OpenAI: The Lehman Brothers of AI? Ed Zitron's thesis

The AI bubble burning billions before it collapses

Ed Zitron has spent months repeating what Silicon Valley doesn't want to hear: artificial intelligence is not an industrial revolution, it's a bubble financed with debt and expectations. The British critic—host of the Better Offline podcast, with over a million downloads a month, and author of the Where's Your Ed At newsletter—argues that OpenAI could be "the Lehman Brothers" of the sector. His upcoming book, "The Hater's Guide To Silicon Valley", arrives in the second quarter of 2027. The question is whether by then there will be anything left to write about.

What the critic who accuses AI of being a mirage claims

Zitron's diagnosis, issued from his research firm EZPR, is that OpenAI and Anthropic burn billions with no visible return, that the boom is largely fake, and that the bust won't stay within the tech sector. He speaks of dragging down the entire economy.

The uncomfortable part is that across the ring there are no slogans, there are figures. It's noted that Anthropic may have closed a month or quarter in profit; with the detailed accounting in front of us, no one has yet confirmed with what calendar tricks. Meanwhile, the price of AI already trades as if the future were signed.

The balance sheets that refute the bubble: orders through 2031

The most annoying argument for doomsayers doesn't come from a sales brochure. It comes from the order book. Nvidia, Seagate, SanDisk, and Micron have accumulated production commitments with caps marked through 2031, according to semiconductor and memory balance sheet tracking. Quick translation: demand for data center components won't dry up next year.

On Thursday, Oracle beat expectations again, amortizing its investment with returns from its clients in its own data centers and with double-digit growth forecasts. And the multiples of Microsoft, Amazon, Alphabet, Meta, Apple, Micron, Western Digital, or Broadcom are not those of a company in a tailspin: several are paying off long-term debt they had been carrying for years.

Future purchase orders: the emperor's new clothes of 2008

The most repeated parallel is not with the dot-coms, but with 2008. The real estate data in 2006 and 2007 were equally spectacular, and came from the same kitchen: it used to be called asset securitization and distributed risk; now it's called future purchase orders. The structure holds as long as no one touches the stick that supports it.

The underlying suspicion is circular economics: tech companies buying capacity from each other to inflate results, borrowing more and more, and asking the government to guarantee loans for their mega data centers. Pure hype, in that reading, with little real business base behind it.

The scenario gaining weight is not the apocalypse, but rerating: a return to logical valuations that wipes out half the market cap. The other half is at stake the day some player has to close up shop. No one disputes the orders through 2031. Whether that demand survives the first bankruptcy, no one is willing to bet on either.