ABS or, I don't know, a solar power ABS versus a data center ABS or CMBS. Sure. So a cell tower ABS, really when you're looking at that, you just have the tower.

It's on land. You may have somebody that's coming along in order to mow the lawn around it. but they don't have much in terms of the actual equipment that's on top of the cell tower. When you have a residential solar securitization, you put the solar panel on top of somebody's roof. You're contracting the cash flows off of that. Again, on the cell tower, you're contracting the cash flows off of that. You pull those together and then you put a bond on top of it. With a data center securitization, what you're doing there is you have the cash flow from the data center lease.

You have the cash flow from other aspects of the business that may be going along with that, like related to the power and some of the additional services. You pool those. You have long-term contracted cash flows. And then a rating agency can rate those as well. But at the end of the day, you're trying to make sure that you have a special purpose vehicle that has all of the assets that it needs within it. You have a separate entity that's handling the billing in the collection, an entity that's

handling the operation and maintenance of the asset. and that's supposed to be a standalone product that can move at least the five to seven years until you get to the anticipated repayment date. And often it's about a 30-year rated final maturity. So I have a question, and it's come up a couple times on previous podcasts, but you mentioned having a special purpose vehicle for this. Why does it seem that so many big tech companies who are in the AI space who are presumably very, very cash rich and profitable. Why are so many of them financing off-balance sheet? They have better things to do with their money. So the return on their invested capital is much better on the tech side than it is on the infrastructure side. Infrastructure is usually low margins, but the benefit with that is

it's long-term, reliable, understood assets, such as, again, if you're doing a large energy project or even the powered shell itself. It's land, it's a building, it's got the fiber connection, it's got the power connection, and that's something that can continue over a long period of time, whereas on the tech side, they make a lot more money and have much greater margins, as we all know. Right. So this aspect of the business, which could be very profitable, but it's stable, it's very predictable. It's like, let's outsource this. Let's have some other entity housed that risk. What are the risks in data center finance? When you think about the other side, not the tech companies, but the lenders to the SPVN, we'll talk about the different structures that that credit

form can take. What are the risks? You know, it's a fairly stable, long-term thing, but how can things go wrong? So it depends on what exactly you're financing, right? So in a securitization and ABS, what you're doing is you are financing the business of this powered shell or turnkey solution. The tenant quality is very important. Do I have somebody that? that has the investment grade tenant quality or diversified pool of high credit quality tenants. Just to be clear, you're securitizing the cash flows from the data center leases, right? Yeah, so exactly. So on an ABS, it's a focus on the business of this data center. The actual lease is in the securitization. And then the hard assets are in there as well.

In a CMBS, you have the lease as well, but it's more focused on the mortgage. And then we have credit tenant leases as well, which are kind of directly to the tenant. That's useful. Thank you. So risks associated with securitization are usually focused on the tenant quality. Yeah. They're focused on the term of the lease. How long is the lease? If you have a 10 or 15-year lease, if you have a shorter term lease, you're going to look a little bit more at, well, is this something that's going to need to be released in that time frame? What is the rate on the lease? If it's a below market lease, it's much less likely that somebody's going to, want to get a new lease and leave you. And then the facility itself is the facility going to be able

to withstand kind of technology risk over a long period of time. So how are people structuring the deals at the moment to compensate for that risk? So you have the time mismatch, you have maybe tendency, rollover risk as well. Are people like adding extra credit protection or how are they making these palatable to investors? So one is the loan to value. You're, you're, you're, your advance rate. So typically, if you're looking for something that's an investment-grade securitization, you're going to have it around 40 to 50%. The structure of the bond itself will have a rated final maturity that's much further out than the lease, maybe 25, 30 years. But you're anticipating that the full principal amount of the bond would be able to be repaid within five to seven

