Outside tech, AI is still paying off for no one in corporate America, says Apollo

“AI may be an excellent calculator, but in business, intelligence is mostly about judgment.” The line appears in a note published on Saturday, October 24, by Torsten Slok, chief economist at the American asset manager Apollo. Its title leaves little room for doubt: “No sign of AI spillovers outside the tech sector.”

Torsten Slok argues that the technology has so far delivered “no financial gain” to American companies outside tech. To make his case, he sets the “Magnificent Seven,” the seven tech giants listed on Wall Street, against the other 493 companies in the S&P 500 index. The note has since been circulating in the American business press, as investors question whether the colossal spending on AI will pay off.

Five charts, one flat line

The argument rests on five charts. The first tracks revenue per employee, adjusted for inflation since late 2019: it rises for the seven giants and stays flat for the other 493. The second shows the same pattern for profit margins. Both curves rely on data from Bloomberg and Macrobond.

The third chart looks at overall productivity in the economy, measured by a series from the Federal Reserve Bank of San Francisco: the author sees no sign of an AI-driven acceleration. The last two use data from Ramp, a corporate payment card company. They show that AI spending is extremely concentrated in a small number of companies, and that AI's “circularity” also shows up in spending on tokens, increasingly driven by venture-backed start-ups.

Torsten Slok draws one conclusion: traditional companies are starting to see their AI spending as a cost that weighs on their margins, not as a growth engine. In his view, the future of AI will be decided at these 493 companies, not at the giants.

What these curves do not prove

The word “no” is strong, and the argument has limits. These curves describe averages: they say nothing about a specific company, job or use. There is also no point of comparison. The seven giants were already very profitable before AI, and the margins of the others also depend on interest rates, wages and tariffs.

Timing matters too. Many large companies are only just emerging from their pilot projects, and the phenomenon recalls the “Solow paradox”: in 1987, economist Robert Solow noted that computers could be seen everywhere except in the productivity statistics. Finally, the data cover listed American companies. They say nothing directly about European businesses.

What business leaders can take away

Above all, the note raises a question that every executive committee will eventually hear: how much is AI earning us? The companies that will be able to answer are those that set a measurable indicator before deploying a tool: time to process a case, error rate, revenue per salesperson. They also track the real bill, including subscriptions and usage, and count the time spent checking what the machine produces.

The debate goes beyond financial markets. The bank Barclays recently cited AI to cut 5,000 jobs. Large companies' upcoming quarterly results will show whether these gains are starting to appear in the accounts, or whether they remain, for now, a promise.