Apollo economist says AI has delivered zero profit payoff outside Big Tech

Seven companies are carrying an AI boom that, by one economist’s math, has yet to show up anywhere else in corporate America. That gap is now the central argument of the AI bubble skeptics.

Torsten Slok, chief economist at Apollo Global Management, published a note on October 24 making a specific, checkable claim: revenue per worker and profit margins have climbed for the Magnificent Seven since the AI spending wave began, and have stayed essentially flat for the other 493 companies in the S&P 500. Slok’s own phrase for it, reported by Bloomberg, was that AI has delivered “zero payoff” to that broader group. Companies outside the Mag 7 are mentioning AI constantly on earnings calls, he found, but the mentions aren’t translating into the numbers that are supposed to follow: faster revenue growth, fatter margins, or real cost savings.

Slok’s framing, also picked up by Fortune, is that mainstream corporate America is starting to treat AI token spending less like an investment and more like a cost center weighing on margins. That’s a blunt reversal of the pitch every enterprise AI vendor has been selling for two years. If the companies buying the tools can’t point to the return, the sales pitch has a hole in it.

Here’s why the distinction between the Mag 7 and the other 493 matters so much right now. Microsoft, Meta, Alphabet, Amazon and Nvidia are major sellers of AI chips, cloud capacity and related infrastructure, while all seven companies are investing heavily in AI. Microsoft reported $41 billion in capital spending in its latest quarter, up 70% from a year earlier, and said an accounting change would put its calendar-year 2026 guidance at roughly $175 billion while leaving the underlying investment plan unchanged. Meta has raised its 2026 capex guidance to $135 billion to $145 billion. Consensus estimates cited by Goldman Sachs put total 2026 capital spending by U.S. hyperscalers at about $800 billion, although not all of that spending is exclusively for AI.

That’s the spending side. Slok’s aggregate charts show no comparable improvement in revenue per worker or profit margins among the other 493 companies. The comparison does not isolate AI’s effect at individual companies or establish that AI caused the gap, but it shows that broad financial gains have yet to appear in those measures.

Why the bond market got nervous first

Slok’s note landed less than two weeks after a sharper, faster scare in the market that funds the AI buildout. In mid-October, a chart about CoreWeave’s debt load went viral on X after venture capitalist Chamath Palihapitiya flagged that the yield on $1.25 billion of the company’s bonds due in 2032 had briefly touched 15%. CoreWeave, which is backed by Nvidia and had about $35 billion in debt plus roughly $16 billion in debt-like lease liabilities at the end of June, became a focal point for worries about AI financing. Its shares fell more than 15% on October 14, while yields on its unsecured bonds subsequently remained near 15%, as Oracle and several other AI infrastructure stocks also sold off.

The CoreWeave scare was about leverage: a company borrowing heavily against future AI income it hasn’t earned yet. Slok’s note is a different and arguably more fundamental worry, about whether the income shows up at all once you look past the handful of companies both building and buying the infrastructure. Put those two stories next to each other and the AI trade starts to look like it’s being carried by a very small number of balance sheets, both on the supply side and the demand side.

Frankly, this is the number every founder selling an “AI productivity” tool into the enterprise should sit with. If S&P 500 companies with finance teams, procurement departments and CFOs under pressure to show ROI can’t produce measurable gains from AI spending yet, that’s a harder sell for a startup asking a mid-market customer to trust the same promise on a much smaller budget and a much shorter sales cycle. It doesn’t mean the tools don’t work. It means the case for them still runs on a story about the future, not a number from the last quarter.

Also read: A $29 Lifetime Analytics Tool and a Free One Showed Up Within Hours of Each Other • Open-weight DeepSeek beats Google’s gated Gemini model at finding bugs • SoftBank’s Son Plans $50 Billion Malaysia Tech Push, PM Says