years. But the lease itself is a 10-year lease or a 15-year lease. And then that can get re-opt for another 10 years on top of that. So when you're looking at your cash flows, the cash flows are going to support your ARD prior to even worrying about the anticipated repayment date of the bond prior to even worrying about your release. So one of the things that sort of like people tweet about and people even write about, including us and we've talked about even. This whole thing about GPU life and some of these questions regarding the long-term value of the assets and people like, you know, like to claim, oh, they're not going to be as valuable long-term as people think. Does this come up in your work and like what is, what's

really going on here? Sure. So it depends on what the financing structure is that you're using. So the GPU life itself is more a question on a private credit story or an equity story. On the securitization side, they're more focused on the Turnkey data center itself. All right. The useful life comes up a lot when you're doing the accounting, and that rolls through to the earnings per share of these public companies. You've got a useful life of a few years ago. It was three years for these GPUs. Now some of them are using six years.

If you're doing straight line amortization, over three years, you're taking about a third off. Over six years, it's obviously less than that. And so you have much less amortization depreciation expense on it. That helps your right expert share. This is Tom Keene, inviting you to join us for the Bloomberg Surveillance Podcast. It's about making you smarter every business day. I'm Paul Sweeney. We bring you complete coverage of the U.S. market open.

We cover stocks, bonds, commodities, even crypto, all the information you need to excel. And I'm Alexis Christophores. Bloomberg Surveillance also brings you the analysis behind the headlines. We do that through conversations with the smartest names in economics, finance, investment, and international relations. We do all this live each and every weekday that bring you the best analysis in our daily podcast. Search for Bloomberg Surveillance on Apple, Spotify, YouTube, or anywhere else you listen. On the East Coast, listen at lunch. And on the West Coast, listen as soon as you wake up.

That's the Bloomberg Surveillance Podcast with Tom Keene, Paul Sweeney, and me, Alexis Christophorus. Subscribe today, wherever you get your podcasts. Bloomberg Surveillance, Essential Listening, each and every business day. What's your impression of how private credit lenders have been handling the GPU aging problem so far? Like, are they conscious of it or are people still kind of basing a lot of their expected cash flows off of GPUs that last, you know, longer because they were previously being used in cloud computing or something like that? Yeah. So they're very aware of it. particularly because if you look at the new,

Nvidia had the H-100s and the H-200s that had come out, then it was the Blackwell, the B-100s, 200s, and now Ruben's going to come after that, right? So every two to three years you have these new chipsets, and it's almost like, do I want to be financing these laptops that you've got in front of me for six years or seven years? What's the replacement cycle on that?

So when do I need to get repaid on my debt. The interesting part that comes into it is it's not just one laptop. We're talking about entire data centers full of these things, and that's in the tens, hundreds of millions of dollars. So how are you actually going to get your cash flow back when maybe those aren't going to be used for training the same way? The reality, though, is the useful life of these and the economic life is not just that initial usage. You can use it for other things. So starting out with training,

needing to get training done as quickly as possible, if the data center is well-located in Virginia or another Tier 1 location, maybe there's inference, if it's close to a city or other location. And then for the CPUs, data centers that have those, you can use them for compute and analytics

and other technologies. and use cases that maybe that wasn't what you originally underwrote, but you knew that that was coming down the line. Since you mentioned Virginia, can you talk about where we are in 2025 with data center sighting and picking location? So I know that there's tons around northern Virginia, and I know that there are some huge projects in the middle of nowhere in Texas where they're not going to offend any neighbors or whatever, et cetera.

But what are the big themes that companies are thinking about right now when they think about location? Yeah, so one of the great things about Virginia is connectivity to the subsea cables. So getting to Europe, getting to Africa, getting to other locations, and the rest of the United States. There's a great regulatory environment there in terms of power, in terms of actually building out the infrastructure there and other data centers so that you can have, I hate to use the word co-located, but connectivity within other data centers makes it very attractive.

