Securitization: The principles of data center financing

Transcript:

Intro: Hello and welcome to Tech Law Talks, a podcast brought to you by Reed Smith's Emerging Technologies Group. In each episode of this podcast, we will discuss cutting-edge issues on technology, data, and the law. We will provide practical observations on a wide variety of technology and data topics to give you quick and actionable tips to address the issues you are dealing with every day.

Claude: Hello, everyone, and thanks for joining us for Tech Law Talks and the latest in our data center series. Now, if you've been reading the papers or listening to the news, you will be aware of the explosive growth in data centres. And certainly, we've explored some of the reasons and the issues relating to data centres in previous episodes of this series. However, as always, these things need to be financed. And although there are a varying number of means by which you can finance the construction and operation of a data center. One of them is the securitization. This is a form of financing, which is proving increasingly popular, not only for data centers, but also for the digital age generally. So it will probably be helpful if we start by looking at where this market developed in the context of data centers, which is the United States, and how that history has come forward and the drivers behind that moving on. So I'd like to ask my colleague here, Jeff Stern, perhaps if you can give us a bit of background to this and how the markets come into being and how it's growing. Jeff.

Jeff: So, as Claude had noted, there has been a remarkable growth over the last several years and continuing on for the foreseeable future in data center development. There are several drivers of that growth. One of them is the migration of enterprise workloads to the cloud. That's a primary driver. Naturally, another one is the remarkable growth of generative AI and large language models. That area in particular has a dramatic increase in the demand for high-performance computing and storage, and AI is in fact projected to represent roughly 60% of the total growth in data center power demand through 2028. And the last major driver for growth of the data center development area is digitization of business processes, so e-commerce, software as a service, digital content delivery, etc. As Claude noted, the U.S. has been the global leader in data centers. As of January of this year, the U.S. had about 60% of the world's installed data center infrastructure, with Northern Virginia being its principal data center order. There were over 5,000 data centers in the U.S. as of Q4, 2024. There are a number of reasons for the U.S.'s outsized role in the growth of data center. One of the most significant of those is the presence of major hyperscalers in the U.S. Amazon, Apple, Google, Microsoft, Meta, Oracle. There is, in general, stable and affordable energy, robust connectivity, and a favorable regulatory environment. I'll mention, among several other initiatives, the Stargate initiative by President Trump. that seeks to invest a minimum of $500 billion into AI data center infrastructure over a four-year period. It bears mention on a related note that the regulatory environment for foreign direct investment in U.S. Data centers has tightened significantly. So the Committee on Foreign Investment in the United States, CFIUS, is quite focused on data center transactions, particularly if they're located near sensitive sites or involved in critical technologies. Just looking ahead in this market, again, talking about the development of data centers, the top six hyperscalers are due to spend over $380 billion on CapEx this year, and that should rise to about $540 billion in 2029. So very, very significant investments by the major hyperscalers. 2025 has also been a record year sort of migrating in the direction of financing 2025 has been a record year for data set of financing with about 11.5 billion of abs and 7.4 14 billion of cmbs so that total is predicted to double next year so very very rapid growth in the financing.

Claude: Clearly, with those sort of numbers, particularly in the financing need, the developers and operators of these data centers will be looking at all available sources of financing. And as you've already touched upon cmbs and rmbs as commercial mortgage-backed securitizations and residential mortgage-backed securitizations offer some building blocks which have been well developed in the market and established over time for sectors other than data centers but even within the securitization of data centers there's a number of different models of how to go about constructing the financing. And Simon, who is one of my colleagues, Simon Hugo, based here in London. Simon, perhaps you can start off by talking about how securitizations are particularly relevant to the design, build, and operation of data centers.

