AI Delivered "Zero Financial Payoff" Outside Tech, Says Apollo's Chief Economist

Strip the seven biggest tech companies out of the S&P 500 and something odd happens to the AI story: the numbers stop moving. That's the finding behind a research note from Torsten Slok, chief economist at Apollo Global Management, which argues that for corporate America outside the tech bubble, AI has produced essentially no financial return so far.

Slok built what he calls the S&P 493 — the 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla — and looked at the fundamentals. Profit margins: flat. Revenue per worker: flat. No visible productivity lift. As reported by TechCrunch, his conclusion was blunt: "Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech."

Why "cheap intelligence" isn't showing up in the accounts

The generous reading is that everyone outside the big seven is simply lagging — they bought tools, they're piloting, the gains will land later. Slok didn't buy it. He argued it will be hard for companies to convert "increasingly cheap intelligence" into higher revenue growth and higher profit margins, which is the part most AI pitches quietly skip over.

That framing matters because the cost side of AI is easy to measure — licences, seats, GPU hours, integration work, the engineer who keeps the pipeline alive. The revenue side is much harder. Unless a tool changes what you charge for, how fast you deliver, or how many people you need, it's a cost line with a nice demo attached.

The line that stuck with me

Slok's sharpest argument wasn't about models at all. "AI may be a great calculator," he wrote, "but in business, intelligence is mostly judgment." He went on to point out that AI doesn't grasp accountability, trust, culture, unwritten rules, ethical trade-offs or fairness — the things that actually shape reputational risk and, eventually, revenue and margins.

Anyone who has shipped software inside a real company recognises this. The hard part was never producing an answer. It was deciding which answer to bet on, who owns the consequences, and how to keep a dozen people aligned while the ground shifts. Those are the hours AI is supposed to give back, and those are precisely the hours it hasn't yet.

A counterpoint worth keeping on the table

Two caveats stop this from being the final word. First, the S&P 493 still contains enormous companies with long procurement cycles — a rollout that takes 18 months to reach the P&L looks identical to no rollout at all in a quarterly margin chart. Second, there's selection bias in the index itself: the firms that invested hardest in AI also tend to be the ones Slok removed.

Still, the direction of travel is consistent with other work. Several recent studies have found AI has yet to deliver significant economy-wide productivity gains, even as spending climbs. The gap between "this model is impressive" and "this quarter was better" is where most of the disappointment lives.

What I'd do with this if you're building

Treat it as a measurement problem, not a verdict on the technology. Pick one workflow where you can name the number before you start — invoices processed per hour, tickets closed without escalation, days from brief to deploy. Ship something narrow enough to finish. Then check the number, honestly, three months later.

AI is genuinely good at compressing the boring middle of a task. It is not good at absorbing blame, reading a room, or deciding which customer to say no to. Build for the first half, keep humans on the second, and don't let a vendor's slide deck stand in for your own baseline.

The takeaway: the value was never going to come from access to a model. It comes from redesigning a specific process around one — and if nobody on your team can point at the metric that moved, you haven't shipped AI yet, you've only paid for it.