Views from the Floor The Mispriced Debt Powering the AI Boom

A surge in borrowing has created a structural gap in the market for data centre debt.

Technology giants are flooding the bond market to pay for the physical footprint of the artificial intelligence boom. Total outstanding debt for AI-related borrowers, including hyperscalers and cloud providers, is growing at roughly four times last year's pace, according to a Morgan Stanley report in June. To fund new data centres, these companies are issuing everything from corporate bonds to asset-backed securities (ABS).

The mechanics of securitised debt live in the more misunderstood corners of the credit market, but at its simplest, ABS pool together the contractual cash flows that assets generate (in this case, the lease payments from data centre tenants), and uses them to service tradable bonds, with the underlying data centre real estate and equipment pledged as collateral. This year alone, hyperscalers are estimated to spend up to $800 billion on AI infrastructure, which has propelled data centre securitisation into one of the fastest-growing segments of structured credit.

Source: Bloomberg. Data as of 7 September 2026.

Problems loading this infographic? - Please click here

The sheer volume of capital required to build these facilities seems to be changing the shape of the global bond market and it has also exposed a glaring anomaly in how their underlying risk is priced. Investors are currently being paid a spread premium to hold debt secured by hard assets and contractual cash flows versus the unsecured bonds of the very same firms building them.

Structured credit investors may also use data centre ABS as a diversifier, gaining exposure to a sector largely uncorrelated with traditional collateral like auto loans, credit cards or residential mortgages.

The pricing disparity is clearest when comparing three ways to gain exposure to the data centre build. Investors lending to an investment-grade rated hyperscaler via unsecured corporate bonds have been earning tight yields with no structural claim on physical assets and no protection beyond the issuer's general balance sheet. Those lending to newer AI-focused providers (neocloud) via high-yield debt take on materially greater credit risk, again unsecured and without the benefit of any asset collateral backstop.

Data centre ABS offers a third route, benefitting from the same fundamentals as corporate bonds but with added structural debtholder protection the unsecured market doesn’t offer: subordination (junior tranches absorb losses before senior noteholders), overcollateralisation and covenants that protect noteholders if cash flow performance deteriorates. Noteholders also hold a direct claim on the underlying real estate and equipment. Yet, these securities offer bondholders a wider spread than the similarly rated unsecured corporate paper of the firms building this infrastructure.

Source: Bloomberg as at 7 September 2026.

Problems loading this infographic? - Please click here

The comparison is clearest within a single rating band. Since early 2025, AA-rated data centre ABS have offered spreads of 155 to 175 basis points. Unsecured corporate debt with the same rating traded at just 45 to 70 basis points over the same period. That is a gap of roughly 100 basis points, and the higher yield belongs to the instrument with a direct claim on the buildings and the rent.

The pattern repeats further down the credit spectrum, where A-rated ABS routinely yield more than lower-quality BBB- unsecured paper.

We believe the pickup in yield for data centre ABS is the clearest evidence that this pricing gap is a liquidity and complexity premium rather than a credit signal.

As the market struggles to absorb the volume of new supply, buyers demand extra compensation to hold a less familiar, less liquid asset. ABS typically trade less frequently than corporate bonds given the smaller size of the market and fewer participants.

Further, a meaningful portion of the extra spread on offer is attributable to structural complexity, independent of credit risk, where investors are effectively demanding a premium for structural safety the market hasn't yet learned to underwrite.

A distinct pressure point is technical downgrade risk. If a tenant slows infrastructure spending or negotiates a lower rent upon renewal, the data centre's appraised value mechanically drops. That valuation drop can breach loan-to-value thresholds and trigger an automatic downgrade, even if the tenant never misses a single payment. Ratings-constrained insurers are then forced to sell into a thin market, pushing yields higher – a dislocation that is almost entirely technical and detached from deal solvency.

Other potential vulnerabilities include concentration risk and technology obsolescence, which must be underwritten appropriately. Many deals reference a small pool of big tech tenants, so a change in a single tenant's leasing strategy or capex appetite – not necessarily a default – can potentially affect multiple transactions at once. Computer hardware also depreciates faster than the 15-to-20-year real estate cash flow assumptions embedded in many deals, a mismatch that has to be priced in.

Data centre ABS sit at the intersection of one of the most powerful secular themes in markets and have the potential to reward investors who underwrite the underlying tenant, technology and liquidity risks with rigour and scrutiny.

When geopolitical volatility spiked earlier in 2026, a new high-quality data centre deal stalled simply because the broader market had turned risk-off. The underlying tenants’ ability to pay remained untouched, yet the bonds struggled to find buyers in the primary market. Spreads compressed once the market recovered and the bonds traded into much stronger demand.

For managers with the capacity to tactically act as a source of liquidity when others cannot, such episodes may be a significant driver of returns, reinforcing the view that the potential for mispricing, whether in a stalled syndication or in the spread basis, is likely structural rather than one-off.

All data Bloomberg, unless otherwise stated.

Author: Jaime Dede, a Credit Analyst focused on global structured credit investments and Hugo Richardson, a client portfolio manager at Man Group.

For further clarification on the terms which appear here, please visit our Glossary page.