Simon: Yeah, I mean, at its heart, securitization is a financing method that allows the pooling of assets and income streams to create and fund securities and returns to investors. Now, in the context of a data center, data centers are particularly suited for this type of transaction, particularly given they're real estate centric with regular cash flows. And they'll be familiar to many real estate asset class investors. They have, to a certain extent, income stabilization, which is required for an ABS. And it's also a cost-effective way to refinance existing bank debt, which some real estate assets are particularly reliant on at the moment. We've seen the increased demand for data storage and processing capabilities and the expansion of development of data centers that Jeff's mentioned, really accelerating that growth in the different financing tools and the expansion into different methods. In terms of those methods, typically data center securitizations are structured in one of two ways, either as an ABS or an asset-backed security structure. This is where the debt is backed by the lease revenues from the data centers, as opposed to a CMBS or commercial mortgage backed securitization where the debt is backed by loans to data center operators secured on the data center property. So, you know, here we have two different structures which may be suited for different investors and different operators alike. So from a market perspective, I think Europe isn't as mature as the US, but it's growing rapidly and forecast to grow significantly over the next few years. We've had two significant deals done so far. One's a securitization of two data centers in Cardiff, who recently had a TAP issuance and refinancing in the last few weeks. And another one was a securitization of four data centers in Berlin and Frankfurt. These notably were all data center ABS deals, and we expect to see a lot more European data center ABS coming out of Western European cities, including London, Dublin, Paris, Frankfurt, Amsterdam, and potentially some emerging hubs as well.

Claude: Thanks, Simon. Jeff, you know, given that the U.S. Is leading the way in the data center construction, I think it's reasonable to assume that you're also leading the way in the securitization market for data centers. So what are you seeing over in the United States?

Jeff: It certainly has been an extremely active market over the last several years. You've seen a fair amount of both CMBS and securitization. I would note that securitization, broadly speaking, offers much greater flexibility than CMBS. It typically allows for upsizing and changing in the collateral pools so that increases revenue and leverages costs. ABS also notably includes an incentive to refinance after about five years, but the issuer is not required to refinance until the final maturity date, which is typically 24 or 30 years. By contrast, CMBS is normally a fixed pool and has typically a hard maturity of five years. And so at maturity, you must refinance or face forced liquidation. So the securitization allows greater flexibility both in terms of the duration of financing and not having a hard maturity at such a short period of time, but also allows for growth of the pool through a master trust structure. As you can imagine, securitization of data centers is considered an esoteric asset and is typically executed in a 144A or 482 private placement offering in the U.S. rather than a public offering. An interesting point is that with a 482 placement, not only have I gotten some market feedback that you're seeing larger checks written for 482 placements, which is a little bit surprising. It may actually allow for easier execution, but it also may be easier to share detailed information about the tenants in a 482 deal where there is a more limited pool of investors and where leases may have strict confidentiality provisions that can only be shared with a limited number of people. So that's an interesting point without a 482 placement. Just a few words about the structuring of a securization of data centers. The typical form is a master trust so that leases are subject to eligibility criteria and rating confirmation because as i mentioned you may have a a pool which changes over time and and allows for the addition of additional leases it often employs variable funding notes to align with the dynamic nature of lease agreements and the portfolio again final maturity is typically 30 years, but it generally includes an anticipated repayment date, not a hard maturity of around five years, and there's an incentive to refinance at that time because it's to create an increase of interest at that five-year period. These deals commonly include DSCR performance triggers, that is, debt service coverage ratios. A breach of the ratio can trigger cash reserves, cash traps, and amortizations. For hyperscale data centers, leases are generally long-term and utilization approach is 100%, so DSCR triggers may be less relevant. But for enterprise data centers, with many customers, contracts are shorter, and that creates a repricing and cash flow risk. So I'll just finish on this sort of overview, talking about what the collateral is for securitizations. As Simon mentioned, they are typically leases. There are also separate accounts set up for the SPVs. There are rights under the customer contracts. There may be, if the sponsor owns the property, there may be also a grant of the data center itself so that the rating agencies may be able to look to residual value for purposes of their ratings. There's typically a reserve account as well as a collection account. And there's typically a pledge of the equity of the issuer and the other SPVs, including an asset SPV and a lockbox SPV that are set up. So that's a general overview of the structure and the differences between CMBS and ABS.

