Dario Amodei, the CEO and co-founder of the artificial intelligence lab Anthropic, offered a stark assessment of the current AI frenzy on Tuesday, suggesting that some competitors are taking reckless financial and safety risks—or “YOLOing”—in the race to build advanced models.
Speaking at The New York Times DealBook Summit in an interview with Andrew Ross Sorkin, Mr. Amodei described a delicate high-wire act facing the industry: the need to commit billions of dollars to infrastructure years before knowing if the revenue will materialize to justify it.
While Anthropic, the maker of the Claude chatbot, has seen its own revenue skyrocket, Mr. Amodei cautioned that the massive capital expenditures required to build data centers create a “cone of uncertainty” that could threaten the stability of major players.
“There is a real dilemma deriving from uncertainty in how quickly the economic value is going to grow, and the lag times on building the data centers,” Mr. Amodei said. “I think there are some players who are YOLOing… who pull the risk dial too far, and I am very concerned.”
Mr. Amodei, a former executive at OpenAI, did not name specific competitors, though the industry is currently defined by an arms race between Anthropic, Google, and OpenAI to achieve dominance. He noted that the lead time for building compute capacity is roughly “a year or two,” forcing CEOs to make massive bets on future demand.
“If I buy too much compute, of course, I might not get enough revenue to pay for that compute,” he said. “In the extreme case, there is kind of the risk of going bankrupt.”
Despite these warnings, Anthropic’s own growth trajectory has been historic. Mr. Amodei revealed that the company’s revenue has effectively grown “10x a year every year for the last three years.” He detailed the trajectory: “Zero to 100 million in 2023, 100 million to a billion in 2024, one billion to… it’s going to land somewhere between eight and ten at the end of this year.”
Scaling Laws and National Security
Throughout the interview, Mr. Amodei adhered to the theory of “scaling laws”—the idea that simply adding more data and computing power to AI models will continue to yield massive improvements in intelligence without needing a radical scientific breakthrough.
“The thing that is most striking about all of it is, as you train these models in this very simple way… they get better and better at every task under the sun,” Mr. Amodei said. He predicted that eventually, these models will resemble “a country of geniuses in a data center.”
This potential for “singular capability” drove Mr. Amodei’s forceful defense of United States export controls preventing the sale of advanced AI chips to China. Despite Anthropic’s partnership with chipmaker Nvidia, Mr. Amodei argued that providing authoritarian regimes with cutting-edge hardware poses an existential threat to democratic nations.
“If it’s plopped down in an authoritarian country, I feel like they can outsmart us in every way… intelligence, defense, economic value,” Mr. Amodei said. “I worry that they’ll be able to oppress their own people, that they’ll be able to have a perfect surveillance state.”
He added, “We need to beat them, but we need to not do the things that would cause us to become them.”
The Economic Impact and the Future of Work
Mr. Amodei was candid about the potential for artificial intelligence to disrupt the labor market, particularly regarding entry-level jobs. He acknowledged that while AI can make humans more productive, companies face a choice between innovation and pure efficiency.
“They can increase efficiency by basically having AIs do what humans used to do… insurance claims processing or know-your-customer,” he noted. “I think we’ll need a lot less humans for them.”
While he expressed optimism that society can adapt, calling the potential upside of the technology “machines of loving grace,” he admitted that the free market alone might not suffice to handle the transition.
“I think the government is going to need to step in,” Mr. Amodei said, suggesting that significant wealth redistribution or tax policy changes may eventually be required. “The structure of a society that has built powerful AI is just going to have to be different.”