Mark Zuckerberg de Meta says that it is “definitely a possibility” that we are in a bubble AI

Deutsche Bank called him “summer AI has become ugly”. For weeks, with each new proof that companies failed to adopt the AI, the fears of a bubble of AI have intensified, fueled by the realization of the way in which the S&P 500 has increased, as well as warnings from the main leaders of the industry. A August MIT study revealed that 95% of AI pilot programs fail to offer a return on investment, despite more than 40 billion dollars paid in space. Just before the MIT report, the CEO of Openai, Sam Altman, sounded the AIA bubble bubbles, expressing his concerns about the overvaluation of certain AI startups and the intensity of investors’ enthusiasm. These trends have even drawn the attention of the president of the Fed, Jerome Powell, who noted that the United States witnessed “unusual quantities of economic activity” in the creation of AI capacities.
Mark Zuckerberg has similar thoughts.
The meta-PDG has acknowledged that rapid development and rise in investments in AI should form a bubble, potentially exceeding productivity and practical yields and risking a market accident. But Zuckerberg insists that the risk of over-investment is preferable to the alternative: to be late that it considers a technological transformation defining the era.
“There are convincing arguments to explain why AI could be an aberrant value,” covered Zuckerberg during an appearance on access podcast. “And if the models continue to grow in capacity from one year to another and demand continues to grow, then there may be no collapse.”
Then Zuckerberg joined the Altman camp, saying that all capital spending bubbles such as the construction of the IA infrastructure, largely seen in the form of data centers, tend to end in a similar way. “But I think there is certainly a possibility, at least empirically, on the basis of the large buildings of past infrastructure and the way they led to bubbles, that something like that would happen here,” said Zuckerberg.
Bubble echoes
Zuckerberg highlighted the past bubbles, namely the railways and the Dot-Com bubble, as key examples of the infrastructure accumulations leading to a collapse of the stock market. In these cases, he said that bubbles had occurred due to companies which contract too many debts, macroeconomic factors or the decrease in products, which led companies to cope and to leave precious assets.
The comments of the meta-PDG echoes Altman, who also warned that the Boom of AI shows many signs of a bubble.
“When bubbles occur, intelligent people are excited about a core of truth,” said Altman The penisAdding that AI is this nucleus: transformer and real, but often surrounded by irrational exuberance. Altman also warned that “cash frenzy chasing everything that labeling” AI “” can cause swollen assessments and a risk for many.
The consequences of these bubbles are expensive. During the DOT-COM bubble, investors paid money into technological startups with unrealistic expectations, motivated by media threw and a frenzy for new Internet companies. When the results dropped, the stocks involved in the DOT-COM bubble lost more than 5 billions of dollars in total market capitalization.
An AI bubble should have similar economic impacts. In 2025 alone, the largest American technological companies, including Meta, spent more than $ 155 billion on AI development. And, according to Statita, the current AI market value is around 244.2 billion dollars.
But, for Zuckerberg, losing the potential of AI is a much greater risk than losing money in an AI bubble. The company recently hired at least $ 600 billion in American data centers and infrastructure until 2028 to support its AI ambitions. According to Meta’s financial director, this money will go to all the buildings of the American data center of the American giant and to national commercial operations, including new hires. Meta also launched its superintendent laboratory, recruiting talents aggressively with job offers of several million dollars, to develop an AI that surpasses human intelligence.
“If we end up mistreating a few hundred billion dollars, it will be very unhappy of course. But I would say that the risk is higher on the other side,” said Zuckerberg. “If you build too slowly, and the superintendent is possible in three years, but you have built it, assume that it would be there in five years, then you are out of position on what I think is the most important technology that allows the creation of new products and innovation and value in history.”
Although it sees the consequences of not being aggressive in the investment in AI winning it on overinvestment, Zuckerberg has recognized that Meta’s survival does not depend on the success of the AI.
For companies like Openai and Anthropic, he said: “There is obviously this question open to what extent they will continue to collect funds, and it depends both to a certain extent on their performance and the way in which AI, but also on all these macroeconomic factors which are out of control.”
https://fortune.com/img-assets/wp-content/uploads/2025/09/GettyImages-2235448475-e1758309777180.jpg?resize=1200,600