Dario Amodei, head of Anthropic, worries that some AI companies investing huge sums in data centers and computing power could be erratic. “I think some players are ‘YOLO’ and they’re raising the stakes too high,” the CEO said today (December 3) at the New York Times Dealbook Summit.
Even Anthropic, which Amodei says is more “conservative” than its rivals, recently committed $50 billion to building data centers across the United States. Other companies pay much higher. OpenAI, for example, has struck over $1 trillion worth of data center and GPU deals in 2025 alone.
Artificial intelligence startups are increasingly caught in a delicate balance: the lengthy time required to build data centers and the uncertainty of the technology’s ultimate financial returns. Navigating these variables comes with “a degree of irreducible risk,” Amoudi said. He warned that not all players were managing this risk responsibly, but he did not name any specific companies.
“Even if the technology is really powerful and delivers on all of its promises, I think for some players in the ecosystem, if they stray just a little bit, bad things could happen,” he said.
Like its rivals, Anthropic prefers revolving financing deals with chipmakers and cloud providers. Under these arrangements, hardware companies invest in AI model developers, who then use the funds to purchase their computing products. The deals have raised concerns 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 have yet to turn a profit.
Amoudi said there was nothing inherently “inappropriate” about such transactions. “Of course, you can overextend,” he added, noting that companies must balance the dangers of spending too much on computing against the risk of not getting enough resources to serve customers.
One area where executives are less concerned is competition. Earlier this week, OpenAI CEO Sam Altman reportedly announced an internal “code red” to improve ChatGPT after Google’s latest Gemini model surpassed the chatbot in benchmark tests. Amodei said the announcement did not faze Anthropic, stressing that his company is targeting a different market, focusing primarily on enterprise rather than consumer products.
“I’m very grateful that Anthropic has taken a different path,” he said. “We don’t have to do any code red.”
