AI shows zero payoff outside big tech so far, Apollo economist says

Torsten Slok, chief economist of the US investment firm Apollo Global Management, argued in a research note dated October 24, 2026, that artificial intelligence has so far delivered no visible financial payoff to most large American companies outside the biggest technology firms. Comparing the seven largest tech companies in the S&P 500 stock index with the other 493, he wrote that revenue per worker and profit margins have risen for the tech giants but stayed flat for everyone else, and that there are still no signs of AI lifting productivity growth.

Companies around the world are spending heavily on artificial intelligence: on chips, data centres, software licences and staff training. The case for all this spending rests on one promise, that AI will make businesses more productive, so that each worker produces more and profits rise. Slok's note asks a simple question: is that showing up in the numbers yet? His answer, for ordinary large companies, is no.

A few terms help explain the argument. Revenue per worker is a company's total sales divided by its number of employees; if AI lets the same staff do more, this number should rise. Profit margin is the share of each rupee or dollar of sales that is left as profit. Productivity growth, measured across the whole economy, is how much more output is produced per hour of work. If AI were spreading benefits widely, all three should start to move. According to the note, they are moving only for the handful of companies that build and sell AI and cloud services, such as the makers of chips and the owners of the largest data centres.

The point about inference spending matters too. Inference is the work an AI model does each time it answers a question or completes a task, and it costs money in computing power. If almost all of that spending is concentrated in the top tenth of firms, most companies are either not yet using AI at scale or are using it in ways too small to change their results. Slok's framing is that cheap intelligence alone is not enough: companies still have to turn it into higher sales or lower costs, and that is a business problem, not a technology one.

There are limits to what the note shows. Flat company-wide figures are consistent with AI having no effect, but they are also consistent with gains that are small, slow to arrive, or held back because many firms are still experimenting. The note does not settle that debate; it describes where the numbers stand today.

For India, the argument is worth following closely. Indian IT services companies earn much of their revenue helping large Western companies adopt new technology, and AI projects are now a central part of what they sell. Whether those client companies see real returns from AI will shape how much they keep spending, and therefore the demand for Indian technology services and the jobs that come with them.

Apollo note: two groups compared (as reported) Measure | Magnificent Seven | S&P 493 Revenue per worker trend | Rising | Flat Profit margin (approx.) | about 30% | about 10% Visible AI payoff | Yes | None yet

Short-term: The note adds to investor questions about whether the huge sums going into AI will earn a return, and may increase pressure on companies to show concrete results from their AI projects.

Long-term: If ordinary companies start converting AI into higher productivity, the figures for the S&P 493 should begin to rise; if they do not, spending on AI outside the biggest firms may slow.

Who is affected: Investors in AI-linked companies; large businesses deciding how much to spend on AI; technology suppliers, including Indian IT services firms that sell AI projects; and workers whose jobs are being redesigned around AI tools.

Questions and Answers