Artificial intelligence has delivered no measurable financial return to US companies outside the technology sector, according to Torsten Slok, partner and chief economist at Apollo Global Management, in a Daily Spark note published on October 24, 2026. The note, titled No Signs of AI Payoff Outside the Tech Sector, sets five charts side by side: revenue per worker, profit margins, productivity growth, the concentration of AI spending and the financial backing of the heaviest AI buyers. Its conclusion is blunt. "Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech," Slok wrote.

In Short

A top economist at the investment firm Apollo looked at whether ordinary big US companies, not the tech giants, are making more money because of AI, and found no sign that they are. This matters because the huge sums being poured into AI chips, data centres and models only make sense if the thousands of companies buying AI eventually earn more from it. For now, the data shows the gains staying with the biggest tech firms, while spending on AI services sits with a small group of mostly venture-backed companies.

The split between seven companies and 493

The note's central device is a division of the S&P 500 into two unequal parts. The "Mag 7" - the seven largest US technology companies, usually counted as Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia and Tesla - sit on one side. The remaining 493 constituents, which Slok calls the "S&P 493," sit on the other. Apollo's note does not list the seven names; the grouping is the conventional one used across financial markets.

"Revenue per worker is rising for the Mag 7 but is flat for the S&P 493," Slok wrote. "Profit margins are rising for the Mag 7 but are flat for the S&P 493."

The first chart, titled "Revenue per worker moving up for tech but not for the rest of the economy," tracks inflation-adjusted revenue per employee, in thousands of dollars at 2019 fourth-quarter prices, from the first quarter of 2019. On PPC Land's reading of the chart, the Magnificent 7 started the period at roughly $695,000 per employee, fell to a trough of about $565,000 around the turn of 2022 and 2023, and then climbed steadily to roughly $780,000 in the most recent quarter shown, with the sharpest jump at the very end of the series. The S&P 493 line moves within a much narrower band: about $495,000 in early 2019, a low near $465,000 in 2021, a peak around $548,000 in late 2022, and roughly $527,000 in the latest reading. Apollo credits Bloomberg, Macrobond and its own chief economist's office for the data.

The second chart, "No signs of profit margins going up outside the tech sector," goes back further, to the first quarter of 2015. Reading from Apollo's chart, the Magnificent 7 profit margin rose from around 11% in 2015 to about 22% in early 2022, dropped to roughly 14% in 2023, and has since risen to close to 30%. The S&P 493 margin stood at about 8.5% in 2015, dipped to around 6.5% during the pandemic, peaked near 11.7% in early 2022, and has hovered between 9% and 10% since 2023. These figures are approximations taken from the plotted lines; Apollo did not publish the underlying values in the note.

The shape matters as much as the level. For the largest technology companies, both revenue per worker and margins have accelerated in the period that coincides with heavy AI investment. For the rest of the index, the lines are close to flat over the same years.

Productivity: no break in the trend

The third chart widens the lens from corporate accounts to the whole economy. Titled "No signs of AI boosting productivity growth," it plots two measures of year-over-year productivity growth from January 1985: labour output per hour from the US Bureau of Labor Statistics, and total factor productivity, or TFP. The TFP series is the utilisation-adjusted measure maintained by the Federal Reserve Bank of San Francisco (Fernald, 2014), which corrects for cyclical swings in how intensively capital and labour are used.

On the chart, output per hour spiked to around 7% during the 2020 pandemic, when lower-productivity jobs disappeared first, and then dipped below zero. In the most recent readings, output per hour growth runs at roughly 2% to 2.5%, while TFP growth has slipped to around zero and appears slightly negative at the end of the series. Neither line shows the kind of sustained step up that would mark a technology-driven productivity boom. "There are still no signs of AI boosting productivity growth," Slok wrote.

TFP is the measure economists usually watch for the effect of a general-purpose technology, because it captures output growth that cannot be explained by adding more workers or more machines. A flat or falling TFP line, set against record AI capital spending, is the core of Slok's case.

From growth engine to cost centre

The note goes further than describing flat lines. It makes a claim about corporate behaviour. "With no impact on the bottom line, mainstream corporate America is pulling back and starting to view token spending not as a source of growth but as a cost center that weighs on margins," Slok wrote.

