Apollo's Slok finds no AI payoff yet outside tech

The AI boom has a receipt problem, and Apollo's chief economist just put it in five charts.

Torsten Slok, chief economist at Apollo, wrote in an Oct. 24 note that AI has so far shown no financial payoff for American companies outside the tech sector.

Five charts, one flat line

Slok's headline claim is blunt. He writes that "the technology has delivered zero financial payoff to corporate America outside of tech."

His evidence is a set of charts drawn from Bloomberg, Macrobond, the Federal Reserve Bank of San Francisco, the Bureau of Labor Statistics and Ramp.

Revenue per worker, adjusted for inflation, is climbing for the largest tech companies. In his words: "Revenue per worker is rising for the Mag 7 but is flat for the S&P 493."

Profit margins outside tech show no rise. Productivity growth, measured with the San Francisco Fed's utilization-adjusted series, shows no AI lift. And AI spending, per Ramp's data, is concentrated in a small number of firms, while the Ramp AI Index shows circularity in token spending too.

His conclusion: "The bottom line is that the AI story will be decided in the S&P 493, not the Mag 7."

The note publishes charts, not a table of figures, so read the charts themselves before you quote a number from them.

The bill is visible, the return is not

Yes, productivity data lags new technology. Slok's note does not argue AI will never pay off. It argues it has not yet, and that the place to look is the 493 companies that buy AI rather than the seven that sell it.

That matters to you because the companies that buy are where your budget lives. When a Wall Street economist says the payoff is missing, finance teams read it. The question in your next budget meeting gets sharper: what did last year's AI spend return?

It also explains the mood. TopFive's top story today is a price war between OpenAI and Anthropic. Buyers who cannot yet show a return push hard on price.

How to answer the chart in your own company

Do not argue with macro data using anecdotes. Answer it with your own numbers.

Pick one workflow where AI does real work: support tickets closed, code merged, documents reviewed. Measure output per person before and after, in hours or dollars, with the same quality bar. A small, clean number beats a big, vague one. Your own evals tell you whether quality held.

Then cut what does not pay. Seats nobody uses are exactly the spend this chart is pointing at.

If you sell AI, the same applies with the roles swapped. Your customers will ask for the number. Have case studies with before and after figures, not adjectives.

Watch the next earnings season

Watch revenue per employee and margin commentary when non-tech companies report their next quarter, and whether any of them credit AI by name with a number attached.

Show the receipt, or lose the budget.