Apollo economist finds no AI payoff outside tech

By VK

Published 26 Oct 2026, 12:21 UTC

Reporting from TechCrunch

Apollo's Torsten Slok says margins and revenue per worker are flat across the S&P 500 minus its seven tech giants, with no sign of AI gains.

What it means for founders

Expect harder ROI questions. If you sell AI to companies outside tech, Slok's note suggests buyers may start booking your product as a cost to trim, so expect demands for proof. Bring a before and after on a metric the customer already reports, not hours saved.
Sell judgment support, not raw intelligence. Slok's case is that cheaper intelligence alone will not move margins. Products that fit into approval chains, keep a named person accountable and log decisions answer his objection more directly than another chat window.
Count the hidden cost. The Berkeley findings suggest AI can add work instead of removing it. Track whether your own team's hours rise alongside output before telling investors AI made you leaner.
Watch the next data. The signal that would answer Slok is margins or revenue per worker in the S&P 493 starting to climb, in later Apollo notes or in company earnings.

The story

Apollo Global Management chief economist Torsten Slok says AI spending has yet to show up in the numbers outside the tech sector. In a Daily Spark note dated Saturday, October 24, and covered by TechCrunch the next day, Slok writes that the technology has produced "zero financial payoff to corporate America outside of tech." TechCrunch notes that Apollo is a former owner of the publication.

How Slok ran the numbers

Slok splits the S&P 500 in two: the seven tech giants he calls the Mag 7, which TechCrunch lists as Nvidia, Microsoft, Apple, Alphabet, Amazon, Meta and Tesla, and everyone else, the S&P 493. His charts show revenue per worker and margins climbing for the giants but flat for the 493, and he sees no sign yet of AI lifting productivity growth.

Slok also says mainstream companies, seeing nothing in profits, have begun cutting back and now tend to view token spending as a cost that squeezes margins rather than a growth driver. Venture-backed firms keep spending more on tokens, he writes, but established firms neither need the newest models nor want to pay for them. Citing Ramp data, he argues the economy-wide effect will stay small while that spending remains so concentrated.

The split puts the AI sellers on one side and their would-be customers on the other. Enki has covered the seller side, including Anthropic's 12x revenue growth; Slok is looking at the buyers.

Why he doubts the gap will close

Slok doubts the gap is simply a lag. Even as models turn into commodities and intelligence gets cheaper, he writes, converting it into faster growth and fatter margins may be hard, because most business intelligence is judgment. He argues AI has no real grasp of fairness, trust, accountability, ethics or a company's unspoken norms, which shape reputational risk and, eventually, revenue and margins.

In February, TechCrunch reported that UC Berkeley researchers who spent eight months observing a tech firm with 200 employees saw AI tools expand workloads rather than shrink them. The same report cited a trial in which experienced developers took 19% longer on tasks with AI while believing they were faster, and an NBER study that put time savings at about 3%. Both of those studies have been disputed.

What we don't know yet

How deep the pullback in token spending runs, since the note's spending charts rest on Ramp data rather than company filings.
Whether gains inside individual companies are real but too small to register in aggregate figures.
How results differ by sector within the S&P 493, since an average can hide clear winners.