AI and fixed income: Booming demand for data center ABS and CMBS

Artificial intelligence (AI) has captured the imagination and enthusiasm of investors around the globe.  While the market often focuses on AI in the context of equity opportunities, the fixed income market is likely to play an increasingly important role in financing digital infrastructure growth. Data center operators and developers are tapping a range of debt financing sources to fund their AI-related infrastructure investments, including the asset-backed securities (ABS) and commercial mortgage-backed securities (CMBS) markets. We have been finding attractive value in some of these securitizations backed by data centers and other technology infrastructure, although a thorough analysis of both the collateral and the structures of the deals themselves is essential.

AI drives increasing data center demand

Growth in AI applications has supercharged the two more traditional demand drivers of data center capacity: the migration from on-premises computing to the cloud and broader internet connectivity. This increase in data center demand is evidenced by lower data center vacancy rates, elevated leasing activity, and significant pricing power. Data center vacancy rates as a percentage of megawatt total capacity¹ are approaching new lows across nearly all major U.S. geographic markets.

Despite a significant increase in total deliverable power capacity as developers work to fulfill pre-leased contracts with new builds, we expect most markets to remain supply constrained for the foreseeable future. Across the U.S., new leasing activity in the sector grew 82% between 2022 and 2023. In the first half of 2024, leasing volumes reached a record year‑to‑date high of 2.9 gigawatts,² on pace for a 25% sequential increase from 2023’s record‑high annual leasing volume.

We’ve seen higher demand result in data center operators wielding greater degrees of pricing power, particularly in supply constrained markets. Rent per megawatt of electricity capacity has turned meaningfully higher, in part because of generative AI demand. We expect this dynamic to persist, at least in the  short to medium term, within most markets given healthy supply/demand fundamentals.

Data center pricing power supports fundamental credit, but risks remain

The current fundamental backdrop remains supportive of data center operators’ pricing power, primarily driven by a significant structural increase in demand from generative AI applications. While data center supply is optically elevated, tenants are placing a premium on securing space early and reliably, which results in the vast majority of new data centers coming to market having been leased prior to construction.³ Additionally, most data center tenants are either highly creditworthy—think of Amazon or Google—or rely on the data center for mission-critical tasks. This makes tenants less likely to move at lease expiration, adding to the overall credit resilience of the relevant ABS and CMBS deals.

Despite the currently supportive backdrop for data center operators, there are risks to their newfound pricing power that we consider and evaluate. Most notably, we are carefully monitoring for any potential oversupply or demand degradation, whether in a particular geographic market or across the sector. For example, demand degradation might emerge if companies that have been investing heavily in data center capacity and computing resources to meet projected AI use cases were to slow their capital expenditures in response to weaker‑than‑expected AI monetization.

Securitized credit to play a significant role in financing tech infrastructure

We expect ABS and CMBS to play significant roles in financing capital expenditure on data centers and other forms of technology infrastructure. Both ABS and CMBS are forms of securitized credit, which involve a stream of cash flows from underlying secured collateral. Each deal has a defined structure for how those cash flows are allocated, with the highest-quality slices, or “tranches,” having priority on the cash flows generated by the collateral.

The cash flows in a CMBS deal come from payments on the mortgage used to acquire or finance the underlying data center or centers. Data center ABS are backed by property interest in the underlying property, typically held in a “bankruptcy remote” master trust. Both ABS and CMBS structures allow investors to look toward the underlying assets to service their debt rather than relying on the financial health and corporate-level cash flows of the data center operator that sponsors the transaction.

Analysis of underlying asset is essential

Regardless of whether a specific data center securitization is ABS or CMBS, we begin our credit analysis by examining the quality of the underlying asset. This includes the credit quality of its sponsor, the supply/demand dynamics of the geographic location, lease rollover risk, tenancy composition, in-place rents versus both current and projected market rents, power accessibility and cost, and proximity to population centers. We also incorporate environmental, social, and governance (ESG) considerations into our analysis, evaluating factors like power usage efficiency and water usage in cooling systems.

Securitization structure also drives some potential differences in outcomes

After examining the quality of the asset underlying a data center securitization, we review important structural factors, including implied loan-to-value ratio,⁴ debt yield,⁵ callability or other issuer optionality,⁶ equity economics and incentives, cash flow triggers,⁷ and resilience against credit stress scenarios, as part of our determination of an individual investment’s attractiveness in the context of its market price. Ultimately, the intersection of asset quality and structure‑specific features drives our assessment of a specific bond’s value.

Searching for value relative to other types of ABS

We balance the merits and risks of each individual data infrastructure deal across both ABS and CMBS to identify opportunities. In doing so, we assess any trade-offs between collateral quality, securitization structure, and relative value.

For example, data center ABS are structurally similar to whole-business securitizations, which are generally backed by an interest in a company’s primary revenue-generating assets (often franchise fees and royalties). We compare the relative value in data center ABS with that of whole-business securitizations, as well as with that of ABS with more traditional collateral, such as auto or equipment loans, in an attempt to identify the best risk‑adjusted opportunities across the investment universe.

Of course, data centers also require other types of digital infrastructure to function as a linchpin of AI technology, and those other types of digital infrastructure are often also financed in securitized credit markets. For example, we have evaluated ABS structures backed by more novel types of digital infrastructure, including fiber‑optic networks and internet protocol (IP) addresses. Because the securitized credit market is an essential part of the funding ecosystem for AI‑related infrastructure, we expect to see ongoing issuance in the relevant ABS and CMBS markets. We will continue to evaluate and selectively participate in the new deals in this evolving landscape.