Summary
AI data center operators and the hyperscale tech companies that lease from them raise hundreds of billions of dollars by selling bonds backed by lease payments and facility assets. These bonds come in two main forms, asset-backed securities (ABS) and commercial mortgage-backed securities (CMBS). Together they reached roughly $26 billion in U.S. issuance in 2025 and are on track to exceed $60 billion a year by 2028. CREFC E-Primer, Jan. 2026, Morgan Stanley, July 2025 Most transactions are sold as private placements under Rule 144A, not registered with the SEC, and a definitional gap in Regulation AB restricts eligibility for registered offerings. 17 C.F.R. § 229.1101(c)(1), Securitization analysis, Regulatory analysis The largest single AI data center financing to date is a $27 billion SPV structure tied to Meta’s Hyperion campus, and the largest CMBS deal is a $3.5 billion floating-rate transaction for QTS centers. The market now faces fresh scrutiny from lawmakers and regulators asking whether this fast-moving, lightly regulated financing could pose a systemic risk.
What is an asset-backed security and how does it apply to AI data centers?
An asset-backed security is a bond whose payments come from a specific, isolated pool of financial assets. Under Regulation AB, an asset-backed security means a security that is primarily serviced by the cash flows of a discrete pool of receivables or other financial assets, either fixed or revolving, that by their terms convert into cash within a finite time period, plus any rights or other assets designed to assure the servicing or timely distributions of proceeds to the security holders 17 C.F.R. § 239.44, 17 C.F.R. § 239.45, 17 C.F.R. § 229.1101(c) When an AI data center secures long term leases with investment grade tech tenants, those lease payment streams are the receivables that can be securitized. The operator (or an affiliated special-purpose vehicle, SPV) sells bonds to investors and uses the lease payments to pay interest and eventually principal.
The operator benefits because it can raise large sums at a lower interest rate than corporate debt, often without adding the full obligation to its balance sheet. The investors get a bond backed by hard assets and a predictable cash flow from a tenant they may already view as creditworthy. The bonds are structured with credit enhancement, such as over collateralization or a reserve account, to earn investment-grade ratings.
The statutory definition in the Exchange Act, added by Dodd-Frank, is broader than Regulation AB’s definition. It covers managed pools and master trusts, structures that do not fit neatly inside Reg AB. 15 U.S.C. § 78c(a)(79), SEC Concept Release on RMBS Disclosures § IV.A, SEC Concept Release, SEC Concept Release, Release Nos. 33-11391, 34-104102 That mismatch is why almost all AI data center ABS deals choose the private placement market.
What are the main legal structures for AI data center securitizations?
ABS structure
In an AI data center ABS, note issuance is typically uses a master trust structure where the issuer includes in the security package the equity interests of wholly owned subsidiaries established solely to hold the AI data centers, along with the related tenant leases and cash flows. ARC Ratings, Oct. 2024 The SPV issues notes, and investors receive payments solely from the lease collections. The structure is designed so that even if the parent operator gets into financial trouble, the SPV and its cash flows stay insulated.
Most AI data center ABS use a master trust. That allows the sponsor to add new qualifying AI data centers to the trust and issue additional notes over time, subject to the deal documents. Diamond Hill, Mar. 2026 The master trust creates a scalable funding platform that matches how a multi asset operator grows.
