Is there an AI bubble? What Apollo's "no signs of AI payoff outside the tech sector" charts show, and what they don't

Is there an AI bubble? The question got a new round of attention on October 24, 2026, when Torsten Slok, Apollo's chief economist, published a note headlined "No Signs of AI Payoff Outside the Tech Sector." It argues that the money pouring into AI data centers hasn't shown up in the revenue, margins or productivity of companies outside tech. It came nine days after the October 13–15 AI stock and credit sell-off and three days after Meta raised its spending plans again on October 21.

This analysis is based on Apollo's October 24 note; Bureau of Labor Statistics productivity releases; the Census Bureau's Business Trends and Outlook Survey; the Ramp AI Index as published on FRED; a Federal Reserve FEDS Note and a speech by Governor Lisa Cook; the IMF's October 2026 blog and Reuters's report on its outlook; and earlier 9to5AI reporting on company filings, all checked October 26, 2026. It describes evidence and arguments. It isn't investment advice, and it doesn't predict market moves.

The short version

• Apollo's claim: revenue per worker and profit margins are rising at the "Mag 7" tech companies but flat across the rest of the S&P 500, and a utilization-adjusted productivity series shows no AI boost. The note publishes charts without values, date ranges or definitions.
• Adoption is real and rising: Ramp's card data show 55.9% of its business customers paying for AI in September 2026, up from 43.6% a year earlier. Census's broader survey puts firm-level use much lower, at 17% to 20%.
• Payoff is thin in the data so far: BLS productivity growth of 2.2% in the year to Q2 2026 is solid but not exceptional on its own, and only 13% of CIOs in Gartner's survey report significant value.
• Spending keeps climbing: Meta expects 2026 capex of $135–145 billion and has about $825 billion of future commitments in its latest filing.
• No one in this story forecasts a crash as fact. The IMF models a correction as a scenario; Apollo's note is a chief economist's view; the Fed's Cook says the timing of gains is uncertain.

What does Apollo's chart actually measure?

Apollo's note contains five charts, each credited to outside data plus "Apollo Chief Economist":

The bottom line Slok draws is that there is "no evidence that revenue growth or margins are improving outside the tech sector," and that the "S&P 493" is where the AI story will be decided. He also argues that AI can't capture the judgment, accountability and trust that drive revenue and margins.

What the charts don't establish matters as much. The page doesn't define the "Mag 7" or list which seven companies are in it, doesn't give the period each chart covers, and doesn't show values a reader could check. A flat margin line for 493 companies averages very different industries together; it can't show whether AI helped some firms and not others. And Apollo's own disclaimer says the views are the author's, may change without notice and aren't investment advice.

What do official productivity numbers show?

They show solid growth, not a clear AI signature. The BLS revised second-quarter 2026 figures, released September 3, put nonfarm business labor productivity up 1.4% at an annual rate in the quarter and 2.2% from a year earlier, with output up 2.5% and hours up only 0.2% over the year. Manufacturing productivity rose 1.1% from a year earlier.

Rising output with nearly flat hours is the pattern AI optimists expect to see. But the BLS figures don't say what caused it, and labor productivity moves with the business cycle, which is why Apollo used the San Francisco Fed's utilization-adjusted series instead. The next BLS reading, for the third quarter, is scheduled for November 5, 2026.

Governor Lisa Cook, speaking on September 28, 2026, didn't claim AI is visible in productivity statistics. "Currently, I anticipate that productivity gains will provide modest disinflation within the next few years," she said, adding, "I am very uncertain as to the breadth and timing of these channels."

How many businesses use AI, according to the data?

It depends heavily on who is asked and how.

• Census Bureau (Business Trends and Outlook Survey): between 17% and 20% of US firms reported using AI in the previous two weeks from December 2025 to May 2026. For the period ending May 3, 2026, the national rate was 19.8%, with 39.7% in the information sector and about 14% in retail. Firms with 250 or more employees were at 37%. Census changed the question in November 2025 to ask about AI in "any business function."
• Federal Reserve: an April 3, 2026 FEDS Note by Jeffrey S. Allen compared surveys and found estimates ranging from 18% of firms (Census, December 2025) to 78% on an employment-weighted basis (the Atlanta Fed's Survey of Business Uncertainty, November 2025). It attributes most of the gap to sampling and what each survey counts.
• Ramp AI Index: the share of Ramp's more than 70,000 business customers with an observed payment for an AI product or service in a month.

