AI in action · Markets

S&P Global Says the Biggest Tech Companies Probably Owe More for AI Than Anyone Can See, and Still Won't Change Their Ratings

"It is safe to assume that there is more [debt] out there that we are not seeing," S&P Global's head of digital infrastructure said of the AI build-out on October 19, 2026. Mark Habib pointed to guarantees hyperscalers give to data-centre projects and smaller AI firms. S&P expects six of them to spend more than $7 trillion on AI capex from 2025 to 2030, but sees no need to change their ratings yet.

Oct 25, 2026

Credit rating agencies rarely say out loud that they cannot see all of a company's obligations. Mark Habib, S&P Global's head of digital infrastructure, said something close to it at a roundtable in London on October 19, 2026, reported by Reuters' Marc Jones. Asked about the debt behind the AI build-out, he said: "It is safe to assume that there is more [debt] out there that we are not seeing." The reason is structural. "If it's not reported, then it's not part of the disclosure by the hyperscalers," he said, and if S&P does not rate the transactions where a liability sits, it may simply not have the visibility.

The structures he described are what Reuters called "interconnected financing": arrangements in which a hyperscaler effectively guarantees a large data-centre project built by someone else, or stands behind a smaller AI company, through long-term leases or capacity commitments rather than through bonds on its own balance sheet. Such commitments let the largest technology companies add capacity without adding reported debt, and they spread the financing across lenders, private credit funds and smaller cloud operators who borrow against the hyperscaler's name. S&P has no estimate of how much additional debt this circular financing could add, Habib said, and is trying to take a "holistic" view.

The scale explains why the question matters. S&P expects the six largest US hyperscalers, Amazon, Microsoft, Alphabet, Oracle, Meta and SpaceX, to spend more than $7 trillion on data centres and other AI-related capital expenditure between 2025 and 2030, and to have negative free cash flow until at least next year. Meta reported its first negative free-cash-flow quarter since 2012 on October 21. The agency has already acted once: it downgraded Oracle to BBB-, the lowest investment-grade rung, in July, citing its AI spending and "an uncertain path to profitability," and bonds of neocloud operators such as CoreWeave have traded at distressed levels this month.

Habib's conclusion was nonetheless that nothing needs to change yet. "For now, we don't have the view that there is a major disconnect that needs to force our ratings to change at this point," he said, adding that S&P's ratings are designed to look through market swings. What could change that view, he said, are "further increases in the investments" or "a view that the revenue rate of competition is getting too intense", in other words more spending, or less certainty that it will pay. The biggest risk to the ratings, in Reuters' paraphrase of his remarks, remains whether the expected profits fail to arrive.

For investors the remarks are a caveat on the numbers they will read this week, when Alphabet, Microsoft and Amazon report. Capital expenditure and debt on the balance sheet are disclosed and audited; guarantees and lease commitments to third-party data centres are disclosed less clearly, if at all. When a rating agency says the visible debt is probably not the whole of it, the gap between the two is the part of the AI trade that no earnings call fully answers.

Sources: Reuters (Marc Jones) via Business Day — Hidden AI debt risks cloud tech giants' credit outlook (Oct 19, 2026); iTnews — Top hyperscalers likely to have more AI-linked debt than now known (Oct 2026)