The problem is it's saturated. It's a very saturated market. And so if you want to build that next large giant data center and you need power or you need water or you need other parts of the infrastructure and you need to do it quickly, you're often going to look elsewhere. That's one of the reasons that Texas and Orcott have become so attractive. Yeah. So one of the things we hear constantly is that chips and financing are not necessarily the big choke points for doing this. It's more the power. And I guess I'm curious, is it really a power shortage problem or is it more the distribution of power is not in the places where these data centers, you know, want to be because of convenience like co-location?

It's... You can say both as well. Yeah, it definitely is both. I will say, though, that with power, it's interconnection. It's getting that actual permission both for the load. So the data center itself is drawing power from the grid and for the generation. And those are two separate interconnections on these larger facilities that you're going to be getting at relatively the same time. And if you can't have both, then you're probably not going to be able to support such a large facility.

So on the generation side, if you're trying to get interconnection, you go through a long process, potentially with FERC and with your ISO or RTO, you know, ERCOT if you're in Texas, to do typically at this scale, like the large data center scale, like a five-year process in order to get that generation approved and interconnected. That's just the approval, five years, and then you have to actually do the connection. Yeah, and then you've got to build it. Wow. The fun part about that is that timeline, a lot of people walk in and they think, oh, I've heard that it's going to be 18 to 36 months in order for me to get this done.

And then a month or two goes by and they get a new date and it's been pushed out. And then a few months go by and they get a new date and it's been pushed out. And one of the analogies we use is like you're waiting in the airport to get on the plane. And they call, I fly United all the time. They're like, oh, it's global services. Okay, global services go forward. And now armed services. Okay, armed services.

And people with young children. And you realize that that first class ticket that has, you know, boarding group one doesn't mean that you're getting on first. Well, you all get there at the same time at least. This is how I can, you know, I don't like waiting there either, but I was like, you know, I'm going to arrive at the destination the same as all these people with children and everything. So it's hot. Anyway, you know, there was a story recently, Corrieve, one of the big neocloud companies sort of lowered a growth forecast because of delay in a third party data center project, which is exactly sort of what you're talking about, these things.

How does that affect the financing this uncertainty of when you can actually plug these things in or when they're going to get approved? because that sounds very frustrating and time is money. And so how does that interact with the credit component? Yeah. So you kind of have three categories in the timeline of one of these projects. You have your development capital that's often equity funded. You have your construction capital. So you get a big construction loan once you've got your permits and your power.

And then after that's been completed, then you do your takeout financing. Okay. That may be the securitization and the like. Has the regulatory, I guess, political environment, does it feel like it's significantly changed in the last six months? So you talk about like, you know, very few people were talking about water. My impression is that the water component specifically is very overrated based on things I'm right. But it obviously is a matter. But, you know, the public is really concerned about water.

The public is clearly concerned about electricity prices. People are showing up to town halls to protest or to voice their opposition to new data center projects in the areas where they live. Does it feel like the environment today when we're talking in December 2025 is meaningfully different than it was at the start of the year in terms of public awareness of all this stuff? So politically, absolutely, right? I think that the community organizers and there are a lot of organizations that kind of make their money based on outrage around these types of things, they've realized that this is a great opportunity for them. And if you look at the sustainability reports of a lot of the hypers and others, they talk about being water positive. It doesn't mean that they're just happy about water. It means that they're actually trying to have a positive impact,

and they'll start putting money into the watersheds and the like, and water replacement. And that actually is really meaningful. And it's a part of the story that I think a lot of the folks that are thinking about the political aspect of this don't realize, which is when these data centers come in and the balance sheets of these hyperscalers come into an area that's largely been overlooked. You know, it's that New Yorker magazine cover where, you know, you're looking from, The East River is from Manhattan and then it's California, right? Exactly. A lot of these areas are underinvested, particularly in infrastructure.

And a hyperscaler comes in and starts saying, well, you know, this water treatment facility isn't sufficient. We need to build it out. That's very meaningful. Part of that interconnection study that we're doing, you're looking at the electric transmission of the area. That is very meaningful. But water for the developer, it's not a problem unless it's actually a problem. What I mean by that is we have a lawyer on, several lawyers on staff that focus on water.

