Apollo's Torsten Slok Says AI Has Paid Off Only for the Magnificent 7. His Note Prints No Figures
A.I. / news
Apollo's Torsten Slok Says AI Has Paid Off Only for the Magnificent 7. His Note Prints No Figures
The Oct. 24 note says the S&P 493 shows flat margins and no productivity gain. Ramp's September index has 55.9 percent of US businesses paying for AI.
Torsten Slok, Partner and Chief Economist at Apollo, wrote on Oct. 24 that AI "has delivered zero financial payoff to corporate America outside of tech". The note, "No Signs of AI Payoff Outside the Tech Sector", drew 548 points on Hacker News and carries five charts but no numbers in its text.
Slok's Daily Spark post is the whole argument. This article sets it beside the adoption data of one of his own cited sources, Ramp's AI Index.
What Slok claims
He makes four claims, each tied to a chart. Revenue per worker, adjusted for inflation to the fourth quarter of 2019, is rising for tech and not for the rest of the economy. Profit margins are rising for the Magnificent 7 and flat for the S&P 493, the index's other members. Productivity, measured by the utilisation-adjusted total factor productivity series from the Federal Reserve Bank of San Francisco, shows no AI-driven gain. AI spending is concentrated in a small number of firms.
He adds that mainstream firms treat token spending as a cost centre, and that "the AI story will be decided in the S&P 493, not the Mag 7". The post lists Bloomberg, Macrobond, the Bureau of Labor Statistics and Ramp among its sources.
The text gives no margin figure, no revenue-per-worker figure and no productivity reading. A reader can verify the direction of each chart only by looking at it.
What Ramp's index measures
Ramp, a corporate card and bill-pay platform, publishes the adoption series Slok cites. Its figures run through September 2026 and come from more than 85,000 US businesses. Ramp counts a firm as an adopter if it shows a positive transaction for an AI product in a given month. Ara Kharazian, Ramp's lead economist, is the named author.
The index shows 55.9 percent of businesses adopting in September, up 0.6 points from 55.2 percent in August. Ramp warns that its results likely understate adoption because free tools and personal accounts leave no paid transaction.
- Information79.9 %
- Finance and insurance73.7 %
- Manufacturing60.9 %
- Retail trade49.9 %
- Construction43.2 %
- Healthcare42.5 %
- Accommodation and food33 %
Source: Ramp AI Index, data through Sep 2026, accessed 2026-10-25
By provider, Anthropic leads at 43.8 percent, ahead of OpenAI at 40.6 and xAI at 4.1. By size, large businesses stand at 65.7 percent, medium at 62.5 and small at 50.0.
Where the two sources do and do not collide
They measure different things. Ramp counts purchases, and Slok counts financial results. A business can pay for a tool and see no change in margin, which is consistent with both.
The gap is that Slok's note gives no result for the 493 companies that the Ramp data shows are buying: 60.9 percent of manufacturers and 73.7 percent of finance and insurance firms pay for AI, per Ramp. If those firms are the S&P 493, the note's flat-margin claim covers a group that is spending. Ramp's sample is US businesses of all sizes, not the S&P 500, so the match is loose.
Neither source separates AI cost from AI benefit. Slok attributes no margin movement to any named company. The question the note raises is whether the spending shows up as profit, and the note answers it with a chart rather than a figure.
Related coverage
The spending side is visible elsewhere. OpenAI said ChatGPT ads reached 89 countries, with more than 100,000 advertisers having run campaigns. Microsoft reported that MAI-DxO scored 83 percent on NEJM cases in a published study, a case where a measured result exists.
The note prints no chart values. Ramp's index carries both an August and a September reading, and an October one would be the next data point.
Sources
More in A.I.
- 01Eka's 1,500 Picks an Hour Is a Lab Number, and Its Own Investors Say SoThe Cambridge, Mass. startup raised $150 million for a vision-force-action robot model. Its headline throughput comes from induction and bin-picking tests, and customer pilots are unnamed.
- 02DeepSeek V4.1-Flash-1017 Posts Gains on All 11 Listed Benchmarks, Every One Of Them Its OwnThe October 17 update ships 763B-parameter weights under MIT with a 1M-token window, but the model card gives no active-parameter count, hardware guidance or training data.
- 03Cerebras Says Kimi K3 Fits on 59 Wafers, and Gives No Price for Any of ThemA vendor walkthrough puts Moonshot's 2.8-trillion-parameter model at 35 plus 24 wafer-scale chips and claims 1,000 to 2,000 tokens a second. Every input is an assumption.
- 04ChatGPT Ads Reach 89 Countries as OpenAI Says More Than 100,000 Advertisers Have Run CampaignsOpenAI added 26 countries on October 21. Ads stay limited to the Free and Go plans, and the company gave no per-country prices or user counts.