Wall Street’s optimism around AI depends on a simple idea: companies outside the biggest technology names will use AI to become more profitable. Torsten Slok, chief economist at US financial firm Apollo, is warning that the path from AI adoption to higher margins may be much slower than investors expect.
His argument is not that AI has no value. It is that the market may be assuming a faster financial payoff than companies outside tech can realistically deliver.
The bet behind AI valuations
According to Slok, AI company valuations rest entirely on the promise of rising margins at S&P 493 companies. That group means the index minus "the magnificent seven".
This is a critical distinction. The market is not only valuing the companies that build AI systems. It is also assuming that a much wider set of businesses will use those systems to lift earnings.
For that bet to work, AI needs to move beyond experimentation and show up in financial results. It has to improve workflows, reduce friction, raise output, or otherwise help companies generate better margins. Slok’s warning is that, outside tech, there is currently no sign that this is happening.
That gap matters because markets are pricing in fast earnings growth. If the expected improvement does not appear in real cash flows, the current enthusiasm around many AI stocks becomes harder to support.
Regulated industries may move more slowly
Slok points to several industries where the road to AI-driven productivity could be longer: healthcare, banking, energy, pharma, and manufacturing. These sectors are not blank canvases where new software can simply be dropped into place.
They often depend on strict processes, sensitive information, and systems where errors can carry serious consequences. The source specifically notes process overhauls and privacy requirements as factors that could delay productivity gains.
That does not mean these industries cannot benefit from AI. It means the benefits may take longer to reach the income statement. A company may need to redesign how work is done, decide where AI can be used responsibly, and create controls around data before any margin improvement becomes visible.
Slok says these delays could push productivity gains "well beyond what the market currently projects." That is the center of the risk: the market may be valuing a near-term earnings effect while the actual operating changes require a much longer timeline.
Productivity is not the same as profit
There is another challenge even when AI helps individual employees work faster. In knowledge work, those improvements can be difficult to measure.
If a worker saves time, produces a draft more quickly, or handles a task with less effort, the benefit may be real. But if management cannot clearly track the improvement, it may not lead to a financial decision. The gain can be absorbed into daily operations instead of appearing on the balance sheet.
This creates a measurement problem. Investors may expect AI productivity to become visible as higher profit margins, but companies need clear metrics before they can act on those improvements. Without that, the benefits may remain operational rather than financial.
For AI valuations, this is an important distinction. A tool can be useful without immediately changing earnings. A business can become more efficient in small ways without producing the kind of margin expansion markets are expecting.
The timing risk for AI stocks
Slok’s warning comes down to timing. If the productivity bump takes five years instead of five months, many AI stocks face a painful repricing.
That sentence captures the pressure inside the current market narrative. Investors are not only betting that AI will matter. They are betting that it will matter quickly enough to justify today’s valuations.
If cash flows trail behind expectations, the market may need to reset. The source also notes that falling token costs could cap hyperscaler revenue, adding another pressure point for companies tied to AI infrastructure and usage.
The issue is not whether AI can make employees more productive in specific tasks. The larger question is whether those gains can spread across non-tech companies, be measured clearly, survive regulated environments, and turn into higher margins within the timeframe markets appear to expect.
What investors should watch
The source points to a practical way to evaluate the AI story: look beyond adoption and focus on evidence of margin improvement outside tech.
Important questions follow from Slok’s argument:
- Are S&P 493 companies showing visible margin gains from AI?
- Are regulated industries able to overhaul processes quickly enough?
- Can companies measure productivity improvements in knowledge work?
- Are real cash flows keeping pace with market expectations?
- Could falling token costs limit hyperscaler revenue?
Until those questions have stronger answers, the market’s AI expectations remain exposed to a timing problem. The technology may be moving fast, but the financial payoff across many industries may move much more slowly.