Token spending refers to the fees companies pay to AI providers for processing text and data, measured in tokens. The note does not quantify the pullback it describes, and it does not cite a survey for the shift in attitude. Other recent data points in the same direction. A KPMG survey of 2,145 senior leaders found that 49% had scaled back, narrowed, delayed or paused AI agent rollouts after costs began to outweigh value, while only 7% reported established return on investment. An August TransUnion study found that 53% of enterprise marketers reported meaningful ROI from AI, with most gains so far coming from time savings rather than revenue.

Slok also draws a line between who buys AI and who does not. "VC-backed companies are spending more and more on tokens, but mainstream corporate America does not need the latest model and is unwilling to pay for it," he wrote. "As long as agentic AI and token spending are confined to a small part of the economy, the macroeconomic effects of AI will be very limited."

Ramp data: spending sits with the top 10%

The fourth and fifth charts carry that argument with spending data from Ramp, the corporate card and expense management company, which publishes the Ramp AI Index based on business transaction data. The fourth chart, titled "AI spending is extremely concentrated in a small number of firms," is a table credited to Ara Kharazian, chief economist at Ramp, and Apollo's chief economist. It shows the share of spending in each category that comes from the top 10% of firms versus the bottom 90%. The note does not explain how the top 10% is defined - whether by total spend, category spend or company size - or what period the table covers.

Spend category Top 10% of firms Bottom 90% of firms
Model serving and inference 99.5% 0.5%
Neocloud 99.0% 1.0%
OpenAI + Anthropic 95.6% 4.4%
Cloud computing 95.7% 4.3%
SaaS excluding AI 91.8% 8.2%
Paid ads 91.9% 8.1%
CRM 84.2% 15.8%

The most AI-specific categories are the most concentrated. Spending on model serving and inference cost - the computing bill for running AI models once they are trained - is almost entirely in the hands of the top tenth of firms, at 99.5%. Neoclouds, the specialist GPU cloud providers that rent computing capacity outside the large hyperscaler platforms, draw 99% of their spend from the same group. Direct payments to OpenAI and Anthropic, the two leading model developers, are 95.6% concentrated.

Yet concentration is not unique to AI. Paid advertising, at 91.9%, is about as concentrated as conventional software-as-a-service spending, at 91.8%. CRM software is the least concentrated category in the table, with the bottom 90% of firms accounting for 15.8% of spend. The comparison suggests that the top tenth of firms dominates business spending in general, and AI spending more so.

For the advertising industry, the paid ads row is a reminder that a small share of companies already accounts for most ad budgets in Ramp's dataset. If AI spending follows the same pattern but more steeply, the companies best placed to deploy AI in marketing are, for now, the same ones that already carry the largest media budgets.

Circularity in token spending

The fifth chart is titled "Circularity in AI shows up in token spending too." Based on the Ramp AI Index, it splits businesses by AI spend intensity and shows whether they are backed by venture capital. Among the top 1% of AI spenders, 80.4% are VC-backed and 19.6% are not. Among median spenders, the proportions flip: 13.1% are VC-backed and 86.9% are not.

The chart's title links that finding to a wider debate about circular financing in the AI industry, in which investors fund AI start-ups that then spend much of that money on cloud computing and model access from companies tied to the same investors. Slok's note does not spell out the mechanism. Its data shows only that the heaviest token buyers are overwhelmingly funded by venture capital, while the typical business buying AI is not.

The implication is that a large share of measured AI demand comes from companies whose spending depends on fresh capital rather than on operating profits. That is not the same as demand from mainstream companies paying for AI out of improved margins - the outcome the S&P 493 charts say has not yet arrived.

Judgment, accountability and the next question

Slok frames the open question as whether falling prices change the picture. "The question looking ahead is whether, even as models become commoditized, mainstream corporate America can turn increasingly cheap intelligence into higher revenue growth and higher profit margins," he wrote. "That may be harder than it sounds."

His reasoning is qualitative. "AI may be a great calculator, but in business, intelligence is mostly judgment," Slok wrote. "Judgment depends on context, and the answer varies with which human or which model you ask." He added that "AI does not understand accountability, trust, culture, unwritten rules, ethical trade-offs or fairness, all of which shape reputational risk and ultimately drive revenue growth and profit margins."

The note closes on a single sentence that summarises the argument: "The bottom line is that the AI story will be decided in the S&P 493, not the Mag 7, and so far, there is no evidence that revenue growth or margins are improving outside the tech sector."

