Apollo Warns Debt-Funded AI Spending Is Raising Hyperscaler Credit Risk

Apollo Global Management is warning that markets are placing more risk on the debt of the giant cloud companies funding the AI build-out. The signal is a roughly 60-basis-point gap between their credit-default-swap prices and those of banks, up from about zero since October 2025.

The warning came Wednesday from Apollo chief economist Torsten Slok, who argued that pricier insurance on hyperscaler bonds is not simply a technical effect of banks hedging a growing volume of newly issued debt. Instead, he said, investors are reassessing the borrowers themselves.

A signal from the bond market

Credit-default swaps are contracts used as insurance against a borrower failing to repay. Apollo’s test is comparative: if banks’ own spreads had widened alongside those of the cloud companies, dealer hedging could offer an explanation. But Slok said the growing difference instead indicates that hyperscaler credit risk is being priced on its own terms.

Apollo attributes that reassessment to a debt-financed AI capital-spending cycle: leverage is rising, some companies have negative free cash flow, and the eventual payoff is uncertain when the assets being bought depreciate. The point is not that the companies have already failed to meet obligations. It is that investors are demanding more protection against that possibility.

Cash flow separates the companies

The financial picture is not uniform. FactSet figures cited by CNBC show Alphabet, Amazon and Meta with negative forward free cash flow, while Microsoft is reported to have positive forward free cash flow. Their forward debt-to-equity ratios also vary materially.

Reported forward financial measures

• Alphabet: negative free cash flow of $25.7 billion and debt-to-equity of 13%.
• Amazon: negative free cash flow of $30 billion and debt-to-equity of 23%.
• Meta: negative free cash flow of $25.7 billion and debt-to-equity of 34%.
• Microsoft: positive free cash flow of $33.4 billion and debt-to-equity of 7.34%.

Those figures do not establish how any one company’s AI investments will perform. They do show why a single story about hyperscaler debt can obscure meaningful differences in financial flexibility. Microsoft’s reported cash-flow position, for example, contrasts with the three peers cited as negative on a forward basis.

The countercase rests on earnings

Technology investor Paul Meeks offered the opposing view: it is too early to decide how the build-out will turn out, because hyperscalers may not add enough capacity through 2027 and 2028 to make the result clear. He said their margins are beginning to improve, and sustained improvement would reduce the concern around their borrowing.

That is the unresolved next move. Apollo’s warning says credit markets are already charging more for uncertainty around AI infrastructure returns. The more optimistic case says operating performance can still catch up. The decisive evidence will be whether higher margins persist while the capacity funded today begins to produce enough cash to support the debt behind it.