Claude: Great. Really interesting point there about the identity of the lessees in there. One of the things when the first data center securitization occurred in Europe, which is the one in Wales that Simon mentioned earlier on, was that it was subject and had approval under the Green Bond principles. That attracted a lot of interest partly because it was novel to see those principles applied in the context of a data center and secondly it prompted the question why that was the case for a data center because I think the environmental and social concerns relating to data centers have been well flagged in the public domain. I mean, obviously, the size of these things makes their construction quite a large-scale undertaking. And so sustainable construction is sort of key, particularly in Europe, of building these things in a way that doesn't impact adversely on the environment. But I think the operation of them probably raises more issues in that, clearly, they take a lot of energy. It's well accepted that the large language models and the AI programs that Jeff highlighted earlier on take a lot of energy to run and operate. So a huge drain on the local energy market. But of course, not only do you need to pull the power in, but you need to be able to transmit the data into the data centers and also, more importantly, the data centers out so there's a large sort of demand for energy in any data center no matter what its size relative to any other activity. As always, the processing operations within a data centre generate a great deal of heat. And while some of that can be addressed by air conditioning, thus adding to the amount of energy consumption in most of the data centres, water cooling prevails. And so water consumption is another environmental impact of large scale data centres. Because of the needs to get close to power and data transmission there is always a debate about whether you have a greenfield site development which is clearly easier for construction or a brownfield development which although more expensive often places you closer to the power sources and the water cooling sources that you need although one would think that these data centers would be great for the local economy, certainly in the construction phase. It transpires that there's limited social benefits when it comes to their operation. A high degree of automation and the fact that the data centers are full of self-maintaining or minimal maintenance technology means that employment is relatively low, really focusing on a number of operators and supervisors, and in many cases, the security around the site. But actually, as areas of high employment, they don't feature large. So in reaction to all this, the approach of making data centers subject to a form of green labeling, whether that's green bond principles or some other generally accepted means of marking them as a problem, environmentally and socially acceptable certainly helps make them more attractive and marketable to investors and helps to allay or maybe dispel the perceptions that they do create large-scale ESG issues. And certainly the use of sustainable energy is something that's being actively pursued by the UK government at the moment in developing an ambitious plan for securitization construction. Which I think really leads us all on to where does this market go from here? We all know that technology is a fast-moving and fast-changing environment. And so where do we see this market going? Increased demand for computing power, yes. But on the other hand, is there a built-in obsolescence? So perhaps we can, all three of us, just think now and share with our listeners the way we see some of the challenges and issues on the market growth of data centers and the securitization of their financing.

Jeff: Simon, you've raised some fascinating points about some of the challenges facing the market. Data center energy issues are clearly front and center, and they raise both supply and environmental issues. For what it's worth, U.S. data center electricity consumption is predicted to more than double by 2030, so the challenges are quite significant. One development that seems to offer promise is the emergence of a small modular reactor market in the U.S. and elsewhere. I know a number of the hyperscalers are actually dedicating significant investment energy and funds to developing that technology and executing on them. The other big issue that I think is worth mentioning, and you touched upon it as well, is water use. Water use is a huge issue because the data centers need to be constantly cooled. Roughly 40% of the energy electricity used in a typical data center is used to chill water. I just note that in June, China began to build a wind-powered underwater data center to address just this issue. And the project is expected to use at least 30% less electricity. So that's an interesting development, not one that probably can be replicated in a lot of locations. But again, an interesting and innovative way to deal with both cooling, water demand and energy constraints. Maybe the last point I'll raise, which I see as a challenge for the marketplace, is a strain on supply chain, critical equipment and skilled labor. However, there's been such a demand that the labor, you see sort of bidding wars for certain leading figures in the space, but also you have many projects that face timelines that exceed three years because of the durations required to order and receive the equipment needed to build these things out.

