Apollo Economist: AI Delivers 'Zero Financial Payoff' to Most Companies Outside Tech
On October 25, 2026, TechCrunch reported that Apollo Global Management chief economist Torsten Slok sees “zero financial payoff” from AI for most large companies outside the tech sector. The claim matters because it pushes back against the strongest AI-productivity narratives using a simple macro observation: across the S&P 493, profit margins and revenue per worker look flat.
What Apollo’s Torsten Slok actually measured
Apollo’s note is a top-line observation about the S&P 493, not a firm-level return-on-investment study. Slok defines the S&P 493 as the S&P 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla. In that group, he finds that profit margins and revenue per worker appear flat, with no evidence that AI is improving productivity.
Slok’s central 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.”
He is also skeptical that non-tech companies will simply lag tech giants for a while and then catch up. He argues it will be difficult for corporate America to “turn increasingly cheap intelligence into higher revenue growth and higher profit margins.” His explanation is that business value depends on more than calculation: “AI may be a great calculator, but in business, intelligence is mostly judgment.” He adds 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 strongest counterargument: the data cannot isolate AI’s payoff
The Apollo note cannot prove that AI spending has produced no value, because it does not measure AI spending against firm-level outcomes or establish a counterfactual. A top-line revenue chart for large non-tech companies can look flat even if those companies are getting real value from AI, especially if AI spending is small relative to total revenue.
Hacker News commenter hyperpape made this critique directly:
“This is a complete misrepresentation of what the Apollo piece can actually show, because it's not measuring the 'payoff' of AI investments, but only the top level revenue of those companies.”
hyperpape also argued that the analysis needs a counterfactual: what would those companies’ revenue and margins look like if they had not spent on AI? And he noted that “while AI spending is increasing rapidly, normal companies aren't spending 1% of their revenue on it.” If AI spending is a small share of revenue, a noisy top-line chart is a weak test of whether AI is useful.
Commenter cl42 raised a related measurement problem: the S&P 493 may not be where most AI spending is happening. cl42 wrote:
“What I'd really want to see is revenue and profit per employee over time tracked against token usage/spend, and not limited to the S&P 493.”
That is a more testable framing: connect AI usage or token spend to revenue per employee, profit per employee, cycle time, or output quality at the firm or process level, rather than inferring ROI from aggregate margins.
Where AI is showing measurable operational gains
Some Hacker News commenters reported concrete gains in narrow, well-defined workflows, which supports the view that AI’s payoff is uneven rather than universally zero. Commenter ManBeardPc described using AI in data science and specialty chemistry to find and optimize formulas:
“We can get from idea to prototype in a day compared to 1-2 weeks. All that boilerplate to process data took a long time to code by hand, especially for the data science/math guys who are not programmers. Now we can actually try all ideas instead of guessing the most promising and only try that.”
That is a process-level productivity gain: faster iteration, more experiments, and less time spent on boilerplate. It does not necessarily show up as higher revenue per worker across an entire S&P 493 company, especially in the short term.
Other commenters emphasized that AI works only when it fits the task. doginasuit wrote that AI “can certainly speed up development but you have to walk a narrow path. It generates garbage much easier than valuable product.” Damogran6 made a similar point: “When it's the right tool, it's a silver bullet. When it's the wrong tool...or not the right tool yet, you can spend more time steering it than you would doing it yourself.”
Organizational bottlenecks can also absorb AI gains. victor9000 described a sales process that still requires sending engineering, product, and sales staff to a customer site: “I'm personally much more productive using AI, but that only goes so far when closing a deal still follows a process stuck in the 50s.”
The zero-sum and “AI tax” hypothesis
One explanation for flat aggregate margins is that AI productivity gains get competed away. If every company in an industry uses AI to improve output, no single company may keep the surplus; customers or competitors may capture it instead. Commenter kozikow described this dynamic:
“Your company uses AI and it gets ahead a lot quicker. However, your competition does so as well. Many areas of economy are zero sum game - so your $ ends up being the same as previously. However, now with a 'tax' paid to AI model providers...”
