Anthropic Chief Dario Amodei warns of AI bubble as competition with OpenAI sharpens

Anthropic CEO Dario Amodei offered a measured assessment of whether the artificial intelligence sector is entering bubble territory during an appearance at The New York Times DealBook Summit on Wednesday, while also taking a pointed swipe at an unnamed rival widely understood to be OpenAI. He stated that the situation was too complex for a straightforward yes-or-no answer and instead laid out his broader view of AI economics, stressing both optimism about the technology’s long-term potential and concern that some industry players may be misjudging the timing of economic returns.

Amodei said he remained bullish about AI’s promise but cautioned that companies could make timing errors or face adverse outcomes as they chase future gains. He stated that there was inherent risk given the uncertainty around when economic value would materialise, and emphasised that firms were compelled to take risks not only to remain competitive but also to counter authoritarian adversaries — a reference highlighted in reports as pointing to China. However, he argued that some competitors were not managing their exposure responsibly and were taking what he described as unwise risks.

He stated that much of the dilemma stemmed from the unpredictability of how quickly AI’s economic value would expand and how to align that with the long lead times required to construct new data centres. He said this created a genuine challenge that Anthropic was attempting to handle responsibly, while warning that other companies were effectively “YOLO-ing”, pulling the risk dial too far, and this, he said, was deeply concerning.

Amodei also addressed a separate industry pressure point: the depreciation timelines of AI chips. He informed the audience that the chips themselves continue to function for many years, but their value drops as newer, faster and cheaper versions enter the market. He said Anthropic was taking conservative assumptions on this and other fronts as it plans for an uncertain future.

The CEO highlighted Anthropic’s rapid revenue growth, noting that the company’s earnings expanded tenfold annually over the past three years — from nothing to $100 million in 2023, then from $100 million to $1 billion in 2024, with projections landing between $8 billion and $10 billion this year. He stated that it would be unwise to assume this trajectory would simply continue, adding that the outlook could vary wildly and that such uncertainty was disconcerting. He said he therefore planned on the lower end of expectations.

Amodei explained that AI companies must decide years in advance how much computing capacity they will need and how much to invest in data centres. Underestimating demand could leave them unable to serve customers, while overestimating could saddle them with unsustainable costs or even push them towards bankruptcy. His remarks followed reporting last month that OpenAI faced a public relations crisis after its CFO suggested the US government should “backstop” the company’s infrastructure loans — a suggestion she later retracted.

Without naming individuals, Amodei warned that those inclined towards bold or impulsive risk-taking could overextend themselves, a comment reports noted as an implicit reference to OpenAI CEO Sam Altman, as reported by TechCrunch. He concluded by saying that Anthropic believed it would remain stable in almost all scenarios, but he could not make the same assurances for others in the industry.