Apollo says the credit behind the AI build-out is repricing

Apollo has told clients that the widening in hyperscaler credit default swaps is not a technical, dealer-inventory story. “What the market is repricing is hyperscaler credit fundamentals, namely a debt-financed AI capex cycle with rising leverage, negative free cash flow”, the firm wrote^1,2. Apollo chief economist Torsten Slok is behind the call³.

Briefs Finance summarised the message as a warning that cheap credit for AI’s cloud backers is ending as default-swap costs climb⁵. One outlet took the comparison further, setting hyperscaler AI debt stress against the 2008 housing collapse⁴. The distinction Apollo draws matters: a dealer-inventory move would unwind on its own, while a repricing of fundamentals does not^1,2.

Why it matters to investors

The spending is not slowing. Hyperscaler capex is still expected to reach around 795 billion dollars in 2026⁶. Data centre capital spending grew 92 per cent in the second quarter of 2026, with US and Chinese hyperscalers doubling spending as server costs rise⁷.

The problem is the funding mix rather than the demand. Hyperscaler capex is growing faster than revenue, which puts pressure on free cash flow⁸. Semafor’s review of analyst estimates tracked by S&P Capital IQ puts 66 billion dollars of capital spending at stake across Amazon, Alphabet, Meta and Microsoft, published under the argument that slowing AI development could improve hyperscaler balance sheets⁹.

Private credit is increasingly part of the pipe. The American Federation of Teachers told Treasury Secretary Scott Bessent it is gravely concerned about AI economic risks, arguing that private credit has become increasingly prevalent in financing AI infrastructure, including data centres and chip purchases¹⁰.

What to watch

Two prices now tell the same story from opposite ends. The default-swap spread on hyperscaler debt is the credit market’s verdict on the capex cycle^2,5. The 10-year Treasury yield is the discount rate applied to the earnings that cycle is supposed to produce, and one market commentary framed the question explicitly as whether a 5 per cent 10-year still leaves room to buy technology⁸.

The portfolio consequence is already visible. Wealth Professional argued that AI capex and rising real yields together are reshaping portfolio construction, and that the barbell global capital flows are tracing may reflect a hedge against uncertainty rather than a confident strategy¹¹. For investors the test is whether spending plans survive a higher cost of the debt funding them^1,6.