Artificial intelligence has produced “zero financial payoff” to U.S. companies outside of the tech sector, argues Apollo Chief Economist Torsten Slok. Slok claims that while revenue per worker and profit margins have risen for the “Magnificent 7," both have stagnated across the S&P 493 — with no clear indication of AI-driven productivity growth. “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,” Slok said.
The AI Boom Has a Nothingburger Problem Outside of Tech
Here's the uncomfortable truth about the AI revolution: for all the billions poured into data centers, chips, and frontier models, the payoff has shown up in exactly seven companies. Everyone else? Still waiting.
Look at the numbers, and it's hard to argue. Revenue per worker is climbing for the Magnificent 7 while it stays basically flat for the other 493 companies in the S&P 500. Profit margins tell the same story — widening for Big Tech, stuck in place for mainstream corporate America. And productivity growth, the thing AI was supposed to supercharge? Still no visible bump. Zero, zip, nada.
That's not just a curiosity. It's starting to change how corporate America behaves. When there's no measurable impact on the bottom line, AI stops looking like a growth engine and starts looking like a cost center — another line item eating into margins. Companies that were experimenting enthusiastically a year ago are now quietly asking their CFOs whether all that token spending is actually buying them anything.
The Two Economies of AI
There's a real bifurcation happening. VC-backed startups are burning through tokens at a ferocious clip, building agentic AI products on the latest and greatest models. Meanwhile, mainstream corporations look at those same frontier models and shrug. They don't need cutting-edge. They need cheap, reliable, and good enough — and they're increasingly unwilling to pay a premium for more intelligence than their use cases require.
This matters enormously for the macro picture. As long as heavy AI spending is confined to a small sliver of the economy, the aggregate effects will be limited. A handful of hyperscalers retooling themselves doesn't move national productivity. The S&P 493 — the banks, retailers, manufacturers, insurers, and logistics companies that actually employ most of America — has to come along. And it hasn't.
So Will Cheaper Intelligence Save the Day?
Here's where it gets interesting. Model prices are collapsing. Intelligence is becoming commoditized, and tokens keep getting cheaper. The optimistic take is that this eventually trickles into everything — every workflow, every decision, every process — and the S&P 493 finally converts cheap intelligence into revenue growth and fatter margins.
Maybe. But I'd caution against assuming it's automatic, because there's a trap hiding in the word "intelligence."
AI is a spectacular calculator. But in business, intelligence is mostly judgment. Judgment is contextual — the right answer depends on who's asking, what's at stake, and what the organization can actually absorb. Ask five experienced executives the same question, and you'll get five defensible answers. Ask five models, and you'll get the same problem, just faster.
And judgment isn't the hardest part anyway. The hardest part is everything AI can't touch: accountability, trust, culture, unwritten rules, ethical trade-offs, fairness. These aren't soft skills garnish — they're what determine reputational risk, and reputational risk is what drives revenue growth and margins over any horizon that matters. A model can draft the termination email. It can't be the person who has to look the employee in the eye, and it certainly can't absorb the fallout if the call turns out to be wrong.
What I'm Watching
The question going forward isn't whether models get better. They will, and fast. The question is whether mainstream corporate America can take this increasingly cheap intelligence and convert it into something the P&L actually reflects. That's an organizational problem, not a technical one — and organizations are slow, political, and full of exactly the human messiness that AI doesn't understand.
The bottom line is simple: the AI story will be decided in the S&P 493, not the Mag 7. And so far, the evidence says the revolution hasn't left Silicon Valley. Until revenue growth and margins start moving outside of tech, the smart money should treat AI hype with exactly as much skepticism as it deserves — which is some, but not infinite.