What the data does not establish

The charts show correlation in time, not cause. Rising Magnificent 7 margins could reflect AI-driven sales of cloud capacity, chips and advertising, but they also reflect cost cutting, layoffs, pricing power and cyclical factors that predate the current wave of generative AI. The flat S&P 493 lines do not prove AI has no effect at company level; gains in some firms or sectors could be offset by losses elsewhere, or could take longer to show up in aggregate figures. The note's claim that mainstream companies are "pulling back" on token spending is not backed by a specific data series in the note itself.

The Ramp figures also come with limits. They reflect businesses that use Ramp's cards and expense software, not the economy as a whole, and they are not the same population as the S&P 493 in the first two charts. The note does not state the sample size, time period or definition of the top 10%.

Apollo's earlier notes on the same theme

The October 24 note is the latest in a run of Daily Spark entries in which Slok has tracked the gap between AI investment and returns outside technology. On June 30, 2026, he argued in a note titled "AI: The ROI Runway Could Be Long Outside the Tech Sector" that sectors such as health care, banking and energy face longer timelines to productivity gains than software companies, and that AI company valuations depend on margins eventually rising in the S&P 493. On August 9, 2026, in "The Buyers of AI Are Still Waiting for the Payoff", he broke S&P 493 margins down by sector and argued that the capital spending boom was showing up in the margins of AI sellers, not buyers. The October note adds the revenue-per-worker comparison, the productivity series and the Ramp spending data.

Why this matters for marketing

The advertising market sits on both sides of Slok's divide. Alphabet, Meta, Amazon and Microsoft are among the Magnificent 7 and earn much of their revenue from advertising, which in turn funds their AI infrastructure. Alphabet guided to $175 billion to $185 billion in 2026 capital expenditure when it reported fourth-quarter 2025 results in February. Microsoft said in April that its AI business had passed a $37 billion annual revenue run rate, up 123% year over year, with quarterly capital spending of $31.9 billion. Google said this month that Gemini 4 Argon processed almost 25 trillion tokens in its first 72 hours, and that the company remains capacity constrained.

The advertisers on the other side of those platforms are, for the most part, S&P 493 companies and smaller businesses. When their margins are flat, marketing budgets come under the same scrutiny Slok describes for token spending. Many of the AI features built into ad platforms - automated bidding, generated creative, agentic AI tools that run campaigns - are sold to these buyers on the promise of efficiency. Whether that efficiency reaches their own profit and loss accounts is the question Apollo's charts say has not yet been answered.

Measurement on the buyer side remains thin. A Basis survey published on October 19 found that only 6.5% of 184 organisations measured AI returns against defined benchmarks, even as 88.6% paid for premium AI tools. Survey-based studies tend to show more optimism than the aggregate data: Meta's YouGov survey of 3,983 Asia-Pacific small business owners found 76% reporting a productivity gain, though those were owners' own assessments rather than measured outcomes. Google's AI and Economy ATLAS report found AI usage in 68% of occupations but a median of 21% of tasks within those jobs.

The broader market debate has sharpened too. Ray Dalio said earlier this month that his bubble indicator puts AI level with 1929 and 2000. Analyst Ed Zitron argued in August, citing UBS estimates, that OpenAI and Anthropic could make up more than 48% of Google Cloud revenue in 2027 - a concentration risk that echoes the Ramp data in Slok's note. Apollo's chief economist adds a different kind of evidence to that debate: not valuations or financing, but the plain accounts of the companies that are supposed to be AI's customers.

Timeline

Summary

Who: Torsten Slok, partner and chief economist at Apollo Global Management, using data from Bloomberg, Macrobond, the Federal Reserve Bank of San Francisco, the US Bureau of Labor Statistics and Ramp, including Ramp chief economist Ara Kharazian.

What: A Daily Spark note arguing that AI has delivered "zero financial payoff" to US companies outside tech. Revenue per worker and profit margins have risen for the Magnificent 7 but stayed flat for the S&P 493, productivity growth shows no AI effect, the top 10% of firms account for 99.5% of model serving and inference spend in Ramp data, and 80.4% of the top 1% of AI spenders are venture-backed.

When: Published October 24, 2026, following related Apollo notes on June 30 and August 9, 2026.

Where: The United States, with the analysis centred on S&P 500 companies and Ramp's business spending data.

Why: The returns on AI investment depend on mainstream companies turning AI into higher revenue and margins. Slok argues that this has not happened yet, and that while AI spending stays concentrated among a small group of mostly venture-backed firms, its effect on the wider economy will remain limited. For advertisers and the platforms that sell to them, the question is whether AI-driven efficiency will reach buyers' own profit and loss accounts.