OpenAI’s Sam Altman reportedly announced a ‘code red’ after Google revealed its new Gemini 3 model, setting off fresh tension in the race to stay ahead. Just as this drama gathered attention, Dario Amodei, the CEO of Anthropic, stepped into the spotlight with pointed remarks that seemed aimed at his fast-spending rivals. While he didn’t name anyone, he still raised questions about the recently created rush and the huge bets some companies are making. He showed his concerns and warned that a single wrong move could bring heavy losses. His comments landed at a moment when pressure around competition was already rising, making his subtle swipe feel even sharper.
Speaking at The New York Times DealBook Summit, Dario raised concerns about the growing financial risk in the fast-moving AI sector. He explained that while the technology behind AI is improving quickly, the money side of the industry is entering uncertain territory.
He further added that companies are spending huge amounts to build large data centres that will take years to complete. These centres are needed to train and run powerful AI systems, yet no one can confidently predict how much demand there will be by the time they are ready.
Amodei further added that the gap between future needs and present spending has created a wide “cone of uncertainty” that could damage companies if they get their calculations wrong.
During his appearance at the event, Amodei drew a careful line between Anthropic’s approach and that of unnamed rivals. He did not mention OpenAI directly, but his comments appeared aimed at competitors that are moving very fast and spending heavily. He said some companies are “YOLOing”, meaning they are taking big financial bets without enough caution. According to him, even a small mistake in timing could cause serious losses.
Also read: Motorola Edge 70 India launch date, specifications, price and everything else we know so far
Amodei also highlighted the trend of ‘circular deals’, where chip makers invest in AI firms that then use that money to buy more of their chips. He admitted that Anthropic has taken part in such arrangements but on a smaller level. He added that these deals can work well only if the numbers remain realistic. For example, building a new data centre of a large size can cost around 10 billion dollars over five years. If companies start assuming future earnings of more than 200 billion dollars a year to justify such spending, they could be putting themselves in danger.
He said that the bigger issue is predicting demand. Data centres take one to two years to build, which means that the decisions for the year 2027 must be taken now. If a company buys too little computing capacity, it will lose customers. If it buys too much, then it would be hard for them to survive any financial hits. This is termed as the ‘cone of uncertainty’ that Amodei says every major AI company is currently facing.
Also read: Sam Altman quietly explored buying a rocket company to rival Elon Musk’s SpaceX
He went on further and explained that Anthropic’s revenue has grown very quickly in the last three years, reaching up to 10 billion dollars in 2024. However, even with the strong growth, he still cannot predict whether the company will make 20 billion or 50 billion dollars the upcoming year. This unpredictability is one of the concerning factors making him believe that careful planning is essential for the AI companies.
During the final wordings Amodei said that Anthropic is trying to manage its risks by focusing on business customers, which usually provide more stable income. He stressed that the company aims to avoid sudden or extreme decisions and instead follow a steady path as the AI industry continues to expand.