And one of them, for example, knows all 98 water districts in Texas and knows which one you are supposed to go to if you want to get your project done quickly and which one to avoid. That's something that if you mess it up could really affect your timeline. Wait, now I'm really curious. What would make one water destination more attractive versus the one you want to avoid? Is it just regulatory hurdles? or is it like quality of the water? So it's not quality of the water. You know, if it's potable water, it's potable water.

It's both water supply and then water offtake. So one of the issues in oil and gas in Texas that they've had is, where do we get all the water and what do we do with all of this wastewater? It's the exact same thing with data centers. Obviously, they're putting different chemicals into the water in order to put it into the facility, but it's still a chemically treated water that has to be processed before it can go back either into the drinking supply or into, you know, the rivers and streams. And so just back to the political environment, like, what's that doing on the ground today?

This big change that's occurred because everyone's up in arms about all this stuff. How is that affecting some of the project planning that exists today? So I think that the developers have been very mindful in the past, but they're even more so now because of the sensitivity around it. In certain states, they're considering proposals that might involve moratoriums on data. centers. That's not the case in Texas. If you look at what Governor Abbott and the legislature have been doing, they're pushing more data centers, more tax incentives, more infrastructure. I think Virginia and several other places realize how important it is to their economies as well and to their tax base because data centers produce a lot of tax revenue for these states,

which go to help fund schools and the like. Is anyone talking about public-private partnerships in the data center context? I feel like this is something that like comes in ways. people start getting really, really excited about like public-private partnerships for infrastructure build-out, and then you don't hear about it for like five years. And then it comes back. Are we in one of those waves right now when it comes to data center spend or not really? Yes. So I'm thinking about what I can say. So the short answer is, yes, those are being explored. They're both from the perspective of public-private partnerships, but also from the perspective of like the Department of Energy Loan Program Office and the Department of Interior and the others.

Yes, they can provide financing support. And the Department of Energy has been very vocal about what the loan program office offers. The Department of Defense, Department of War, I guess now, also has programs that they can make available. Some of these are actual grants. Some of them are loans. But separate and apart from that, there's loan guarantees. And so you have a backstop by the federal government of a loan that, let's be honest, if you're at 40 percent loan value and you've got investment-grade tenant that's going to be paying the lease longer than the term of the bond. Do you really need it? Probably not, but it can help lower the cost of capital anyways. Actually, say more about that because we did several episodes, or we've done several episodes, about the idea of loan guarantees or particularly the loan programs office. But from the private side, like talk to us about how that functionally turns into, okay, we're going to get a better credit rating or this is going to crowd in private capital as they like to say, what, happens, what is the steps via which the government's role suddenly unlocks this financing?

I think in all credit, when you're trying to underwrite any kind of loan, you want to see what are the cash flows or what's the collateral and am I going to get paid back? What's the likelihood, right? And one of the things that the frameworks I think of are, show me who you walk with and I'll show you who you are, right? Okay, fine. If you have the United States government that is providing a backstop, then I can be confident I'm going to get paid back. If it's Microsoft or Google or another who's providing a guarantee on a project or the loan, then I'm very comfortable that I'm going to get paid back. When you put these types of things together, that means I don't have to charge you an extra incremental amount to protect me from that credit risk. And that extra

money that you keep in your pocket, you can then turn into additional projects. And it's a velocity of money concept. I want to go back to private credit for a second because, you know, as you mentioned earlier, a lot of these deals have been done in the private credit space so far. And companies have a lot of options when it comes to financing. As I mentioned in the intro, they can go the public route, they can securitize, they can go to private credit. And whenever people talk about private credit in the context of AI, they always say vague things like, oh, it provides customizable financing options and stuff like that. What exactly is the attraction of private credit for data centers and AI buildout? So one of the most attractive things that private credit can but doesn't always offer is non-dilutive

