Anthropic's AI bubble 'YOLO' warning

Dario Amodei takes indirect shots at OpenAI's ‘YOLOing’ and big, circular deals. Also: More from the DealBook Summit and Meta's big hire.

Dec 03, 2025

I’m sending this issue from The New York Times DealBook Summit in Midtown Manhattan, where I’ve been hanging out and watching the onstage interviews all day. It has been nice running into some of you. I’m headed back to the West Coast tomorrow.

(Credit: Getty Images for The New York Times)

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

The Anthropic CEO spent a good chunk of the interview with Andrew Ross Sorkin drawing a careful line between his company’s approach and that of a certain competitor. When asked about whether the AI industry is in a bubble, Amodei separated the “technological side” from the “economic side” and then twisted the knife.

“On the technological side, I feel really solid,” he said. “On the economic side, I have my concerns where, even if the technology fulfills all its promises, I think there are players in the ecosystem who, if they just make a timing error, they just get it off by a little bit, bad things could happen.”

Who might those players be? Despite Sorkin’s prodding, Amodei wouldn’t name OpenAI or Sam Altman. But he didn’t have to.

“There are some players who are YOLOing,” he said. “Let’s say you’re a person who just kind of constitutionally wants to YOLO things or just likes big numbers, then you may turn the dial too far.”

He also touched on “circular deals,” where chip suppliers like Nvidia invest in AI companies that then spend those funds on their chips. Amodei acknowledged that Anthropic has done some of these deals, though “not at the same scale as some other players,” and walked through the math of how they can work responsibly: A new gigawatt data center costs roughly $10 billion to build over five years. A vendor invests upfront, and an AI startup pays back its share of the deal as revenue grows.

While he again didn’t name names, he referenced the eye-popping numbers OpenAI has been trumpeting for its compute buildout. “I don’t think there’s anything wrong with that in principle,” he said. “Now, if you start stacking these where they get to huge amounts of money, and you’re saying, ‘By 2027 or 2028 I need to make $200 billion a year,’ then yeah, you can overextend yourself.”

The cone of uncertainty

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

He said that Anthropic’s revenue has grown tenfold annually for three years, from zero to $100 million in 2023, $100 million to $1 billion in 2024, and now somewhere between $8 billion and $10 billion by this year’s end. (Sam Altman, by comparison, has said that OpenAI expects to end 2025 with an annualized revenue run rate exceeding $20 billion.) But even Amodei doesn’t know if Anthropic will hit $20 billion or $50 billion next year. “It’s very uncertain.”

That uncertainty is concerning, he explained, because data centers take one to two years to build. Decisions on 2027 compute needs have to 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 that cone is basically determined by my margins.”

“We want to buy enough that we’re confident even in the 10th percentile scenario,” he said. “There’s always a tail risk. But we’re trying to manage that risk well.” He positioned Anthropic’s enterprise focus, with higher margins and more predictable revenue, as structurally safer than that of consumer-first businesses. “We don’t have to do any code reds.”

Overheard at the DealBook Summit:

“I think scaling is going to get us there.” - Dario Amodei, answering the question I gave Sorkin to ask about whether he thinks major breakthroughs on the order of the transformer will be needed to get to AGI.
“My YouTube is just a bunch of educational content. This morning I watched a video about how trees grow.” - Mr. Beast
“We’re both absorbing the full risk of each other’s splendor.” - Alex Karp to Sorkin during one of the Palantir CEO’s many onstage diatribes.
“Prediction markets are a big deal.” - Coinbase CEO Brian Armstrong, followed by BlackRock CEO Larry Fink saying, “This is not how I’m going to live my life.”
“I don’t read The New York Times anymore… The New York Times is no longer the paper of record.” - Treasury Secretary Scott Bessent onstage at a New York Times event.

A MESSAGE FROM OUR SPONSOR

Finance built for speed and control.

When it comes to managing company spend, founders and finance leaders often have to choose between financial responsibility and operational speed. Brex breaks that tradeoff. Brex is the intelligent finance platform that helps companies spend smarter and move faster. With Brex, you get corporate cards with built-in expense management plus a team of AI agents that handle manual finance tasks for you. That means more time back for you to focus on the business.

Learn more.

Sources on Squawk Box

I joined the CNBC Squawk Box set this morning to talk about OpenAI’s code red, competition from Google, and why Joe Kernen can’t get the chatbot to give him the answers he wants. You can watch the whole segment on CNBC’s website or directly above.

Also: Thanks to Highwire’s Robert Smith for interviewing me about Sources and ACCESS: “The Continued Rise of New Media: Alex Heath on Sources, podcasting, and pitching”

Feed check

Meta hired Apple’s head of design: This is a significant hire and a bad look for Apple. As Bloomberg reported, and I can also confirm, Alan Dye, Apple’s head of user interface design, is joining Meta as chief design officer for Reality Labs. The move suggests that Meta is quite serious about improving the quality of the software powering its growing lineup of devices.
The FT scooped Anthropic’s IPO plans: “Anthropic has tapped law firm Wilson Sonsini to begin work on one of the largest initial public offerings ever, which could come as soon as 2026, as the artificial intelligence start-up races OpenAI to the public market.”
OpenAI is buying Neptune in an all-stock deal: “Neptune makes software that helps OpenAI and other customers analyze training runs and find issues when developing artificial intelligence models. The ChatGPT maker has used Neptune’s tools for more than a year to conduct multiple experiments and compare different versions.”
Nvidia’s CFO told investors: “We still haven’t completed a definitive agreement” with OpenAI for the $100 billion investment announced in… September.
Waymos have started driving more like humans in San Francisco: “Waymo has been trying to make its cars ‘confidently assertive,’ says Chris Ludwick, a senior director of product management with Waymo, which is owned by Google parent Alphabet. ‘That was really necessary for us to actually scale this up in San Francisco, especially because of how busy it gets.’”
The New York Times profiles startups that are recreating websites to train agents: “Mr. Garg’s company, AGI, is among a number of Silicon Valley start-ups that have spent the past several months recreating popular websites so that A.I. systems can learn to navigate the internet and complete specific tasks on their own, like booking flights. If an A.I. system learns to use a replica of United.com, it can use the real site, too.”

Audience survey

I’m thrilled to be a launch partner for Alltogether, a new initiative to partner brands with leading voices in independent journalism. Subscriptions will always be the backbone of my business, but advertising can also play a critical role in my growth. I’m a fan of Alltogether’s approach because it respects not only my readers but also my commitment to editorial independence.

Part of this effort involves getting to know my subscribers better. Please take a few minutes to complete this audience survey. Results are anonymous and not tied to your email. Thank you so much for your support!