Latest / Minds, Bodies, and Terawatts / Apollo's Chief Economist Looked at 493 of America's 500 Biggest Companies and Found AI Has Paid Off for None of Them. The Seven Selling It Took the Whole Gain, and Wall Street Spent the Week Dumping the Builders

Transcript

Apollo's Chief Economist Looked at 493 of America's 500 Biggest Companies and Found AI Has Paid Off for None of Them. The Seven Selling It Took the Whole Gain, and Wall Street Spent the Week Dumping the Builders · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, the show where the whole economy bought the tool and seven companies kept the change. Bloomberg reported on Sunday that Torsten Slok, chief economist at the investment giant Apollo, went through America's five hundred biggest companies. Outside of tech, he found, AI has delivered, in his words, zero financial payoff.
• 0:19
  [Guest] He split the S and P five hundred into the seven big tech names everyone calls the Magnificent Seven, and everybody else, which he calls the S and P four ninety-three. Then he drew the same two lines for each group.
• 0:33
  [Host] Which two lines did he draw?
• 0:35
  [Guest] Revenue per worker and profit margins. For the seven, both are climbing, and margins are near thirty percent. For the other four ninety-three, both lines are flat, and margins sit around ten percent, where they have been through the whole ChatGPT era.
• 0:49
  [Host] The whole ChatGPT era, and the line never moved. Who are the four ninety-three?
• 0:54
  [Guest] That is the point of the chart. They are the airlines, the insurers, the grocery chains, the regional banks. They are the customers, and after years of buying the software, their own financial statements can't tell you it happened.
• 1:07
  [Host] Most of us don't work for the seven. We work for the four ninety-three, so if the tool isn't showing up in my employer's margins, what is it doing?
• 1:17
  [Guest] Slok's answer is that for the typical company it has become a cost center that weighs on margins, his phrase, and that mainstream firms are now pulling back. His third chart is productivity for the whole economy, and the broad measure economists watch shows no sign of AI in it.
• 1:35
  [Host] Does he show that pullback, or just assert it? Somebody is still paying OpenAI's bills.
• 1:40
  [Guest] He asserts it, and the note puts no number on the pullback. What it does show is who is spending. Slok pulled corporate-card data and found the top ten percent of firms account for ninety-nine and a half percent of spending on running AI models, and the very heaviest spenders are overwhelmingly venture-backed startups.
• 1:58
  [Host] So the AI economy is startups paying AI labs with their investors' money.
• 2:03
  [Guest] That is the loop he is pointing at. Venture money pays the AI bills, the labs' revenue justifies the next round of data centers, and the company in Ohio that insures your car is barely in the picture.
• 2:16
  [Host] And this landed in a week when Wall Street was already nervous about exactly that.
• 2:21
  [Guest] Thursday the Nasdaq had its steepest drop in months on that exact worry. And since the market peaked earlier this month, Yahoo Finance's tally has the companies building and equipping data centers falling harder than the big tech firms writing the checks.
• 2:35
  [Host] Let me push back on that. Is zero payoff this early actually damning? Electricity took a while too.
• 2:41
  [Guest] About forty years, in that case. Economic historians have shown that factories had electricity for about that long before productivity moved, because every building was laid out around one central steam shaft, and until someone tore out the floor plan, electricity was just an expensive way to turn the old shaft.
• 2:59
  [Host] So the flat line could be the trough, not the verdict.
• 3:03
  [Guest] It could be, and there is a name for it. The Stanford economist Erik Brynjolfsson calls the modern version a J-curve: you expense the retraining and the rebuild now and count the output years later, so the early chart always looks like Slok's. But the curve comes with a test, because in the trough, somebody has to be doing the expensive rebuild.
• 3:24
  [Host] And are the four ninety-three tearing out the steam shaft?
• 3:27
  [Guest] Slok's spending data says mostly no. The rebuild money is concentrated in venture-backed startups, and his read on mainstream firms is that they don't need the latest model and won't pay for it. You don't get the upstroke of a J-curve if you skip the part where you rebuild the floor.
• 3:45
  [Host] Then where did the gain go? Intelligence got radically cheaper, so that value had to land somewhere.
• 3:51
  [Guest] It landed on the seven. The price of a unit of machine intelligence collapsed, but the electricity, the chips, the warehouse, the app on your phone, and the company you can sue when the answer is wrong did not, and the seven own most of that. Slok's own line is that AI may be a great calculator, but in business, intelligence is mostly judgment.
• 4:13
  [Host] So what should I actually watch this week?
• 4:15
  [Guest] Wednesday and Thursday, when Microsoft, Alphabet and Amazon report, because that is the seller's side of this trade and it will probably look fine. Slok's closing line is that the AI story will be decided in the S and P four ninety-three, not the Magnificent Seven. So the thing to watch is whether a single buyer can show you a margin.
• 4:36
  [Host] So the seven have the margin, the four ninety-three have the subscription, and nobody has torn out the steam shaft yet.
• 4:44
  [Guest] For more on where the money goes when intelligence gets cheap, and who pays while the curve is still in its trough, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.