Does artificial intelligence actually make money for the companies using it? A note published on October 24, 2026 by the bank Apollo Global Management gives an answer that cuts against the prevailing narrative: outside the seven largest US technology companies, the financial return on AI in business remains, for now, almost invisible in the numbers. For an SMB leader who has already invested in ChatGPT, Copilot or Claude licenses, this finding deserves to be taken seriously, without giving up on AI altogether.
In short
- According to a note from Apollo Global Management's chief economist, Torsten Slok, published on October 24, 2026, revenue per employee and profit margins are rising at the "Magnificent Seven" (Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia, Tesla), but remain broadly flat across the other 493 companies in the S&P 500.
- PwC's 29th Global CEO Survey (4,454 CEOs, 95 countries, published January 19, 2026) found that 56% of CEOs have seen no significant financial benefit from AI, and only 12% report gains in both cost and revenue.
- A MIT NANDA study published in 2025 had already found that 95% of enterprise generative AI pilots produce no measurable return, versus only 5% that create real value.
- The cause identified is not model quality, but fragile implementation: isolated pilots, no integration into real processes, and insufficient measurement.
- For an SMB, the lesson is not to stop investing in AI, but to pick one precise use case, truly integrate it, and measure before scaling it up.
What Apollo Global Management's note reveals
Apollo Global Management is a US asset manager whose economic research team closely tracks AI's effect on the real economy. In its October 24, 2026 note, chief economist Torsten Slok compares the financial performance of the "Magnificent Seven", the seven tech giants that dominate US stock indices, against the other 493 companies in the S&P 500.
The finding is clear: revenue per employee and profit margins are rising at the tech giants, which either sell AI infrastructure (cloud, chips, models) or integrate it at massive scale into products used by millions of customers. But at the vast majority of companies, which buy and use AI rather than sell it, those same indicators remain broadly flat. The note goes further: it finds no acceleration in total factor productivity, an economic measure of how efficiently labor and capital are combined, that can be attributed to AI at this stage.
The “Magnificent Seven”
Revenue per employee and margins have risen since the rise of generative AI, according to Apollo Global Management. These companies either sell AI infrastructure or integrate it at very large scale into their products.
The other 493 S&P 500 companies
Revenue per employee and margins remain broadly flat. No visible acceleration in total factor productivity tied to AI, according to the same note.
A direct consequence, per Apollo: some companies are already cutting their AI spending, and more finance departments now treat AI token costs as an expense that eats into margins, rather than as an automatic growth driver.
PwC and MIT confirm the same gap
This finding is not an isolated one. PwC's 29th Global CEO Survey, published January 19, 2026 and based on 4,454 CEOs across 95 countries, puts precise numbers on this gap between promise and result.
The detail matters: CEOs who report gains in both cost and revenue are two to three times more likely to have embedded AI extensively into their products, sales outreach, and strategic decisions, rather than using it at the margins. An earlier MIT NANDA study, titled "The GenAI Divide", had already documented the same pattern in 2025, based on more than 300 enterprise generative AI deployments.
According to MIT, the cause of this gap is not the quality of the models themselves. It comes down to implementation: isolated pilots, disconnected from real processes, unable to retain context from one task to the next, and poorly aligned with the company's actual operational needs.
Why some companies do better
Between the near-total failure of isolated pilots and the success of the tech giants, there is a middle path: the 12% of PwC CEOs reporting real financial gains, and the 5% of MIT pilots creating measurable value. What they have in common is not company size or AI budget, but method.
Pick a measurable use case
Integrate AI into the real workflow
Measure before scaling up
Revisit the math every quarter
Measured optimism
This gap is not a condemnation of AI: it mainly shows that value concentrates on uses that are well chosen, well integrated, and measured. An SMB that moves forward with one precise use case, rather than an "everywhere, fast" AI strategy, is statistically more likely to land among the 12 to 33% that win, rather than the majority that still sees no financial effect.
What the three studies say, at a glance
| Study | Date | Key figure | Lesson for an SMB |
|---|---|---|---|
| Apollo Global Management (Torsten Slok) | October 24, 2026 | Flat margins outside the "Magnificent Seven" | AI's financial effect is still concentrated at the tech giants |
| PwC, 29th Global CEO Survey | January 19, 2026 | 56% of CEOs with no significant financial benefit | Return depends on the scale and maturity of deployment, not company size |
| MIT NANDA, "The GenAI Divide" | 2025 | 95% of pilots with no measurable return | Failure usually comes from implementation, not the chosen model |
FAQ
So is AI not profitable for an SMB?
Not automatically profitable, no. All three studies show that a return exists, but only for a minority of companies that integrate AI deeply into their processes and measure its effects, rather than those that pile up scattered, isolated uses.
Why do large tech companies get more out of AI?
According to Apollo Global Management, these companies either sell AI infrastructure (cloud, chips, models) themselves or integrate it at massive scale into products used by millions of customers, which multiplies the effect of every improvement. An SMB does not have that same leverage, which is why choosing use cases carefully matters so much.
Should a company stop investing in AI until the picture is clearer?
No. PwC's analysis shows that companies deploying AI broadly, with strong governance and integration foundations, report nearly 4 points of extra margin compared to others. The risk is not investing in AI, but investing without a precise use case or measurement of results.
How can an SMB concretely measure the ROI of an AI tool?
By comparing one indicator before and after adoption: time spent on a task, error rate, customer response time, or revenue generated, weighed against the tool's real cost (subscription, API tokens, team training time).
Going further
This data point echoes another trap we've already covered: AI can also intensify work instead of easing it, as shown in our article on the AI productivity paradox. To structure an AI rollout that produces a real return, see how other leaders have done it in our client success stories.