A central feature is the anticipated repayment date (ARD), a soft maturity 5 to 10 years after issuance. Before the ARD, the notes pay only interest. After the ARD, the interest rate steps up according to a preset schedule to encourage the issuer to refinance. The final legal maturity can stretch to 25 or 30 years. Diamond Hill, Mar. 2026
CMBS structure
CMBS deals are backed by a mortgage loan on the AI data center real estate, not by direct ownership of the operating company’s equity. The loan is typically single asset, single borrower (SASB) for a large hyperscale facility. The loan has a hard maturity of around five years. Borrowers cannot prepay, but they can substitute a portfolio of U.S. Treasury bonds that replicate the remaining cash flows, a process called defeasance. Diamond Hill, Mar. 2026
ABS versus CMBS for AI data centers, a quick comparison
| Feature | ABS | CMBS |
|---|---|---|
| What backs the bond | Lease cash flows and equity interests in the property-owning entities | A mortgage loan on the real estate |
| Collateral pool | Often a master trust, can add properties | Fixed pool, typically a single asset or small portfolio |
| Maturity structure | Soft maturity (ARD) at 5 to 10 years, final at 25 to 30 years | Hard maturity at about 5 years, extendable with options |
| Prepayment | Not typical | Not permitted, but defeasance is allowed |
| Issuer flexibility | Can issue multiple series from one master trust | Each issuance tied to a specific mortgage loan |
What regulations govern AI data center securitizations?
SEC Regulation AB and the Exchange Act
Regulation AB sets the SEC’s disclosure and registration rules for asset-backed securities. To register a public ABS offering on Forms SF-1 or SF-3, the security must meet the Reg AB definition. 17 C.F.R. § 229.1101(c) That definition excludes many structures that actively manage the asset pool or use master trusts. Most AI data center ABS rely on master trusts because they want the flexibility to add new properties. Therefore, they fall outside the definition and cannot register publicly.
The Exchange Act’s ABS definition is broader. It includes a collateralized mortgage obligation, a collateralized debt obligation, a collateralized bond obligation, a collateralized debt obligation of asset-backed securities, and a collateralized debt obligation of collateralized debt obligations. 15 U.S.C. § 78c(a)(79) Any security meeting that definition is subject to credit risk retention rules, conflict-of-interest prohibitions, and other Dodd-Frank requirements. But it still cannot use the SEC’s registered ABS shelf forms unless the SEC changes its rules.
The Role of Rule 144A
Rule 144A is a safe harbor that allows the private resale of securities to qualified institutional buyers without SEC registration. 17 C.F.R. § 230.144A Nearly every AI data center ABS and CMBS deal is issued under Rule 144A to avoid the Reg AB definition problem. The bonds trade among sophisticated investors but do not have the same public disclosure obligations as a registered offering.
The SEC’s concept release and potential changes
On September 26, 2025, the SEC published a concept release asking for comment on whether to harmonize the Reg AB and Exchange Act definitions. SEC Concept Release S7-2025-04 The proposal focuses on residential mortgage-backed securities but includes 39 multi part questions that could open a path for AI data center ABS to register publicly if the definitions are aligned. Comments were due December 1, 2025. If the SEC ultimately adopts a harmonized definition, the industry might see registered AI data center ABS for the first time. As of the writing date, no final rule has been announced. SEC Rulemaking Activity
Risk Retention
Under Regulation RR, sponsors of securitizations must generally retain at least 5% of the credit risk of the assets they securitize, unless an exemption applies. 17 C.F.R. Part 246 This rule applies to any asset-backed security within the Exchange Act definition, including AI data center ABS issued under Rule 144A. Sponsors in the AI data center market meet the retention requirement through various methods, such as holding a vertical slice of the notes or maintaining an eligible horizontal residual interest.
How large is the data center securitization market?
The market has grown from a single $900 million ABS by Vantage Data Centers in 2018 to a combined ABS and CMBS issuance of roughly $26 billion in 2025. RBC Capital Markets, Dec. 2025, CREFC E-Primer, Jan. 2026
Annual U.S. issuance by year.
