Apollo's Torsten Slok: AI compute costs exceed user benefits

Torsten Slok's series of notes this past week came as Wall Street retreated from the AI sector, with traders doubting how much longer the capital expenditure orgy can go on. His argument centers on the relationship between compute costs and business returns: companies will only pay a high price if it translates to higher revenue growth, bigger profits, or substantial savings—and he sees little sign any of those exist at scale.

Slok extended his warning to the financing side, suggesting a circular dynamic between AI startups and the venture capital firms that invest in them. He also contrasted the Magnificent Seven, where profit margins and revenue per worker are rising, with the rest of the S&P 500, which has seen no such improvement. In his view, Corporate America overall doesn't need the latest AI model—and doesn't want to pay for it.

The broader market context includes massive planned spending by hyperscalers, estimated at $800 billion this year with capex next year seen topping $1 trillion. Meta, planning to spend $135 billion-$145 billion this year, saw the cost of insuring its debt against default touch a record on Thursday. S&P's Mark Habib also flagged 'interconnected financing structures' between hyperscalers and data center projects or smaller AI companies, warning that more debt may exist beyond what is disclosed.