capital, right? Equity is dilutive and equity is very expensive. There's that risk premium that's associated with it. Private credit, there is an assumption in the story that because of the extra protections that they receive, they don't need the same equity risk premium. So a lot of them are looking at maybe a 15% IRA. They might charge you 10 to 12% up front, but then because of the moik that basically their minimum return on the capital that they'll get paid or an IRA premium over time in the takeout, then they'll have a catch up on the back end. The equity, they're taking a percentage of what you at the end of the day are going to get. And it's usually a much more expensive piece. Private credit. But what about private credit versus other forms of credit like

bank lending or the public bond market? Well, can you get it? Well, I mean, that's the, I mean, what is, that's the, I don't know. The underwriting standards for private credit can be riskier than what you would have from, you know, a bulk bracket bank's credit desk. This is Caroline Hyde. And I'm Ed Ludlow inviting you to join us for Bloomberg Tech, a daily podcast focusing exclusively on technology, innovation, and the future of business. Every weekday, we bring you the top headlines from the world's biggest tech companies. From finance to defense, AI to entertainment, and from startups to the magnificent seven.

We highlight the latest stories of the people and companies pushing the tech sector to new. frontiers and the politics that shape global tech markets. We do this all every weekday, then bring you the most important conversations and analysis in our podcast. Search for Bloomberg Tech on YouTube, Apple, Spotify, or anywhere else you listen. Join us every afternoon on your commute home and stay ahead of the tech news cycle. That's the Bloomberg Tech podcast. I'm Caroline Hyde in New York.

And I'm Ed Ludlow in San Francisco. Subscribe today wherever you get your podcasts. So when it comes to tenant diversification in data centers, how is that judged and evaluated and priced? Because if I think about who's using a data center, correlation is hard to measure at the best of times. And if I think about, you know, like a new business that's growing and everyone is suddenly using it, it feels like the kind of thing that could end up being very correlated as opposed to diversified. So when you are thinking about larger investment-grade bonds and debt in data centers, normally you are thinking about wholesale data centers where there's one tenant that has that facility and that tenant has investment-grade credit quality.

It's much less likely that you are going to have a pool of non-investment-grade tenants in a kind of co-location-style data center. and then because of that, Fitch or S&P or Kroll is going to give you this investment grade rating. So I don't think that the expectation should be that you're going to get a investment grade rating on a pool of non-investment grade tenants, at least not right now. I think that eventually when you've got thousands of tenants across hundreds of data centers, that becomes much more attractive. It's just not really where the market is right now. We did an episode a few months ago in September, actually, where we're talking to this guy, Don Wilson in Chicago who's setting up a futures trading firm for GPUs and theoretically it could hedge

your GPU costs. As you think about the industry going forward, could you see that being a useful instrument for data center financing? I'm a little concerned about, is the value of this collateral going to say, I want to hedge the GPU exposure. Could you see over time at maturing and that being a valuable thing? love new finance tools. I think that that kind of creative creativity is really important. There's always an opportunity there. The thing I have noticed with the data centers, just with other asset classes, is that there's always a cash flow stream that is not being utilized. And so, for example, if you have a securitization or really any bond, you're not

necessarily going to get credit for everything in it. So your contracted cash flows, if that counter party is an investment grade may not get an advance rate, a loan to value on that. Somebody else should be there in order to take advantage of that and provide you with additional capital that you can then reinvest into development. So last week, we had a big outage at the CME and futures were basically frozen for like 10 hours or something. And this was because of an issue at one of the CME's data centers, which is operated by a company called Cyrus I. is there a reputational or operational risk that either deal structures or investors need to be aware of when it comes to these financing arrangements?