| Year | ABS (approximate) | CMBS (approximate) |
|---|---|---|
| 2018 | $0.9 billion (Vantage only) | n/a |
| 2021 | $6.2 billion | $3.2 billion |
| 2022 | $1.0 billion | $0.3 billion |
| 2023 | $5.9 billion | $2.7 billion |
| 2024 | $8.4 billion | $3.0 billion |
| 2025 (est.) | $15 billion | $11 billion |
Source, CREFC E-Primer, Jan. 2026
As of late February 2026, total outstanding AI data center ABS stood at about $40.5 billion and CMBS at about $18.0 billion. Diamond Hill, Mar. 2026 In the first two months of 2026, ABS issuance reached $4.1 billion, putting the year on pace for nearly $25 billion in ABS alone. Diamond Hill, Mar. 2026
Morgan Stanley projects the combined annual issuance could reach more than $60 billion by 2028 ($37 billion ABS and $25 billion CMBS). Morgan Stanley, July 2025 That projection is part of a broader estimate of $2.9 trillion in global AI data center capital expenditures through 2028, of which roughly $1.5 trillion must come from external financing beyond the hyperscalers’ own cash flows. Morgan Stanley, July 2025
Top issuers and notable deals
The most active ABS issuer from 2020 through the third quarter of 2025 was Sabey Data Centers with seven transactions. Centersquare, Compass Datacenters, CyrusOne, and STACK Infrastructure each had six. DataBank raised the most total value at $3.2 billion. CRA Insights, Dec. 2025
A few landmark transactions reflect the market’s scale and the variety of structures.
-
Meta / Blue Owl Hyperion, a $30 billion SPV financing in October 2025, is the largest private credit AI data center deal in history. An SPV called Beignet Investor, co-owned by Meta (20%) and Blue Owl Capital (80%), raised $27 billion in A+ rated debt (anchored by PIMCO at roughly $18 billion) plus about $2.5 billion in equity. The bonds mature in 2049, are fully amortizing, and were priced at 225 basis points over Treasuries. The deal finances Meta’s Hyperion AI data center campus in Richland Parish, Louisiana, a 4 million square foot facility with 2 to 5 GW of power. Meta retains operational control, leases the campus back under a long-term operating lease, and provides a residual value guarantee. S&P said the guarantee was essential to the A+ rating. Global Data Center Hub
-
Blackstone / QTS, BX 2025-VOLT, a $3.5 billion CMBS SASB transaction in November 2025, was the year’s largest CMBS. It backed roughly 10 QTS AI data centers across multiple U.S. markets. The floating-rate loan had a two year initial term with three one year extension options. Demand was so strong that the lower-rated D notes were reportedly oversubscribed by about 23 times. CREFC E-Primer, Jan. 2026
-
Switch (SWCH 2025-DATA) executed a combined $3.5 billion through a $2.4 billion SASB CMBS and a $1.1 billion ABS. Both were designated green bonds. SFA Research Corner, May 2025
-
TeraWulf / Fluidstack / Google, a $3.2 billion senior secured green note offering, finances 522.5 MW of AI/HPC data centers at the Lake Mariner campus in upstate New York. The notes are amortizing over five years, callable after two years, and priced with a 7.75% coupon. Google backstops approximately $1.8 billion of Fluidstack’s lease obligations to support project-related debt financing. The deal has approximately $3.7 billion in contracted revenue. TeraWulf 8-K, Aug. 14, 2025, TeraWulf 8-K exhibit, Aug. 14, 2025, TeraWulf press release, Aug. 14, 2025, TeraWulf 8-K exhibit, Aug. 14, 2025
The market is also expanding internationally. The first EMEA AI data center ABS was executed in early 2024 by Vantage Data Centres, raising £600 million for two properties in Wales. ARC Ratings, Oct. 2024
What do rating agencies look for in AI data center ABS and CMBS?
All three major rating agencies have now published criteria specific to AI data center securitizations.
