Apollo Economist: AI Hasn't Paid Off Outside Big Tech

For all the money going into AI, the big payoff still seems stuck inside the companies that sell it. That's the blunt takeaway from a new note by Torsten Slok, chief economist at the investment firm Apollo Global Management. I spotted it via TechCrunch's post on X, then went to Apollo's original note.

What happened

In a Daily Spark note dated October 24, Slok compares the "Mag 7" (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla) with the other 493 companies in the S&P 500. His charts show revenue per worker and profit margins rising for the seven giants and staying flat for everyone else. He also says there's still no sign of AI lifting productivity growth.

His wording is strong. Outside tech, he writes, AI has delivered "zero financial payoff" to corporate America. He adds that mainstream companies are starting to see token spending as a cost that eats into margins, not a growth driver.

The numbers behind the charts, as reported by Stocktwits on Yahoo Finance: revenue per employee of roughly $780,000 for the Mag 7 against about $525,000 for the S&P 493, and margins near 30% versus around 10%. The same report says the top 10% of firms account for 99.5% of spending on model serving and inference. TechCrunch and Bloomberg also covered the note.

Why it matters

Slok's argument is that the real AI test happens in ordinary companies like banks, retailers and manufacturers, not at the chipmakers and cloud giants. He doubts cheaper models will be enough, because, in his view, most business decisions come down to judgment, trust and accountability, which AI doesn't handle.

A fair bit of caution is due. These are US figures for the biggest listed firms only, and the charts show what happened over time, not what caused it. The Mag 7's rising margins also reflect layoffs, pricing power and cloud sales, not just AI. And flat averages can hide companies where AI is working and others where it isn't. Slok is one economist, writing a short note, not a full study. Then again, TechCrunch points out that other recent research hasn't found big productivity gains from AI either.

What to do

• If you're paying for AI tools at work, put a number on what they actually save or earn. If you can't, that's useful to know before renewal.
• Don't assume the newest, priciest model is the right one. Slok's own point is that most businesses don't need it, and smaller models that run on a single laptop keep getting better.
• If you invest, remember the AI story so far has paid off mostly for a handful of giants. That concentration cuts both ways.

I'll keep an eye on whether the next round of earnings from non-tech companies shows anything different.