On October 24, 2026, Torsten Slok, chief economist at Apollo Global Management, published an analysis in the firm's Daily Spark column titled "No Signs of AI Payoff Outside the Tech Sector." His conclusion is blunt: leaving the tech giants aside, artificial intelligence has yet to deliver a quantifiable financial payoff for American companies.
The Magnificent Seven and the other 493 are on different tracks
Slok splits the S&P 500 into two groups: the seven large technology companies on one side, and the other 493 banks, retailers, manufacturers, insurers and health-care firms on the other. Revenue per employee and profit margins are rising for the first group and flat for the second. The San Francisco Federal Reserve's productivity series shows no sign that AI has lifted overall productivity growth either. In other words, the companies selling AI are making money, while the companies buying it have not yet seen a matching return in their financial statements.
Slok also writes that mainstream companies have started tightening AI spending: token spending is turning from a growth investment into a cost item that eats into margins. Some firms no longer feel they must use the latest model, and are unwilling to pay a premium for it. As long as agentic AI and token spending stay confined to a small part of the economy, AI's macroeconomic effect will remain limited.
Why this verdict lands hard in this earnings season
The backdrop is an asymmetry. Tech giants keep pouring money into computing power and data centers, and cloud providers' future capacity keeps getting booked up in advance. Meanwhile, most companies' financial statements still show no separately identifiable AI revenue. The money flows to the firms selling compute and models, but the payoff has not spread to the firms using them. That is the mismatch Slok wants to expose: the industry is paying for AI, but most of those paying have not yet earned money from it.
Where this verdict stops short
Leave room for the objection first. Quarterly profit margins are a poor yardstick for a technology that plays out over many years; electricity and information technology also took more than a decade to show up in productivity data. Seeing nothing yet does not mean nothing will come. Looking only at reported margins can also miss the convenience consumers gain and the small new firms that appear, neither of which turns directly into the 493 firms' profits.
But for non-tech companies, Slok's warning is immediately usable: do not use token volume to prove AI's value; watch revenue per employee and profit margins instead. Where the latest model is not needed, do not pay for it. Treat token spending as a cost item to be managed, not as an input that will automatically deliver growth. Only when more companies can state separately in their financial statements how much revenue AI brought in and how much cost it saved will the argument about whether AI is worth it gain new evidence. The evidence available now says just one thing: the payoff has not yet spread beyond the tech circle.