S&P Global assesses competitive positioning (location, age, power cost, and power efficiency), portfolio diversification or concentration, expected income changes from delinquencies or vacancies, and operating costs including capital improvements. CRA Insights, Dec. 2025 Through Q1 2025, S&P had rated 42 AI data center ABS issuances totaling $16.2 billion. As of December 2024, 84% of S&P rated data center ABS issuances received A level ratings. CRA Insights, Dec. 2025 Among the 69 ABS issuances analyzed by CRA from 2020 to Q3 2025, all senior tranches with S&P ratings were rated A or A-. The highest-rated junior tranches were BBB. CRA Insights, Dec. 2025
Moody’s February 2025 methodology favors newer assets in primary markets, long economic useful lives, efficient operations, stable historical revenue, tenant responsibility for expenses, tenant diversity, strong over-collateralization, and first mortgage liens held by the trust. CRA Insights, Dec. 2025
Fitch issued an exposure draft in July 2025 asking for market feedback. As of that date, Fitch held no outstanding AI data center ABS ratings. The draft asks whether its existing CMBS Large Loan criteria should also rate AI data center ABS and CMBS transactions, the relevance of residual values to an ABS-style transaction, and whether deals with an anticipated repayment date should be allowed higher leverage. CRA Insights, Dec. 2025
Fitch has also stated that it views AI data centers used for AI training less favorably than those used for cloud computing, because AI training tenants have more location flexibility. That could make demand for a given facility more volatile. CRA Insights, Dec. 2025
What coupons and leverage look like
Mean coupon rates on senior AI data center ABS tranches moved with the rate environment. They dropped to 2.87% in 2021, rose to 5.84% in 2023, and settled at 5.46% in 2025. Spreads on senior AI data center ABS tranches over the five year Treasury tightened from 2.56% in 2020 to 1.45% in 2025. CRA Insights, Dec. 2025
Leverage typically ranges from 8 to 12 times operating income, which translates to a loan-to-value ratio of roughly 65% to 70%. RBC Capital Markets, Dec. 2025
What risks and regulatory scrutiny are emerging?
The rapid expansion has caught the attention of lawmakers and financial stability watchdogs. On January 22, 2026, Senator Elizabeth Warren and three Senate Democrats asked the Treasury Secretary to have the Financial Stability Oversight Council (FSOC) examine AI data center financing structures. The letter specifically named the Blue Owl / PIMCO financing of Meta’s Hyperion project. It warned that if AI companies cannot service their massive debt, the losses could spread through interconnected financial institutions. CREFC E-Primer, Jan. 2026, Tax alert
A Federal Reserve Board study found that up to a quarter of bank loans to non bank financial institutions are now made to private credit firms. Major life insurance companies have nearly $1 trillion tied up in private credit. Quinn Emanuel alert The Bank for International Settlements has also expressed concern.
In January 2026, a bondholder lawsuit was filed against Oracle alleging losses tied to its AI buildout. Quinn Emanuel alert
Meanwhile, tech companies have moved more than $120 billion in AI data center spending off their balance sheets in under two years through off balance sheet SPV structures. Quinn Emanuel alert The speed and opacity of this off-balance-sheet financing are the core regulatory concern.
One positive signal is that no public report was found of a credit event, default, or downgrade in this asset class. But the opacity of private credit markets, where many deals are never publicly reported, means that public data may not capture the full extent of credit exposure. Law firm analysis
Key takeaways
- AI data center ABS and CMBS are the two primary securitization tools. ABS uses lease cash flows and a master trust structure with a soft maturity. CMBS uses a mortgage loan on the real estate with a hard maturity.
- Almost all deals are done as Rule 144A private placements because the SEC’s Regulation AB definition does not fit the master trust structure that operators prefer.
- The SEC is considering a definitional change that could permit registered public offerings, but no final rule exists.
- Issuance has grown from $0.9 billion in 2018 to an estimated $26 billion in 2025. Outstanding ABS alone surpassed $40 billion by early 2026.
- Rating agencies generally award A-level ratings to senior ABS tranches, driven by strong lease terms and credit support. Coupons have tightened as the market matures.
- Federal lawmakers and regulators are scrutinizing the systemic risk of this off-balance-sheet financing, especially the large private credit exposures.
