Even Sam Altman fears an AI bubble
“Investors are too enthusiastic about AI,” says OpenAI CEO
Sam Altman is not the first to warn about the risks of an AI bubble. But his words carry more weight. In mid-August, during a meeting with US journalists, the OpenAI CEO admitted that “investors are too enthusiastic about AI,” comparing the situation to the dot-com bubble of 2000.
He found it “insane” that start-ups with “three people and an idea” could raise tens or even hundreds of millions of dollars, with valuations sometimes exceeding the symbolic billion-dollar mark. “This is not rational behavior,” he said.
Yet, Sam Altman still plans to spend trillions of dollars to reach artificial general intelligence, equaling human capabilities, and eventually superintelligence, which would surpass them.
95% failure rate
Its statements coincide with the publication of a study conducted by MIT. Widely discussed, it claims that 95% of pilot programs implemented in companies turn out to be unsuccessful. Instead, employees prefer to use ChatGPT — often with their personal accounts — for “simple” tasks, but not for “critical” ones, the study explains.
Although some use cases have proven successful, particularly in customer service, many projects to integrate AI into workflows never make it past the POC (proof of concept) stage. In this context, industry players are now betting on AI agents, capable of automating certain complex tasks without human intervention — and thus delivering greater productivity gains for businesses.
Despite all this, generative AI start-ups continue to attract investors. In the first half of the year, they raised more than $100 billion worldwide, according to Crunchbase figures. That’s twice as much as last year. A large part of this money comes from major industry players.
The market reality may soon catch up
But the AI frenzy is also benefiting numerous start-ups, which are securing huge sums just weeks after launching. Most often, this is based only on an idea or the résumé of their founders, without generating any revenue yet. Such is the case of Thinking Machines and Safe Superintelligence, founded by Mira Murati and Ilya Sutskever, two former OpenAI stars, who raised $2 billion and $3 billion respectively.
The market reality may soon catch up with many companies. Several start-ups are already experiencing financial difficulties, forcing them to accept disguised acquisitions by tech giants or scale back their ambitions. And few companies can now afford the costs of training new models capable of competing with those developed by OpenAI, Google, or Anthropic. Even Meta and Apple are considering pulling back.
The market is thus becoming increasingly concentrated around a handful of giants, who keep expanding their reach at the expense of services designed to meet specific needs. For example, OpenAI and Anthropic alone account for nearly 90% of the revenues generated by the 18 largest AI start-ups, according to data collected by The Information.