Reputational or in what regard? Well, for instance, if Cyrus 1 has a big meltdown at one of their data centers, which seems to have happened because one of their cooling centers reportedly malfunctioned, is that something that then gets priced into the financing arrangement? or is that something that investors should be concerned about? So these are definitely things that investors underwrite, too. So if you are a blue chip operator and developer, you are going to have a lower cost of capital than you would otherwise. There are a lot of Johnny Come Lately's into data centers over the last two years since AI has really come to the forefront.

The people that have been doing it for a decade, they know what they're doing, and they're very good at it. And that doesn't mean that there aren't going to be problems. Things blow up sometimes. Squirrels, two wires. Yeah. Hopefully not sub-C ones. There are accidents all the time,

but that's part of why you have these pretty incredible engineering studies that are done in order to actually put these things together. And freak accidents happen. And, you know, that's part of the credit risk. But just compared to other sort of real. estate plays, is there more sort of, I guess, operational risk in a data center than an other type, you know, for just some sort of retail distribution facility? I imagine the sort of operational risk isn't going to be as great in something like that. Yeah. So that's part of why

they have these service level agreements within the data centers themselves, where the operator is agreeing to provide an uninterruptible non-intermittent power. That's one of the big things that people focus on and how do the terms of that contract work and what are the backstops. You could have a reserve account associated with it so that if there are payments that are necessary to be made, that there's just cash sitting there ready to go. Normally people are thinking that that would reduce the cash flow on the bond and that would reduce at the end of the day the equity distribution coming from the SPV as opposed to affecting the payment on the bond. That reminds me actually actually, are you seeing a big surge in demand for like data center insurance or for people,

you know, who want to insure not just against operational risk, but cyber risk and things like that? Yes. So one of the wonderful things about any kind of economic activity is that the insurance market is always there to support it. Well, like the lawyers as well, right? Yeah, definitely. Give us a call, right? The reality, though, is there are products that data centers have had for decades. And then there are new risks that are coming up. just because of how AI training works. You can always go to Lloyds of London in order to get a specific policy if necessary.

But for the most part, these are risks that have already been priced in and there are products that are already there. Now, there's talk of, well, there are products that exist in terms of getting an insurance product to handle your technology risk. So does the product work at the end of the day

or does it become obsolete too quickly? That looks more like a bet almost as opposed to a real insurance policy, but it's something that can be underwritten. How valuable is just having a plug into the grid? You know what I'm thinking about those various deals of companies wanting to buy Bitcoin mining operations, for example, and it seems like, you know what? Yeah, maybe you can make a little money mining Bitcoin, but you have access to 24-7 reliable power. That's a lot more valuable than what's going on inside the shell here.

Talk to us just about that value of anyone who has actually. access to power. So powered land is huge. And that is developing, you know, it's existed for a while, but it's becoming more and more important because there are REITs and real estate developers who don't need to have the sophistication of can I do the new direct-to-cooling technology for a powered shell or turnkey solution. I'm just going to have land and I'm going to make sure that there's sufficient power here for you to build your data center or for you to bring your GPUs into. That's attractive. If you listen to the earnings calls of several of the real estate developers, they're moving into that if they haven't already. So their job is they're

just going to secure land that they can know, that they know they can get power to. So they've already got land. They've got the land. So the interesting part is there are so many different projects over the last several decades where they've been working through interconnection studies for the land for you name it project and they're waiting years and years just like everybody else in order to get those done. Well, why use it for this when we can use it for that? And data centers, you can make a lot more money on it. Isn't the obvious solution to the power problem just for the big guys to build their own

power system and maybe spend a little bit more money plugging it into the grid? Again, if political pushback is becoming an issue or if regulators are worried about this, shouldn't they just do it on their own? Is that the straightforward thing to do? The hypers themselves, I think that goes back to what's the best use of their capital. They're not in the infrastructure business specifically. Though I will say, if you look at, I love Amazon as a good example, they sold books. They had the website.

