AI is paying off for big tech but not yet for the rest of corporate America, Apollo's economist says

Seven American companies now bring in more money per worker than at any point since 2019. The other 493 in America’s best-known stock index bring in about the same as they did then.

That gap is the heart of a short note published on 24 October by Torsten Slok, chief economist at Apollo Global Management, a large private-markets investment firm. Is AI paying off? For big tech, his charts say yes. For everyone else, he writes, the technology “has delivered zero financial payoff to corporate America outside of tech.”

Is AI paying off for companies? What Apollo’s five charts show

Slok splits the S&P 500, an index of large US companies, into two groups. The “Mag 7” are Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. The rest he calls the “S&P 493”.

On Apollo’s first chart, Mag 7 revenue per employee, adjusted for inflation, has climbed to about US$780,000, up from about US$565,000 at its low around the start of 2023. The S&P 493 line has stayed at roughly half a million dollars per worker since 2019.

Profit margins tell the same story. On Apollo’s second chart, Mag 7 margins have risen to about 30%, while the S&P 493 have stayed near 10%. A third chart, built on productivity data from the Federal Reserve Bank of San Francisco, shows “no signs of AI boosting productivity growth” across the economy.

The last two charts use data from Ramp, a US company that runs corporate cards and spending software. The top 10% of firms account for 99.5% of spending on running AI models and 95.7% of cloud computing spend. Among the top 1% of AI spenders, 80.4% are backed by venture capital; among median spenders, those in the middle, only 13.1% are.

Why mainstream firms now see AI as a cost

Slok says ordinary companies are “pulling back”. Tokens are the small chunks of text that AI providers charge for. Firms now see that token spending “not as a source of growth but as a cost center that weighs on margins.”

His deeper doubt is about what AI can do. “AI may be a great calculator, but in business, intelligence is mostly judgment,” he wrote. He adds that AI does not understand accountability, trust or culture, which in the end drive revenue.

There are limits to the evidence. This is a short note with five charts, not a peer-reviewed study. Margins also move with interest rates, wages and tariffs, so a flat line does not prove AI did nothing. Ramp’s data covers only its own customers. And the Mag 7 may be gaining partly from selling AI, through cloud, chips and ads, rather than only from using it.

What the AI productivity paradox means for your work

This is a reality check, not a reason to give up on AI tools. Economists have a name for this pattern, the productivity paradox: a new technology spreads everywhere before it shows up in the numbers. The S&P 493 vs Mag 7 gap suggests AI productivity gains are real for the firms that build and sell AI. For many that only buy it, they are still unproven.

If you run a small business or choose software at work, judge an AI tool by results you can measure, such as hours saved or sales won, not by promises. Meta’s US$825 billion in AI commitments has already made bond investors nervous, and its latest results showed AI spending pushing free cash flow below zero.

Why the 493 will decide whether AI was worth it

On balance, Slok’s note deserves a careful hearing. It asks the right question: not whether AI is impressive, but whether it makes ordinary companies better at their jobs.

Our reading is that “zero” is a strong word for an early stage. Companies often need years to rebuild their work around a new tool before the gains appear. Some are already changing: Barclays says AI is starting to play a part as it cuts about 5,000 roles. Job cuts lower costs, but they are not the revenue growth Slok is looking for.

The fair conclusion is a calm one. So far, the gains sit with the giants that build and sell AI. Whether it pays the companies that only buy it is still an open question, and the honest answer today is “not yet visible”.

Slok puts it plainly: “the AI story will be decided in the S&P 493, not the Mag 7.”

The next clues arrive this week, when Microsoft reports results on 28 October and other big tech firms follow. The bigger test is slower: whether margins outside tech start to rise in the quarters after that.