Signal Scanner · AI & AUTOMATION · 19 September 2026

SEC Staff Put AI Data-Centre Bonds Outside the Asset-Backed Securities Regime

A staff interpretive letter of 29 July 2026 places data-centre securitisations outside the Exchange Act's asset-backed definition, lifting Regulation AB disclosure and the 5% retention rule off a $61 billion channel funding the AI buildout. Inflection 2026 to 2028. Exposed: insurers, pension funds, structured-credit investors, sponsors, supervisors.

Produced using Shaping Tomorrow’s six-stage process for converting external change into decision-ready intelligence.
Scanned and filtered by AI. Stress-tested and judged by humans.

The argument about AI finance is an equity argument: whether the capital spending is a bubble, and what happens to the index if it is. The credit side has meanwhile been reclassified. On 29 July 2026 the SEC's Office of Structured Finance agreed, on the facts put to it, that securities issued in data-centre securitisations are not asset-backed securities under the Securities Exchange Act. Regulation AB disclosure and the Dodd-Frank 5% retention rule fall away with the definition. That lands on a market whose outstanding stock went from $4 billion in 2020 to $61 billion this year, issuing paper with a five-year repayment expectation against a final maturity of 25 to 30. Through 2028 the live question is who carries the renewal risk, and on what information.

Signal Identification

A regulatory pivot delivered by interpretation rather than rulemaking. What changed is the classification of these securities, not their credit quality. Outside the asset-backed definition the sponsor sets the disclosure and keeps no mandated retained interest. The exposure is a fast-growing, long-dated instrument funding short-lived assets, bought largely by insurers and pension funds, and assessed on documents the buyer negotiates rather than on a prescribed schedule.

Time horizon: 1-3 years (staff position in force since 29 July 2026; issuance runs through 2026-2028; first anticipated repayment dates on this vintage fall in 2031)
disclosure terms get set202620272028to 2056
Plausibility band: Medium–High
LowMediumHigh
Geographic / Jurisdictional Scope: Primary: the United States, where the staff position, the esoteric ABS and single-asset CMBS markets and most of the buildout sit. Spillover: the UK through Bank of England supervision; the EU-27 and Japan through insurer and pension-fund holdings of US securitised credit; Ireland and the Gulf as hosts of the next construction wave.
PrimaryUnited States
SpilloverUKEU-27JapanGulfIreland
Sectors exposed:
Data-centre developers and landlordsHyperscale cloud operatorsNeocloud compute providersLife insurers and annuity writersPension fundsStructured-credit investorsCredit rating agenciesBank structured-finance desksSecurities and insurance supervisors

What's Changing

On 29 July 2026 the SEC's Office of Structured Finance told Latham & Watkins that it agreed “the fixed-income or other securities issued in data center securitizations of the type described in your letter are not asset-backed securities as defined in section 3(a)(79) of the Securities Exchange Act of 1934” (U.S. Securities and Exchange Commission, 29/07/2026). Alston & Bird set out the consequence: sponsors “are not required to retain a 5% economic interest in the securitized assets under the credit risk retention rules” (Alston & Bird, 03/08/2026).

The request letter supplies the arithmetic. It describes paper that “typically have an anticipated repayment date of approximately five years and a final maturity of 25-30 years”, written at a loan-to-value ratio “not exceeding 70% of the appraised value” at issuance (U.S. Securities and Exchange Commission, 23/07/2026). The market taking that shape has grown, on Barclays figures, from $4 billion outstanding in 2020 to $61 billion year-to-date 2026, and could reach $180 billion by end-2028, roughly 20% of the esoteric ABS market (Structured Finance Association, 23/07/2026).

Three weeks earlier the Bank of England had named the same gap from the supervisory side: “issuing debt maturities aligned with the asset life of AI chips would imply shorter tenors than are typically preferred by most credit investors”, and the scale of the mismatch “will depend on the terms of the debt agreements” (Bank of England, 07/07/2026). Those terms are exactly what the July letter leaves to the sponsor and the buyer to settle privately.

Four clocks running on the same building

HOW LONG EACH CLOCK RUNS, IN YEARS 0 5 10 15 20 25 30 AI chip asset life short and uncertain, no published figure Anticipated repayment 5 years Lease on one financed site 15 years, triple net Final maturity 25 to 30 years refinancing risk starts here OUTSTANDING DATA CENTRE SECURITISATION 2020 $4bn 2026 to date $61bn 2028 projected $180bn 29 July 2026: SEC staff say these are not asset-backed securities, so Regulation AB disclosure and the 5% retention rule lapse. Barclays figures, restated by the SFA.

Sources: U.S. Securities and Exchange Commission (repayment date and final maturity, from the 23 July request letter; the 29 July staff response); J.P. Morgan (15-year triple-net lease); Bank of England (chip asset life described as short and uncertain); Structured Finance Association, citing Barclays Research (outstanding balances and the 2028 projection).