They needed to expand, so they build Amazon web services. They need a distribution. They've got that. Tech companies have been becoming energy companies for years. companies we know had to become tech companies a decade ago in order to just keep up. But now there's this integration that's happening from both sides, both in terms of what's happening in our operations, but the ownership itself. Just going back to powered land for a second. So what you're saying is that there are these developments that have been going on a long time, but they are initially

planned for something else. But that today in 2025, is they're getting closer to when they could be connected, maybe AI is a, you know what, this connection is a lot more valuable for an AI data center. Absolutely. So this is interesting. Because this to me, like, the only reason I go back to this is because one of the things that when people talk about an AI bubble, if there is one, is this idea of crowding out other productive uses of the economy, right? Are there better things in the long term that we could have done with these turbines?

Are there better things that we could have done with these electrical connection systems, etc. And other companies might have been waiting on some piece of gear and the AI data center outbid them. And I'm not going to, I'm not asking you like it was this sustainable, but there are other things that are going to lose out or not have access to electricity that people wanted to do but aren't going to because that wire is more valuable for an AI company. A hundred percent. Think about the actual interconnection cube two and a half years ago before AI became a big deal. there was, I want to say, 2,600 gigawatts in the interconnection queue, and we were expecting 80% of that would never actually be constructed. Only 20% of it would.

And that's pretty typical, even on a going forward basis. What are you going to do with those projects and with that land that has already started in a process of interconnection studies and the like? well, you can shift that to data centers, especially because a lot of that you were producing solar or batteries, wind and other power gen. Now you add on the data center layer to it. And what may not have been economic before now is. Should I build a data center, Joe? Yeah. I've got a grid connection. I've got water. Yeah. There we go. Okay, next project.

Oddlots builds a data center in protected land in Connecticut. Probably not. Actually, so this sort of piggybacks on a question that Tracy asked already, but why not just for, you know, is these companies, particularly the hyperscalers that look out in the environment and there's all these people showing up at meetings, complaining about the water, et cetera, like that. Why isn't the future just entirely behind the meter in Texas where it's like, we're just going to build it all? We're going to have the natural gas plant on site. We're never going to bother with the grid. We're just going to have the plant right there. Why isn't that just the entire future of data center? One of the issues with behind the meter is, what if the data center goes away?

So, for example, what if Mark Zuckerberg one day decides, I'm not doing the Metaverse anymore? And somebody in AI says, I'm not doing AI anymore. Or we move from GPUs to quantum computing or something like that. You want to have that generative capacity interconnected with the grid. grid so that you don't have a stranded asset. Okay. Oh, yeah, stranded assets.

There's a word I haven't heard for it. Yeah, I know. So just looking forward, and I mentioned earlier, Corweave saying, okay, there was a delay and some of its build out. What are the big choke points? And do you expect them to stack up? We recently did an episode with Travis Kavula at NRG, and he was like, I don't know, like the amount of just the sheer amount that we're adding to the grid, like, it's

going to be tough. And I don't know how it's all going to pan out for some of these projects. But what do you see as the sort of big bottlenecks or choke points that you're thinking about in the coming years? So power continues to be the number one bottleneck. I don't think anybody would dispute that. Water obviously is an issue as well. Getting the turbines and getting what you actually need in order to produce power is very difficult. I do think, though, that when you look at how some of these interconnection requests are prepared,

you have five or 10 different people making applications for the exact same project. So it inflates what the expectations are on the number of projects in the market at a given time. When you get rid of all of that extra wash, you wind up with hyperscalers, large enterprises, and other real investment grade or serious tenants give you projects that can get done. That's infrastructure. That's infrastructure grade. And a lot of the speculative assets, those are PowerPoints that probably are not going to get made, particularly if there's any kind of economic shock later. Travis Wofford. Thank you so much for coming on outlaws.

That was great. Thank you. Thank you so much. Appreciate it. Tracy, I thought that was really helpful. Really clarified a lot of things for me. I'm trying to think, like, the point about the sequencing of the financing, I thought was really interesting or very important to help me understand these things is because in my mind I'm like, wow. they're going to lose so much money.

