S&P Warns of Gaps in AI Debt Visibility

S&P Global warned Monday that major technology companies could carry more AI-related financial exposure than observers can currently identify. Mark Habib, its global head of digital infrastructure, highlighted financing arrangements that link large cloud companies to data center projects and smaller AI businesses.

The warning came with two limits: S&P did not quantify the additional exposure, and Habib saw no immediate reason to change ratings. His principal concern remained whether the expected investment returns would arrive.

The financing structure matters

S&P’s earlier published research helps explain the concern. Its August assessment describes a funding mix that includes borrowing, equity issuance, leases and contractual support. Joint ventures, special-purpose vehicles and residual-value guarantees can complicate the task of assessing credit risk across the sector.

That research projected negative free operating cash flow at the six largest hyperscalers in 2026 and 2027. It identified the durability of demand, possible excess capacity and the eventual monetization of AI investment as areas to monitor. Those were forecasts and analytical assumptions, rather than established outcomes.

For readers of company accounts, the practical question is which entity owes the money and which other entity has promised support. A project can have its own borrower while depending economically on a much larger customer. The contractual links matter when assessing how a problem could spread.

A commitment is not always an immediate cash payment

Consider a hypothetical data center financed by a project company. If a cloud customer agrees to buy capacity for several years, the agreement can support the project’s funding. A separate guarantee could create an additional obligation if specified conditions are met. The contracts determine the exposure; their labels alone do not.

Adding every announced project value, customer contract and guarantee would not necessarily produce a meaningful debt total. The same underlying infrastructure could appear in several figures. A careful assessment needs to distinguish direct borrowing, future purchases, leases and conditional obligations, then check for overlap.

That is why incomplete visibility should not automatically be described as a proven accounting failure. Establishing whether a disclosure is deficient would require examining the relevant contract, financial statement and reporting rules. Monday’s warning concerns the difficulty of seeing the complete economic picture.

Revenue is the next test

TTN’s assessment is that the funding question becomes more pressing when capacity takes longer to earn its expected return. A facility still needs customers after construction, and a customer needs a viable business to sustain its purchases. Strong demand announcements are useful evidence, but they do not settle the timing of collections or the cost of delivering the service.

The explanation of circular AI financing examines how investments and customer spending can reinforce one another. The risk is easier to assess when operating demand, financing support and cash payments are examined separately.

Free cash flow supplies another part of the picture. A large capital program can consume cash before the associated capacity generates revenue. Credit analysis must therefore consider liquidity and the timing of obligations alongside the scale of expected future earnings.

Disclosure needs to follow the obligation

A useful credit comparison would identify who must make each payment, when it falls due and what happens if the associated project is delayed. A lease, a guarantee and a bond can expose the same business to different kinds of risk. Adding their headline values without checking overlaps can overstate the burden, while ignoring commitments outside conventional debt can understate it. Liquidity and payment timing matter alongside the total.

What would change the assessment

The distinction between a warning and a rating action remains essential. A rating expresses an agency’s assessment of creditworthiness; a market price also reflects trading conditions and investor positioning. Neither should be substituted for the other.

For upcoming company disclosures, the useful evidence will be the terms and timing of capacity commitments, the financing attached to new projects and progress toward collecting revenue. S&P’s comments sharpen that scrutiny while leaving the size of any additional exposure unresolved.