Share of Ramp business customers paying for AI, October 2025 to September 2026

Percent of businesses with an observed AI payment in the month

020406055.9

Oct 2025Jan 2026Apr 2026Jul 2026Sep 2026

Show the data

Month	Adoption rate
Oct 2025	43.6
Jan 2026	45.5
Apr 2026	52
Jul 2026	54.8
Sep 2026	55.9

On Ramp's measure, adoption rose from 43.6% in October 2025 to 45.5% in January 2026, 52.0% in April, 54.8% in July and 55.9% in September 2026. That's the adoption side of the argument. Apollo's point is about the next step: whether paying for AI turns into higher revenue or margins outside tech.

What's the case that AI is a bubble?

The bearish arguments, each from a named source:

• No visible payoff outside tech. Apollo's Slok, as above.
• Few buyers see value. Gartner said on October 19, 2026, that only 13% of 2,002 CIOs it surveyed report significant or transformational value from AI, while forecasting $2.67 trillion of AI spending in 2026, as covered in our analysis of AI usage data.
• Spending is outrunning use. Cook said companies "have only spent a small fraction of the $2 trillion in announced plans," and described AI investment as a source of inflation pressure.
• Circular financing. An IMF blog by Athanasios Vamvakidis and Jason Wu, published with the October 2026 Global Financial Stability Report, reviewed 14 deals since 2023 involving about $300 billion of equity and $650 billion of debt in exchange for $1.4 trillion of future revenue commitments. It warns that stress at one company could spread through such arrangements.

What's the case against?

The arguments that the payoff is coming, or is already here in parts of the economy:

• The tech companies are earning it. Apollo's own charts show rising revenue per worker and margins for the "Mag 7." Meta, for example, reported Q3 2026 revenue up 26%, as our Meta earnings story covers.
• Adoption is early and fast. Cook said "AI technology and its adoption are still in their early stages," cited evidence that adoption is faster than PC or internet adoption at comparable points, and said 71% of small employer firms using AI report higher productivity, citing the Fed's Small Business Credit Survey. That's a self-reported measure.
• Productivity is growing. BLS's 2.2% year-over-year figure, with output growing faster than hours, is consistent with efficiency gains, even if it doesn't prove an AI cause.
• A correction isn't a recession. Reuters reported on October 13 that the IMF's severe scenario, a sharp pullback in AI investment, would cut global GDP by 1.1% by 2028 and US output by 1.8%, and that IMF deputy research director Petya Koeva Brooks said it likely wouldn't push growth to recession levels.

How big is the spending that needs a payoff?

Company filings give the clearest numbers. Meta expects 2026 capital expenditures of $135–145 billion, up from $115–135 billion guided in January, per its October 21 results. Its Q3 2026 10-Q lists about $825 billion of leases not yet started, contractual commitments and a contingent cloud obligation, as our 10-Q analysis details. Suppliers are expanding to match: Nvidia says its partners plan 2.5 times more substrate capacity over three years, per our DSX POD story.

That's the scale behind the "AI bubble chart" searches: spending is concrete and dated, while the return outside tech is, on Apollo's reading, not yet visible. Whether that gap is a lag or a mismatch is the question the data can't yet answer.

What's confirmed, reported and uncertain

• Confirmed (primary data): BLS productivity figures; Census BTOS rates; Ramp AI Index values on FRED; Cook's speech; the FEDS Note; the IMF blog's deal figures; Meta's capex guidance and filing.
• Opinion with charts: Apollo's note, whose charts carry no published values or periods.
• Reported: the IMF's 1.1% and 1.8% GDP scenario figures and Koeva Brooks's remark (Reuters).
• Uncertain: how much of 2026 productivity growth comes from AI; whether non-tech margins will move; how large any market correction would be. None of the sources here predicts one as fact.