Corporate America outside the technology sector has yet to show any observable financial payoff from artificial intelligence, according to Apollo Global Management’s chief economist. In an analysis published on October 24, 2026, Torsten Slok argues that AI gains remain concentrated in the Magnificent Seven, while the rest of the S&P 500 is essentially flat. His central point: the real test of AI will play out in the S&P 493.
🔑 Key Takeaways
- At the Magnificent Seven (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla), revenue per employee and profit margins are rising sharply.
- For the other 493 companies in the S&P 500, these indicators are essentially unchanged.
- The exact levels depend on the source and the inflation adjustment used, but all versions of the analysis confirm the gap.
- Research from MIT and FCLT Global suggests AI’s return on investment may arrive well beyond a six-month window.
A Persistent Gap Between Big Tech and the Rest of the Economy
Apollo’s report, titled “No Signs of AI Payoff Outside the Tech Sector,” separates the seven largest technology stocks, grouped as the Magnificent Seven, from the other 493 companies in the index, which Slok calls the “S&P 493.” According to Slok, revenue per employee and profit margins are rising for the first group, while they remain virtually flat for the second. Productivity data show no acceleration in total factor productivity either, despite rapid AI adoption and heavy investment.
| Metric | Magnificent Seven | S&P 493 |
|---|---|---|
| Revenue per employee (inflation-adjusted) | about $780,000 | about $525,000 |
| Profit margin | near 30% | about 10% |
| Trend since 2022 | Sharp increase | Essentially flat |
These figures come from Finance Yahoo, which uses a constant-inflation base. The qualitative message is consistent across coverage: Apollo describes the divide as structural rather than the product of a single quarter.
Figures Diverge Across the Coverage
The published numbers are not identical from one outlet to another. Finance Yahoo cites roughly $780,000 in revenue per employee for the Magnificent Seven on a constant-inflation basis, against about $525,000 for the S&P 493. Business Insider and Ground News work in absolute terms: revenue per employee at technology companies reportedly rose from about $550,000 in 2022 to nearly $800,000 today, while the other S&P 500 companies stayed close to $550,000 over the past seven years.
- Constant-inflation basis: a gap of roughly $255,000 per employee in favour of the Magnificent Seven.
- Absolute basis: an increase of about 45% for tech, against stability for the rest of the index.
- Margins: near 30% versus about 10% according to Finance Yahoo, with non-tech margins reportedly stuck around 10% since 2022.
These level differences do not change the qualitative conclusion. They do show how strongly the chosen series and inflation adjustments shape absolute values. Any precise comparison should state its base, or the gap will appear either sharper or fuzzier than the underlying data support.
Concentrated Spending and Changing Usage
Companies are spending billions of dollars on chips, data centres, software and AI models. That spending, however, is heavily concentrated among a small number of players and sectors. Apollo argues that as long as agentic AI (systems that chain tasks together autonomously) and token consumption (the text units processed by models) remain confined to a limited part of the economy, macroeconomic effects will stay very small.
“Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.”
Torsten Slok, Chief Economist, Apollo Global Management
Behaviour among non-tech companies is shifting. Slok notes that venture-backed firms are consuming more tokens, while large “mainstream” corporations do not need the latest models and are unwilling to pay for them. As he puts it, “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.” Token consumption is thus starting to look like a cost centre rather than a growth engine.
The central question is whether companies can turn increasingly cheap intelligence into higher revenue and margins, even as models become commoditised. Slok says the task “may be harder than it sounds.” He adds that “AI may be a great calculator, but in business, intelligence is mostly judgment,” and that judgment depends on context. In his view, AI does not understand accountability, trust, culture, unwritten rules, ethical trade-offs or fairness. Those elements shape reputational risk and, ultimately, revenue growth and margins.
MIT, the J-Curve and the Question of Timing
Timing sits at the heart of the debate. An MIT study, covered by MIT Technology Review, assessed more than 300 enterprise generative AI deployments, conducted 150 executive interviews and ran 350 employee surveys. Only 5% of pilot projects reportedly cleared a narrow success bar: moving beyond the pilot stage and producing measurable P&L gains within about six months. That result is often summarised as 95% of enterprise generative AI projects showing no measurable profit-and-loss impact in that period.
| Source | Scope | Main finding |
|---|---|---|
| Apollo (Torsten Slok) | S&P 500, macro data | Gains limited to the Magnificent Seven |
| MIT study | More than 300 deployments | 5% of pilots with measurable impact in six months |
| FCLT Global | General-purpose technologies | Returns after several fiscal years |
According to FCLT Global, failures stemmed less from model quality than from poor fit with workflows, fragile tools and insufficient learning or adaptation. Many projects targeted visible, easily measured functions such as sales and marketing, while neglecting automation of administrative functions, which may offer a higher long-term return on investment.
FCLT Global frames AI as a general-purpose technology following a “J-curve.” The model, described by Erik Brynjolfsson, Daniel Rock and Mikkel Syverson in 2021, predicts that early years can show weak or even negative measured growth. Upfront intangible investments in process redesign, employee training and new organisational capabilities depress measured productivity at first, while laying the groundwork for future gains.
Earlier work cited by the same organisation points in the same direction. Brynjolfsson, Hitt and Yang (2002) reportedly found that information technology productivity gains were capitalised into higher equity valuations, but only after firms invested in complementary systems and processes. Basu and Fernald (2008) reportedly observed a comparable pattern in sectors that invested in information and communication technologies, with productivity and profitability rising after a lag.
Six-month payback windows and quarterly pressure push management teams toward gains that are visible immediately rather than toward the foundations of future gains. That is precisely the criticism FCLT Global levels at such return horizons.
Conclusion: A Payoff Delay, Not an Absence of Value
The lack of immediate returns does not prove that AI lacks value. It suggests instead that short-term evaluation criteria are ill-suited to a long-term transformation. For Apollo, the trajectory to watch is that of the S&P 493: if margins and revenue per employee rise outside tech, the case for broadly diffused AI profitability will strengthen.
Three scenarios stand out. In the first, diffusion accelerates as token costs fall and workflows are redesigned, validating the J-curve. In the second, margin pressure builds and AI budgets are trimmed for lack of visible returns, feeding bubble fears. In the third, gains remain concentrated in a handful of giants, with a persistent gap between technology and the rest of the economy. None of these outcomes is settled yet, as public data remain partial and the series still diverge.
Sources
- TechCrunch, October 25, 2026
- Finance Yahoo
- Ground News
- Apollo, No Signs of AI Payoff Outside the Tech Sector
- Business Insider, October 2026
- FCLT Global, AI payoff and time horizons
Cet article est publié à titre informatif et éducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute décision.