In that view, AI becomes a competitive necessity rather than a source of durable excess profit. Companies may have to pay model providers to stay even, while the aggregate margin benefit disappears.
Commenter variadix offered a simpler constraint: “Many industries are simply not bottle-necked by intelligence.” If the limiting factor is regulation, physical capacity, distribution, trust, or sales cycles, better reasoning or code generation may not move revenue much. Commenter tantalor made a related point: unless a company is “selling lines of code, or math theorems,” it may be selling something that AI code generators do not directly unlock.
Judgment, accountability, and organizational transformation
Slok’s judgment argument resonated with some commenters who see AI adoption as an organizational problem, not a model-capability problem. Commenter mgdev argued that most companies do not know which people, processes, or behaviors actually drive their best outcomes:
“90% of the business outcomes are delivered by 10% of the people. The problem is they don't know which people. Therefore they also don't know which processes, if automated or augmented with AI, will achieve the target outcomes. So instead they’re shotgunning.”
mgdev added that existing people-based systems have stabilizing mechanisms that prevent misalignment from taking the whole operation off course, while AI “stresses and often breaks those.” The implication is that successful AI adoption may require treating it as an org transformation: starting a small team that builds ground-up and eventually absorbs or replaces parts of the old workflow.
Not everyone accepted Slok’s calculator analogy. Commenter kazinator objected: “Calculators do not hallucinate.” That is a fair distinction: AI systems are not deterministic calculators, and their failure modes—fabrication, inconsistency, and unreliable judgment—are part of why measurable business payoff can be hard to capture.
Why the headline may overstate the evidence
The Apollo note is a useful corrective to maximalist claims that AI will rapidly transform the entire economy, but it is not a definitive verdict on AI’s business value. Commenter simonw pointed out that Apollo published a widely cited September 2025 note titled “AI Adoption Rate Trending Down for Large Companies,” and said: “I don't think that one has held up very well.”
simonw later softened the criticism after noting Apollo’s publishing cadence:
“I'm going to take this one back a a bit: it turns out Apollo publish articles like this on a daily basis, so the occasional headline that appeals to the 'hungry for negative stories about AI' crowd is less of a pattern than if they published less frequently than that.”
That context matters for interpreting the headline. A daily research note can surface a real macro pattern without being a comprehensive study of AI ROI. The HN thread also pointed to a prior large discussion of the same theme, suggesting the debate is recurring rather than settled.
What would make the debate testable
A stronger test of AI’s financial payoff would connect adoption to outcomes at the level where AI is actually used. That means measuring revenue per employee, profit per employee, cycle time, error rates, customer response time, or experimentation throughput against token usage or AI spend. It also means including companies outside the S&P 493, since much AI spending and experimentation may be happening in smaller firms, private companies, or startups.
A credible study would also need a counterfactual: what would have happened without the AI investment? Without that, flat margins are consistent with “AI is useless,” but they are also consistent with “AI is useful but too small to move aggregate margins yet,” “AI gains are competed away,” or “AI is valuable in specific workflows but offset by organizational friction.”
Bottom line
Apollo’s Torsten Slok is right that the macro data do not yet show a clear AI-driven profit surge outside tech. The S&P 493’s flat profit margins and revenue per worker are a legitimate warning against assuming that cheap intelligence automatically becomes higher revenue or margins.
But the claim of “zero financial payoff” is stronger than the evidence supports. The Apollo note measures top-line outcomes, not firm-level AI returns, and it does not establish what would have happened without AI spending. The Hacker News discussion adds the missing nuance: some teams see large gains in narrow workflows, many industries are not bottlenecked by intelligence, and competitive dynamics may turn AI into a cost of staying even rather than a source of excess profit. The real question is not whether AI has produced zero value everywhere, but where, for whom, and under what conditions AI spending turns into measurable business outcomes.