Disruption Pathway

Stage one runs through 2027: issuance accelerates on the cleared path, disclosure is set deal by deal in negotiation, and the buyer base widens from structured-credit specialists to insurance general accounts looking for duration. Stage two is priced well before it arrives. The first anticipated repayment dates on this vintage fall in 2031, but secondary spreads and rating assumptions about re-leasing will move as soon as one large tenant declines to renew. Stage three is the supervisory answer, and it has two available forms: a Commission-level rule that restores a disclosure floor, or state insurance regulators setting capital charges on paper whose underlying contracts they cannot read.

Stress concentrates in three places. The tenant boundary, where a compute contract of a few years sits inside a lease of fifteen (J.P. Morgan, 10/08/2026) and a bond of twenty-five to thirty. The information boundary, where a buyer without Regulation AB schedules depends on what the sponsor volunteers, at a point when “leverage levels will likely increase for developers focused on hyperscale buildouts to be completed in 2026-28” (Moody's, accessed 19/09/2026). And the supervisory boundary, where the Bank of England can name the mismatch but cannot see the US contracts that allocate it. Two adaptations follow. Investors write their own disclosure covenants into subscription documents, making diligence a privately negotiated good that large buyers can afford and small ones cannot. And insurance supervisors, not securities regulators, become the effective gatekeeper through capital treatment.

Why This Matters Now

This lands on insurance and pension chief investment officers, on structured-credit desks, on the treasurers of data-centre developers, and on supervisors who now hold different halves of the same picture. The assumption to revise is that securitised credit carries a regulated disclosure floor. For this asset class it does not, and the buyer's information depends on what was negotiated rather than on what is filed. On the available evidence the instrument is growing faster than the diligence practice around it: $4 billion outstanding in 2020, $61 billion in 2026, and a first refinancing wave five years out. Investment committees should price that gap, set a minimum covenanted reporting package before the next issuance window, and ask their rating providers what re-leasing assumption carries the senior tranche.

Decision-action posture for this signal: Prepare — the classification changed in July and issuance is running now, but the first repayment dates are five years out, so the work is to set covenanted disclosure and re-leasing assumptions into the next deals rather than to exit a position.

Counter-Argument

The strongest objection is that the classification tracks the economics. These are single-asset real-estate financings, not pools of receivables, and the assets are leased before they are financed: one $4.25 billion offering was “100% pre-leased under a 15-year triple-net lease to a high investment-grade tenant” (J.P. Morgan, 10/08/2026). Stefan Hepp's stress test of the AI financing complex reaches the same place from the other side, finding that “a decline in market capitalization does not cause bonds to default, and a data center with a financially sound tenant can service its debt through a tech-sector sell-off” (Chicago Booth Review, 07/08/2026), with first-loss positions sitting outside the regulated banking system.

Both points hold and neither touches the signal. The claim is not that these bonds default; it is that the information on which they are priced has moved out of a filed schedule and into a negotiation. Hepp's own test names refinancing at maturity as the weak joint, and that is the five-year clock.

Implications

This is a durable change rather than a cyclical one, because it alters who supplies information rather than what a building is worth. The inflection window runs from the autumn 2026 issuance calendar to the end of 2028, by which point Barclays expects outstanding data-centre securitisations to reach $180 billion, roughly 20% of the esoteric ABS market (Structured Finance Association, 23/07/2026). Large insurers with in-house structured-credit teams and sponsors with strong tenants gain: they can negotiate the disclosure others cannot. Smaller general accounts, index-tracking credit funds and any supervisor working from published filings lose ground, and so does the Bank of England's ability to measure a mismatch it has already named.

Early Indicators to Monitor

Disconfirming Signals

Strategic Questions

Keywords

Data centre securitisation; asset-backed securities; Regulation AB; risk retention; Section 3(a)(79); anticipated repayment date; maturity mismatch; esoteric ABS; single-asset CMBS; AI infrastructure finance; insurance general accounts; re-leasing risk

Bibliography

Source tiers: Tier 1, governments, regulators and intergovernmental bodies. Tier 2, think-tanks, academic institutes, major consultancies and quality data providers. Tier 3, quality journalism and specialist trade press. Tier 4, vendor, company and practitioner sources, used only as directional corroboration.

How this was made

The Decision-Grade Intelligence Method

Shaping Tomorrow’s six-stage process for converting external change into decision-ready intelligence.

Scan · Filter · Stress-test · Judge · Trigger · Actionable Intelligence

See full methodology → · How we verify → · Track record →

This Signal Scan is one component of a Shaping Tomorrow intelligence cycle on AI & Automation. Every monitored topic also includes monthly Change Trackers that assess what is materially shifting in momentum, and a quarterly Decision Intelligence briefing that converts the cycle into board-level strategic implications. Cycles may also close with Action Triggers, an end-of-cycle synthesis that reads across the three formats and surfaces patterns no single deliverable carries on its own. Continuous intelligence on this topic is available as a Single-Topic or Multi-Topic programme.

Discuss a Tailored Programme

Prepared by Shaping Tomorrow: 19 September 2026