S&P Global has warned that the artificial intelligence-related liabilities of the world's largest technology companies may be far larger than what has been disclosed, though the agency currently has no plans to adjust its ratings.
Mark Habib, S&P Global's head of digital infrastructure, said at a roundtable in London that hyperscalers may be carrying substantial undisclosed AI-related debt, mainly stemming from implicit guarantee arrangements for data center construction or smaller AI companies. He said:
"It's reasonable to conclude that there is more (debt) that we can't see."
His remarks came amid growing market caution over the financing outlook for AI infrastructure. Bonds recently issued by AI infrastructure providers and so-called "neocloud" companies such as Nvidia-backed CoreWeave have continued to decline in price, reflecting investor concerns about their aggressive borrowing.
Hidden Liability Exposure Difficult to Quantify
Mark Habib specifically highlighted the "interconnected financing structures" between hyperscalers and AI projects — arrangements in which large technology companies effectively provide guarantees for major data center construction or smaller AI firms, creating potential off-balance-sheet liabilities.
"If these liabilities are not disclosed, they will not appear in hyperscalers' disclosures," Habib said. "If the potential transactions these liabilities could generate fall outside our rating scope, we may also be unable to capture the relevant situation."
He said there is currently no specific estimate of how much additional debt these circular financing structures could bring to the balance sheets of the companies involved, but S&P is working to conduct a comprehensive assessment from a holistic perspective.
Capital Expenditure of Six Major Hyperscalers Is Massive
S&P expects that the six major U.S. hyperscalers — Amazon, Microsoft, Alphabet, Oracle, Meta and SpaceX — will collectively spend more than $7 trillion on data centers and other AI-related capital expenditures between 2025 and 2030, and will remain in a state of negative free cash flow at least through next year.
Mark Habib said the biggest risk facing hyperscaler ratings remains whether expected earnings can be realized on schedule.
In July of this year, S&P downgraded Oracle's credit rating to BBB-, the lowest investment-grade tier, citing excessive AI spending and "uncertainty in its path to profitability."
No Pressure for Rating Adjustments Yet, Focus on Sustainability of Support Scale
In response to recent market signals of a selloff in AI-related bonds, Mark Habib said S&P currently sees no major divergence between market pricing and ratings that would require a forced correction.
"At present, we do not believe there is a significant disconnect that would force us to adjust our ratings," he said, explaining that S&P's rating methodology takes a longer-term view.
However, he also made clear the conditions that could trigger a rating review: once the existing scale of spending is deemed unsustainable, the rating stance could change accordingly. "This could come from a variety of factors — AI companies further increasing their investment, or competition in the industry becoming so intense that companies' revenue growth is affected."
Comments