- Counsel and sponsors should watch the SEC rulemaking, the FSOC inquiry, and any rating agency methodology changes. Structure deal documents with an eye toward potential registration if the Reg AB definition shifts.
Frequently asked questions
Q:What is the difference between an AI data center ABS and a CMBS?
A:An ABS is backed by the lease cash flows and the equity interests in the entities that own the AI data centers. It typically uses a master trust that can add new collateral and has a soft maturity (ARD) 5 to 10 years out. A CMBS is backed by a mortgage loan on the real estate, usually a single asset, with a hard maturity of about five years.
Q:Why are AI data center ABS not registered with the SEC?
A:The SEC’s Regulation AB, which governs registered ABS offerings, defines an asset-backed security narrowly. It generally requires a discrete, self liquidating pool of receivables. The master trust structures common in AI data center ABS allow active management and adding new assets, so they fall outside the definition. Issuers instead use Rule 144A to sell to qualified institutional buyers privately.
Q:What is an anticipated repayment date (ARD)?
A:The ARD is a soft maturity date, typically 5 to 10 years after the bonds are issued. Until the ARD, the issuer pays only interest. After the ARD, the interest rate steps up, creating an incentive to refinance. The legal final maturity can be decades later. This structure gives the operator a extended period to refinance or recapitalize.
Q:What is a master trust?
A:A master trust is an ABS structure that lets the issuer add new collateral pools (such as additional data centers) and issue new notes over time, all under one trust umbrella. It scales with operators that are building or acquiring facilities continuously.
Q:How do rating agencies view AI data center ABS versus other assets?
A:Rating agencies look at tenant quality, lease terms, location, power cost, and portfolio concentration. They are generally comfortable assigning A-level ratings to senior tranches when a strong tenant with a long-term lease backs the cash flow. Fitch has indicated it views assets used for AI training as riskier than those serving cloud computing, because training tenants can relocate more easily.
Q:Has any AI data center ABS defaulted?
A:No credit event or default has been publicly reported. However, the market is relatively new and has not been tested by a significant downturn.
Q:What is the largest AI data center securitization to date?
A:The Meta / Blue Owl Hyperion financing is the largest single transaction, a $27 billion SPV bond issued in October 2025. Among standard securitizations, the Blackstone / QTS $3.5 billion CMBS and the Switch $3.5 billion combined deal are among the largest.
Q:Are AI data center securitizations considered green bonds?
A:Some are. Switch’s 2025 transactions and TeraWulf’s 2025 notes were both designated green bonds under each company’s green financing framework. The designation depends on the use of proceeds and the sustainability criteria the issuer adopts.
Q:What happens if the SEC changes the ABS definition?
A:If the SEC harmonizes the Regulation AB definition with the broader Exchange Act definition, AI data center ABS master trusts could register publicly for the first time. That would bring more disclosure obligations but also wider distribution and possibly lower funding costs. Sponsors should watch the rulemaking process.
Q:Who are the biggest investors in AI data center ABS and CMBS?
A:Major institutional investors include insurance companies, pension funds, asset managers like PIMCO, BlackRock, and Apollo, and structured credit funds. The strong demand is evident from oversubscribed deals such as the QTS CMBS and the TeraWulf notes.
Subscribe to The Compute Law Brief
The Compute Law Brief is a free weekday newsletter on the law of AI infrastructure across tax, real estate, construction, power, and deals. The big US build markets and federal law. Three minutes a morning. No paywall, and no email gate to read the blog. Subscribe if you want it in your inbox.
Junde Liu, JD, LL.M. (Taxation) candidate at UF Law. Originally published on Compute Law Blog. This article is general information and does not constitute legal advice. Reading it does not create an attorney client relationship. The reader should not act on the basis of any content here without first consulting a licensed attorney in the relevant state. Last reviewed for accuracy May 23, 2026.