Standard & Poor's Global warns that the scale of artificial intelligence-related liabilities of the world's largest technology companies may far exceed the levels already disclosed, but the institution currently has no plans to adjust their ratings.
Spencer Stover, Head of Global Digital Infrastructure at S&P Global, stated at a roundtable meeting in London that hyperscale cloud service providers may have a significant amount of undisclosed debt related to AI, mainly stemming from their implicit guarantee arrangements for data center construction or small and medium-sized AI enterprises. He said:
It can be reasonably inferred that there is more (debt) that we are not aware of.
This statement comes against the backdrop of the market becoming increasingly cautious about the financing prospects for AI's infrastructure. The bond prices of so-called "new clouds" companies such as CoreWeave, which are provided by AI infrastructure providers and supported by NVIDIA, have been continuously declining recently, reflecting investors' concerns regarding their aggressive borrowing levels.
The exposure to implicit liabilities is difficult to quantify.
Mark Habib highlights the 'interconnected financing structure' between ultra-large-scale cloud service providers and the AI project – large technology companies form off-balance-sheet potential liabilities by essentially providing guarantees for the construction of large data centers or for smaller AI enterprises.
"If these liabilities are not disclosed, they will not appear in the information disclosures of hyperscale cloud service providers," said Habib. "If the transactions that may arise from these liabilities are not within the scope of our rating, we may also be unable to be aware of the relevant situations."
He stated that there is currently no specific estimate of how much additional debt such circular financing structures may bring to the balance sheets of related companies, but Standard & Poor's is working hard to conduct a comprehensive assessment from an overall perspective.
Six major hyperscale cloud service providers have substantial capital expenditure
S&P predicts that Amazon, Microsoft, Alphabet, Oracle, Meta, and SpaceX – these six major US cloud service providers – will collectively spend over $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 for at least until next year.
Mark Habib indicates that the biggest risk faced by the rating of ultra-large-scale cloud service providers still lies in whether the expected profits can be achieved on time.
In July this year, Standard & Poor's downgraded Oracle's credit rating to BBB -, the lowest level of investment grade, on the grounds of excessive AI expenditures and "uncertainty in the profit trajectory."
No immediate pressure for rating adjustments; focus on the sustainability of support scale.
In response to the recent market signals of selling off bonds related to AI, Mark Habib stated that Standard & Poor's currently believes there is no significant discrepancy between market pricing and ratings that requires mandatory correction.
"At present, we do not see any significant disconnects that would compel us to adjust our ratings," he said, explaining that S&P's rating methodology focuses on a longer time horizon.
But he also clarified the conditions that could trigger a re-evaluation of the rating: once the current level of spending is deemed unsustainable, the rating stance may change accordingly. "This could result from various factors—AI companies increasing their investments further, or intense industry competition affecting the company's revenue growth rate."
