Anthropic’s “YOLO” AI Bubble Warning

This is an excerpt from Sources by Alex Heatha newsletter about artificial intelligence and the technology industry, is only distributed to The Verge subscribers once a week.

Dario Amodei took the stage at the DealBook Summit on Wednesday to throw jabs without naming names.

Anthropic’s CEO spent much of the interview with Andrew Ross Sorkin drawing a fine line between his company’s approach and that of a particular competitor. When asked if the AI ​​industry was in a bubble, Amodei separated the “technological side” from the “economic side” and then twisted the knife.

“On the technology side, I feel really strong,” he said. “On the economic side, I have my concerns, even if the technology delivers on all its promises, I think there are players in the ecosystem that, if they make a mistake in timing, they might throw it off a little bit, and bad things could happen.”

Who might these players be? Despite Sorkin’s urging, Amodei did not mention the name of OpenAI or Sam Altman. But he didn’t have to.

“There are some players who enjoy playing,” he said. “Say you’re someone who constitutionally kind of wants to turn things around or just likes big numbers, you might move the dial too far.”

He also touched on “carousel deals,” where chip suppliers like Nvidia invest in AI companies that then spend that money on their chips. Amodei acknowledged that Anthropic has closed some of these deals, although “not on the same scale as some of the other players,” and explained how they can operate responsibly: A new gigawatt data center costs roughly $10 billion over five years to build. The seller invests upfront, and the AI ​​startup pays back its share of the deal as revenue grows.

Although he didn’t mention names again, he did point out the staggering numbers that OpenAI was touting for its computing buildout. “I don’t think there’s anything wrong with that in principle,” he said. “Now, if you start stacking these things so that you have huge amounts of money, and you’re saying, ‘By 2027 or 2028, I need to make $200 billion a year,’ then yeah, you can leverage your capacity.”

The core of Amodei’s argument was the concept he was using internally: the “cone of uncertainty.”

He said Anthropic’s revenues increased tenfold annually for three years, from zero to $100 million in 2023, from $100 million to $1 billion in 2024, and now between $8 billion and $10 billion by the end of this year. (Sam Altman, by comparison, He said (which OpenAI expects to end in 2025 with an annual revenue rate exceeding $20 billion.) But even Amodei doesn’t know whether Anthropic will reach $20 billion or $50 billion next year. “It’s very uncertain.”

He explained that this uncertainty is worrying, because building data centers takes one to two years. Decisions about 2027’s computing needs need to be made now. Buy too little, and you’ll lose customers to competitors. Buy too much, and you risk going bankrupt. “The amount of buffer that is in that cone is basically determined by my margins,” Amodei added.

“We want to buy enough confidence even in the 10th percentile scenario,” he said. “There is always a tail risk. But we try to manage that risk well.” Anthropic’s focus, with higher margins and more predictable revenues, has positioned it as more structurally sound than that of consumer-first companies. “We don’t have to do any code red.”