Apollo Chief Economist: AI Shows Zero Financial Payoff Outside Big Tech
Corporate America is pulling back on AI spending as the S&P 493 sees flat margins and zero productivity gains from the technology.
Business · Source: Hacker News
What happened
Apollo Chief Economist Torsten Slok just dropped a massive reality check on the current artificial intelligence boom. According to his latest macroeconomic analysis, the technology has delivered absolutely zero financial payoff to corporate America outside of the tech sector. The data paints a stark divide in the market. Revenue per worker is rising steadily for the Magnificent Seven tech giants. However, that same metric remains completely flat for the remaining S&P 493 companies. Profit margins show the exact same trend.
Mainstream companies are now actively pulling back on their AI budgets. They are starting to view token spending not as a magical source of growth, but as a heavy cost center that weighs down their profit margins. There is a clear disconnect between startups and traditional businesses. Venture-backed companies continue spending more and more capital on tokens. Meanwhile, mainstream corporate America realizes it does not need the latest frontier model and is entirely unwilling to pay for it.
The broader macroeconomic data backs up this corporate hesitation. Slok points out that there are still no signs of artificial intelligence boosting overall productivity growth in the economy. Total factor productivity remains unaffected by the hype. As long as agentic AI and heavy token spending remain confined to a very small part of the tech economy, the macroeconomic effects of AI will stay severely limited.
Key facts
- Zero — The financial payoff AI has delivered to corporate America outside of the tech sector
- S&P 493 — The group of companies seeing flat revenue per worker and flat profit margins despite the AI boom
- Mag 7 — The only group of companies seeing rising revenue per worker and profit margins
Why it matters
This changes the fundamental math for any founder trying to sell AI software to traditional enterprise customers. You cannot just sell raw intelligence, API access, or faster generation speeds anymore. Mainstream corporate America is waking up to the fact that business intelligence requires deep human judgment. Judgment depends heavily on context. The correct answer often varies depending on which human or which specific model you ask. Artificial intelligence models simply do not understand corporate accountability, workplace culture, unwritten rules, or ethical trade-offs.
The second-order effect here is a brutal wake-up call for the entire AI infrastructure market. These missing human elements shape reputational risk. Reputational risk ultimately drives revenue growth and profit margins for traditional businesses. If the S&P 493 does not start seeing actual financial returns from AI integration, the enterprise software market will freeze over. The entire trillion-dollar AI narrative hinges on widespread adoption by the broader economy. It cannot survive if tech companies are only selling tools to other tech companies.
For builders
Stop selling raw models to enterprises
Mainstream companies do not want or need the latest frontier models. They currently view token spending as a pure cost center that drags down margins. You will lose enterprise deals entirely if you cannot prove immediate margin expansion or direct revenue growth.
Build for accountability and trust
Corporate buyers are terrified of reputational risk and compliance failures. AI lacks context for human judgment, fairness, and unwritten corporate rules. Founders who build strict guardrails that solve for accountability and trust will capture the enterprise budgets that others lose.
Target the S&P 493 specifically
The Magnificent Seven are already seeing AI payoffs, but the real untapped market is the rest of the economy. If you can build a product that actually moves the needle on revenue per worker for non-tech companies, you will build a massive, defensible moat.
My take
I keep saying that building AI products in a vacuum is a death sentence for founders. Slok's data proves exactly what we feared. Selling cheap intelligence is completely useless if your model lacks business context and human judgment. If we want to take enterprise money, we have to solve their actual margin problems instead of just bragging about our token generation speeds.
Original reporting: Hacker News. This is my rewrite and opinion.