The IMF recently warned that an AI bubble bursting could damage the global economy. Commentator Ed Zitron argues the warning deserves more attention than it’s getting, and explains why, in blunt terms, this looks a lot like a bubble built on belief rather than evidence. These are his views and estimates, not settled fact, but they’re worth understanding.
Why The Warnings Keep Getting Ignored
Zitron’s view is that people are numb to these warnings because every one of them gets followed by an equally confident headline saying AI will be the biggest economic driver in history. Faced with a choice between fear and optimism, most people choose optimism, especially when very wealthy people keep telling them to.
The Concentration Problem
Nine companies now make up over 45% of the Nasdaq 100’s value. During the dot-com bubble, the equivalent concentration was around 75%, but the market was far smaller then, and far fewer ordinary people held shares directly or through funds. Today, retail investors hold an estimated $4.65 trillion in AI-related stocks. Zitron’s calculation: a 20% drop across the most concentrated AI stocks would trigger the worst single trading day since the dot-com crash, and it might not even take that much, because Nvidia alone makes up around 8% of total market value.
Nobody Can Actually Measure The Return
The core problem, in Zitron’s words, is that nobody can calculate the return on all this invested capital yet. Companies don’t disclose how much revenue AI actually generates. Meta’s revenue went up and people called that proof the AI bet is working, but Meta doesn’t break out AI revenue separately, so there’s no way to check that claim.
Pension Money Is Already In It
Perhaps the most concerning point: pension and insurance funds are already funding data centre debt through private credit arrangements, and the scale of that exposure isn’t publicly disclosed. Zitron estimates that around a third of the funding at one major private credit firm comes from public pension systems, and calls on pension boards like California’s CalPERS and CalSTRS to review and publish their holdings.
Watch What Banks Do, Not What They Say
Banks that publicly praise AI companies often hold direct financial stakes in them. Zitron’s advice: ignore the public statements entirely and look instead at what banks are actually doing with debt, because that’s harder to dress up. Several major banks have reportedly been quietly offloading data centre debt they bought only months earlier, a sign they may be nearing internal limits on how much risk they’re willing to hold in one sector.
How It Might Actually Unwind
Zitron doesn’t expect a sudden single collapse. He expects a specific trigger, most likely a major data centre project falling through on a force majeure, after which banks and private credit firms could pull back sharply and quickly. Automated trading funds, which already hold trillions of pension dollars, would likely amplify any sudden move in either direction.
Whatever happens to the market, the sensible response for a small business is to build practical skills with the tools rather than bet on the hype. Claude LaunchPad is the free starting point for doing that properly.
