Anthropic CEO Dario Amodei speaks onstage during the 2025 New York Times Dealbook Summit on Dec. 3, 2025 in New York City. Photo by Michael M. Santiago/Getty Images
Dario Amodei, head of Anthropic, is concerned that some A.I. firms may be playing fast and loose with the staggering sums they’re spending on data centers and compute power. “I think there are some players who are ‘YOLO’ -ing, who pull the risk dial too far,” said the CEO during The New York Times’ Dealbook Summit today (Dec. 3).
Even Anthropic, which Amodei decribed as more “conservative” than its rivals, recently pledged to invest $50 billion in building out data centers across the U.S. Others are spending far more. OpenAI, for example, has struck data center and GPU deals worth more than $1 trillion in 2025 alone.
A.I. startups are increasingly caught in a delicate balancing act: the long timelines needed to build data centers versus the uncertainty surrounding the technology’s ultimate economic payoff. Navigating these variables comes with “some amount of irreducible risk,” said Amodei, who, without naming any specific companies, warned that not all players are managing that risk responsibly.
“Even if the technology is really powerful and fulfills all its promises, I think that with some players in the ecosystem, if they get it off by a little bit, bad things could happen,” he said.
Like its rivals, Anthropic has leaned more heavily into circular financing deals with chipmakers and cloud providers. Under these arrangements, hardware companies invest in A.I. model developers, who then use that funding to purchase their compute products. The deals have raised eyebrows across Silicon Valley about whether they are sustainable for companies like Anthropic and OpenAI, which are currently valued at $183 billion and $500 billion, respectively, but are not yet profitable.
There’s nothing inherently “inappropriate” about such deals, according to Amodei. “You can overextend yourself, of course,” he added, noting that companies must balance the danger of spending too much on compute with the risk of not acquiring enough to serve customers.
One area where the executive has fewer concerns is competition. Earlier this week, OpenAI chief Sam Altman reportedly declared an internal “code red” to improve ChatGPT after Google’s newest Gemini model surpassed the chatbot on benchmark tests. The release didn’t perturb Anthropic,said Amodei, who emphasized that his company targets a different market and is focused primarily on enterprise rather than consumer products.
“I’m just very grateful that Anthropic is taking a different path,” he said. “We don’t have to do any code reds.”
