Anthropic’s AI bubble ‘YOLO’ warning
This is an excerpt from Sources by Alex Heath, a newsletter about AI and the tech industry, delivered only to The Verge subscribers once a week.
Dario Amodei took the stage at the DealBook Summit on Wednesday to throw punches without calling names.
The Anthropic CEO spent much of the interview with Andrew Ross Sorkin drawing a careful line between his company’s approach and that of a certain competitor. When asked if the AI industry was in a bubble, Amodei separated the “technology side” from the “economic side” and then twisted the knife.
“Technologically, I feel really solid,” he said. “Economically, I have concerns because, even if the technology delivers on all its promises, I think there are players in the ecosystem who, if they make a mistake in timing, they catch up a little bit, and bad things could happen.”
Who could these players be? Despite Sorkin’s urging, Amodei did not want to name OpenAI or Sam Altman. But he didn’t have to.
“There are players who YOLO,” he said. “Let’s say you’re a person who constitutionally just wants to do YOLO things or just likes big numbers, then you might turn the dial too far.”
He also talked about “circular deals,” where chip suppliers like Nvidia invest in AI companies who then spend those funds on their chips. Amodei acknowledged that Anthropic has done some of these deals, but “not on the same scale as some other players,” and explained how they can work responsibly: A new one-gigawatt data center costs about $10 billion to build over five years. A supplier invests up front and an AI startup pays back its share of the deal as revenue increases.
While once again he didn’t name names, he did reference the eye-popping numbers that OpenAI has touted for its compute build. “I don’t think there’s anything wrong with that in principle,” he said. “Now if you start stacking them up where they’re getting huge amounts of money, and you say, ‘By 2027 or 2028, I have to make $200 billion a year,’ then yes, you can overextend yourself.”
The heart of Amodei’s argument was a concept he uses internally: the “cone of uncertainty.”
He said Anthropic’s revenue has grown tenfold every year for three years, going from zero to $100 million in 2023, from $100 million to $1 billion in 2024, and now between $8 billion and $10 billion at the end of this year. (Sam Altman, by comparison, has said that OpenAI expects to end 2025 with annualized revenue above $20 billion.) But even Amodei doesn’t know whether Anthropic will hit $20 billion or $50 billion next year. “It’s very uncertain.”
This uncertainty is concerning, he explained, because data center construction takes one to two years. Decisions on computing needs for 2027 must be made now. Buy too little and you lose customers to competitors. Buy too much and you risk bankruptcy. Amodei added: “How much buffer there is in this cone is basically determined by my margins. »
“We want to buy enough to have confidence, even in the 10th percentile scenario,” he said. “There is always tail risk. But we try to manage that risk well.” He positioned Anthropic’s business focus, with higher margins and more predictable revenue, as structurally safer than that of consumer-focused companies. “We don’t need to do a code red.”