Simon: I suppose from my perspective, just maybe touching on a European perspective and where there are some headwinds, I think in particular at the moment, there are fewer larger portfolios and some quite marked difference in the concentration risk in Europe as opposed to the US and fewer of the large scale operators in the market. That said, there are many facilities under construction and as sort of M&A activity in the data centre space continues, we expect there to be some more opportunities in the market. Again, the data centre funding has traditionally been with bank debt. Though private credit interest is increasing, which sort of allows for more interest in private securitizations of data centers. One of the challenges with any European securitization is perhaps, particularly in this context, where there are portfolios which might be spread out in different jurisdictions in Europe, is the different number of legal jurisdictions across that portfolio. Now, there are structures that certainly work from a pan-European perspective, but they're not as easy to implement as portfolios within one jurisdiction. I think the improvement in rating methodologies and investor comfort will do a lot to ensure that the market advances and investors get more comfortable with a more standardised product. Now, the structuring challenges within an ABS, particularly from a European perspective, are related primarily to sort of insolvency remoteness requirements, where you have issuers and prop codes. If prop codes are within the group and also have shared services that are introduced in the additional element of risk. Similarly, where you have to move out employees and ring fence different vehicles. It adds to some of the complexity. None of this is insurmountable and has been achieved on the data center securitizations that have been done in Europe. And from a European perspective, I would expect that. That these issues become more standardised in the mitigants that are employed to overcome them?

Claude: I mean, I think at the back of any investor's mind will always be the two trends that we've seen, not just in data centres, but in technology generally in the last 50 to 70 years. One is the risk of technological obsolescence. At the moment, we're seeing a wave of data centres being built which are all pretty much all one vintage but you can already see the signs of newer technology producing newer vintage data centers and that not only produces cheaper and more efficient technology within the data centers but also poses a question as to whether the size and scale of this vintage of data centers will carry on or whether there'll be a reduction in the size of the instructions in which these operations are housed. I suppose the other point is that with technological developments, we also see changes in the future data and computing demands. I mean, clearly data sensors are a reaction to the explosion in the need for data processing, but also a reaction to AI developments. But already we're beginning to see AI models that are less processing intensive require less energy consumption. And on top of that, the technology occurs in both the mechanical space. So, for example, in-chip cooling, which is being developed, will, to reflect Jeff's point about underwater data centers with wind power, reduce the water demand or certainly the water cooling demands. For data centres over time. I think the regionalisation of data centres will also encourage governments of the countries in which they're situated to lobby hard, not only for favourable regulatory treatment to help them develop them as an economic tool, but also to encourage the construction of more efficient data-related infrastructure, be that water or be that energy. And already, to Jeff's point about the small modular reactors, the SMRs, the UK government's paper on the development of its data technology strategy alludes to SMRs being built to power data centres locally to them. And clearly, whilst the hyperscalers may have the financial wherewithal to construct their own SMRs, alternative power sources. A lot will be looking at the government of the region to be able to help create an infrastructure. And I don't think it's any accident that the UK government has very recently announced that the first SMR in the UK will be constructed in Wales, when you only have to look across and see that the first UK data centre that was securitised was also in Wales. So I think there is a lot of potential, but I think there's also a lot of challenges in a fast-moving market, which is based heavily on technology, which is moving as quickly as the demands of the world's population for data processing and the use of technology increases. So i think that brings us to the end of this broadcast uh thank you to everyone who's tuned in to listen and i hope you found it informative both Jeffrey Simon and I are very happy to take any follow-up questions that you may have you can find us all on the Reed Smith website Jeffrey Stern, Simon Hugo or Claude Brown and you can ask us any questions through there and we will get back to you as soon as we possibly can. So this concludes the data center securitization, which is part of our data center series on Tech Law Talks. And I hope it's been both informative and enjoyable. Thank you for listening.

Outro: Tech Law Talks is a Reed Smith production. Our producers are Ali McCardell and Shannon Ryan. For more information about Reed Smith's emerging technologies practice, please email [email protected]. You can find our podcast on Spotify, Apple Podcasts, Google Podcasts, reedsmith.com, and our social media accounts.

Disclaimer: This podcast is provided for educational purposes. It does not constitute legal advice and is not intended to establish an attorney-client relationship, nor is it intended to suggest or establish standards of care applicable to particular lawyers in any given situation. Prior results do not guarantee a similar outcome. Any views, opinions, or comments made by any external guest speaker are not to be attributed to Reed Smith LLP or its individual lawyers.

All rights reserved.

Transcript is auto-generated.