Major cloud companies may face greater AI-related credit exposure than their reported debt indicates because data-center leases, chip purchases and other financing arrangements can create obligations before they appear as conventional borrowings. Mark Habib, S&P Global Ratings’ digital infrastructure sector lead, said the agency is incorporating those commitments into its credit analysis to provide a fuller view of companies’ obligations. Such structures can allow a company to enter a lease or purchase order instead of raising debt directly, helping manage cash flow while shifting when the liability becomes visible.
The credit pressure is already evident at Oracle. S&P Global Ratings cut Oracle’s long-term issuer rating to BBB-, one level above speculative grade, after the company’s AI infrastructure expansion increased leverage and weakened cash generation. Habib said other hyperscalers remained within their rating limits but had less headroom than they did one or two years earlier and were moving closer to downgrade thresholds.
Its analysis identified debt, leases, guarantees, joint ventures and special-purpose vehicles as increasingly important parts of the financing mix, while highlighting monetization, demand durability and overcapacity as credit questions.
Anthropic’s preliminary registration statement illustrates the scale of obligations confronting newer AI companies. The filing says the company has entered into purchase commitments and lease agreements, including leases that have not started, with an aggregate committed spend left blank in the preliminary document; it also says the commitments require payment regardless of actual usage and are substantially related to compute and infrastructure. The filing warns that revenue growth may not be sustainable, leaving the central credit question unresolved: whether future revenue will be sufficiently durable and predictable to support the infrastructure commitments already being made.