Apollo's chief economist: AI has delivered 'zero financial payoff' to big companies outside tech — S&P 493 margins and revenue per worker are flat

"Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech." — Apollo Global Management chief economist Torsten Slok's latest note plants this week's sharpest thorn in the AI debate: focusing on the "S&P 493" (the S&P 500 excluding the seven tech giants), he finds profit margins and revenue per worker "appear flat," with no evidence that AI is improving productivity^[1].

Slok's data: AI's dividends belong to the Seven alone

Slok's framework deliberately excludes the tech seven — outside Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla, the "S&P 493" comprising the other 99% of the index is the real sample for testing AI's broad benefits. The conclusion is cold: AI has brought no bottom-line benefit to any large corporation beyond tech. Margins flat, revenue per worker flat — between the massive investment and the actual return, no transmission has appeared^[1]. TechCrunch's report presses the key question: "Is this simply a matter of non-tech companies lagging behind?" — technology diffusion typically takes time (the internet and cloud both went through infrastructure-before-applications lag), and whether Slok's data shows "AI doesn't work" or "AI hasn't arrived yet" cannot currently be falsified either way. But the data resonates dangerously with this week's credit markets: if AI cannot improve ordinary companies' profitability within a foreseeable horizon, then the trillions in AI-buildout financing have only the Seven's own capex loop as their repayment source — precisely the other side of last week's Morgan Stanley flow map where "Nvidia takes two-thirds."

Massive investment versus tardy returns: the gap in coordinates

Placing Slok's data against the investment side sharpens the gap: hyperscalers' $7-trillion-plus 2025-2030 capex plans, over $450 billion of AI-linked debt issuance this year, JPMorgan's $3-trillion-plus chip spending projection through 2030 — all of these repayment logics rest on the assumption that "AI will lift economy-wide productivity." Slok's flat S&P 493 margins mean the assumed demand side (corporates paying for AI productivity) has not yet appeared in earnings. The counterarguments exist too: Goldman notes hyperscalers' earnings estimates haven't been revised down since July and datacenter demand remains strong — but that is supply-side profit (selling AI), not application-side return (buying AI); Slok measures exactly the latter. Together, the two datasets form the full picture: AI is making the shovel-sellers rich, while the vast majority digging with those shovels haven't struck anything yet.

What it means for investors: the next test is application-side earnings

The investment implication points to a clear observation post: in coming quarters' corporate earnings, non-tech giants' disclosures of "AI-related revenue" or "AI-driven cost savings" will displace cloud capex guidance as the core variable of the AI narrative. If S&P 493 margins show no improvement over the next two or three quarters, Slok's "zero payoff" upgrades from an economist's observation to the market's consensus pricing — and the impact would reach beyond AI stocks into the whole credit chain financing the buildout. Conversely, any single non-tech giant producing verifiable AI productivity data becomes heavy ammunition for the bulls.

What to watch

First, this week's Google, Microsoft and Amazon earnings for case disclosures of "quantifiable AI returns for enterprise customers." Second, the quarter-on-quarter margin trend of the S&P 493 overall. Third, the marginal direction of enterprise AI-spending surveys (CIO polls and the like). Fourth, Slok's follow-up work — if he breaks the data down by industry, it would supply the first empirical map of the "AI diffusion curve"^[1].

Who is Slok: from 'AI-bubble flag-bearer' to data chaser

Slok is no newcomer to the AI-skeptic camp — the Apollo chief economist has spent two years publishing warnings about AI capex concentration (his chart showing 'AI concentration exceeding the internet bubble's peak' was widely cited), among the earliest systematic voices on AI crowding risk. His 'S&P 493' framework is less a new discovery than a natural extension of his bubble thesis: if AI spending concentrates in the Seven and returns have not diffused to the other 493, the structure's fragility depends not on the technology but on how long the Seven keep funding returns that have not yet diffused. His next data move is worth tracking: the market needs to know how much of the Seven's AI revenue ultimately comes from the other 493's AI spending — if that closed-loop share keeps rising while 493 margins stay flat, the AI economy is becoming an internal circulation — and that is the true shape of a bubble.