They're breaking ground on all these deals. And then what if they don't actually get implemented because you're waiting forever? And so hearing him describe the sequencing of different financing at different stages, like, all right, well, at least that makes sense to me. Yes. I still feel like there's kind of a mismatch issue here. Go on. Well, it just feels like, you know, you're talking about, A, a technology that has, like, its own upgrade risk, let's say. Like, you know, people are developing new chips pretty fast.

and you don't know when the next one is coming down the line and when you might want to replace all your chips with something else. So that's one thing. Tenancy rollover risk. I know he pushed back a little bit on the diversified point. But again, my understanding is for a lot of the ABS structures, maybe CNBS as well, diversification is part of the proposal. And I think he mentioned it earlier. And that seems difficult to me to accurately measure. If you have a bunch of tenants all, you know, doing something in the cloud, drawing something from.

a data center if there's a big macroeconomic downturn or something like who's to say that they're not all going to renege on their lease at once. The levels of uncertainty I just seem so extreme because you're talking about, okay, there's the economic downturn. There's the fact that maybe a lot of this AI stuff could completely fizzle out and doesn't produce a return even in normal times. Then there's the technological questions. Then there's the operational questions about are you actually good at operating and building a data center. Not everyone who's going to be the same.

And then there's the grid interconnection and all of these things about like reliability of power, et cetera. So it feels like, yes, on the one hand, I'd very much buy that on paper, yes, you know,

this is the 2025 iteration of what used to be cell towers or what used to be rooftop solar or anything else. But with just an incredible, I mean, he mentioned with the cell tower, for example,

you have one guy whose job is to mow the lawn, right? And that's like that's just going round and round the cell tower. That's like the main like operational component. I get the, you know, most of the time the tower is just there, right? And you have to make sure that it's in a safe area, et cetera. But the degree of complexity, of operational complexity, of technological energy complexity for these just seems like exponentially higher. Yeah.

And I keep thinking back to just the sheer scale of it and like the numbers that got thrown around, literally trillions of dollars in the next few years. Well, I'm sure we'll do more episodes on it. There are so many sub-episodes we could do, including the best of the Texas is 98 water districts and how you find. No, no, no. Joe, you know what sub-episode we could do. What? Sub-C. Cables. We got to do, we didn't get into that, but we should do.

That was a pun. That was a pun. Yeah. Well, we should do more sub-cacable episodes. And actually, just in terms of data center citing, like access to, you know, latency risk and where it needs to be. is something we should talk about more. Shall we leave it there for now? Let's leave it there.

This has been another episode of the Odd Thoughts podcast. I'm Tracy Alloway. You can follow me at Tracy Alloway. And I'm Jill Wisenthal. You can follow me at The Stallworth. Follow our producers, Carmen Rodriguez, at Carmen Armin, Dashel Bennett at Dashbot and Kill Brooks at Kilbrooks.

For more Odd Lots content, go to Bloomberg.com slash Oddlots. We have the daily newsletter and all of our episodes. And you can chat about all of these topics 24-7 in our Discord. Discord.g.g. slash oddlines. And if you enjoy this conversation, if you want us to do a sub-episode on sub-c cables, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes, absolutely add free. All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there.

Thanks for listening. Hello, I'm Michelle Hussein, and for more than 20 years, I was at the BBC. But all the time I was delivering the headlines, I wanted to go further than the news of the day. To spend more time with the people shaping our world. And that's what I'm doing here on this podcast. Speaking to people from Nigel Farage, to love you trying ever so hard. To tech journalist Karaswisher.

And the tech industry is running wild. You know, they've gotten what they wanted and they've seen a huge run-up in their stock prices. This will be a place where every weekend, you can count on one essential conversation to help make sense of the world. So please join me, listen and subscribe to the Michelle Hussein show from Bloomberg Weekend, wherever you get your podcast.

You certainly ask interesting questions.