For two years, “AI bubble” was a phrase you could safely ignore — the province of permabears and reply-guys. That’s no longer true. The people saying it out loud now include one of the most famous short-sellers alive, the founder of the world’s largest hedge fund, and the chief economist of a $700-billion asset manager. You don’t have to agree with them. You should know what they’re actually saying.
Michael Burry bet the whole thing
The “Big Short” investor didn’t just tweet a warning — he deregistered Scion Asset Management in late 2025 and relaunched as a paid Substack, Cassandra Unchained, dedicated to his AI-bubble thesis. Per disclosures reported by Benzinga, as much as 80% of his book sat in put options on Nvidia and Palantir — roughly $1.1 billion in notional exposure, including puts on a million Nvidia shares struck at $110 into 2027. His framing, via Yahoo Finance: this feels like “the last months of the 1999-2000 bubble.” That is not a hedge. That is a man betting his reputation on a date.
Ray Dalio, Torsten Sløk, and the concentration problem
Ray Dalio told Bloomberg TV on June 3, 2026 that his bubble indicators are sitting at “1929 and 2000 levels.” Apollo chief economist Torsten Sløk has been hammering what he calls the “diversification illusion” — the fact that the S&P 500’s returns are lashed to a handful of AI names, with a reported 87% of venture funding now flowing to AI. Burry himself cited Sløk’s data showing AI-linked borrowers accounting for a large share of investment-grade and high-yield debt issuance. The bear case isn’t “the tech is fake.” It’s “the financing is concentrated and circular.”
The AI tool stack actually worth paying for
One email a week. The models, tools and moves that matter, stripped of hype and filtered so you don’t have to drink from the firehose. Free, and you can bail anytime.
The revenue hole nobody has filled
This is the oldest and most stubborn number in the debate. Goldman’s Jim Covello asked in 2024 what trillion-dollar problem AI actually solves to justify the spend. Sequoia’s David Cahn framed the same gap as “AI’s $600 billion question” — the chasm between infrastructure capex and realized revenue. Independent analyst Ed Zitron has spent 2026 documenting the debt side of it, pointing at roughly $178.5 billion of US data-center debt financed increasingly through private credit. Two years on, capex has ballooned past $700 billion a year while consumer AI revenue is still measured in low tens of billions. The hole didn’t close. It got bigger.
The honest read
None of this proves a crash, and every one of these voices has been early or wrong before — Burry especially. As Fortune put it in June, 2026 is “looking like 1999” — which is worth remembering cuts both ways, because 1999 had another year left in it. The point isn’t to panic. It’s that the bubble argument now comes with named authors, dated positions, and hard debt figures attached. That deserves a look before your next allocation, not a reflexive eye-roll.
Not investment advice — I’m a blog, not your financial advisor. These are other people’s positions, laid out so you can judge them yourself.