AI godfather warns: Musk's xAI has "failed", AI industry may face a "big bubble burst"

LeCun currently serves as the founder of AMI Labs, and in an interview with CNBC, he bluntly stated that the core founding team of xAI has largely left. Musk is now finding it extremely difficult to recruit top AI talent, and xAI will be unable to compete with OpenAI and Anthropic in the forefront of the AI field. At the same time, he pointed out that the operating costs of AI services are not decreasing as quickly as expected, and major labs are currently relying on investors' funds to subsidize user usage, a model that is difficult to sustain. LeCun's statement once again raised doubts in the market about the valuation of top AI companies, and poured cold water on the current hot investment enthusiasm in the AI industry. His own AMI Labs, founded in March of this year, just completed a $1 billion financing round, with a pre-financing valuation of $3.5 billion, focusing on what he believes represents the next phase of AI research, the "world model." The "failure theory" of xAI: founding team leaves, talent dilemma is hard to solve LeCun's criticism of xAI directly points to its talent base. "To be honest, xAI is already to some extent a failure because the founding team has already left," he said. Over the past year, several co-founders of xAI have left one after the other. LeCun believes that Musk's previous treatment of the team has greatly reduced its attractiveness in the top AI talent circle. "Elon is now in a very, very difficult situation, it is difficult to recruit top AI talent because the way he treated the former team... was not very good." On the business side, xAI is also under pressure. In February of this year, Musk merged SpaceX with xAI, valuing xAI at a staggering $1.25 trillion. However, according to financial data, the AI business sector (including xAI) under SpaceX operated at a loss of $2.5 billion in the quarter ending March 31. LeCun also pointed out that xAI has two large data centers, Colossus 1 and Colossus 2, in Memphis, and rents out computing power to companies like Google and Anthropic, "because that is the only way he (Musk) can recover his costs." He expressed that he is "not optimistic" about the future of xAI. The unsustainable gap between costs and revenues LeCun has deeper doubts about the business model of the entire AI industry. He pointed out that while the price of AI services is increasing, the rate of decrease in operating costs is not fast enough, leading to continued losses for major companies, with users essentially being subsidized by investors. "This situation is not sustainable, right?" he said. He warned that top labs like OpenAI and Anthropic are facing three paths: raise prices, cut costs, or face a "great bubble burst." This assessment is in line with concerns within the industry. According to CNBC, OpenAI CEO Sam Altman stated in a live broadcast this month that enterprise customers are increasingly discussing AI spending issues, calling AI costs "a huge issue." Technical route dispute: Limitations of LLM and bet on "world models" LeCun is reserved about the current mainstream AI technology route. He has long criticized the limitations of large language models (LLM) - LLM predicts the next word by learning the rules of language, performing well in reasoning and programming fields, but he believes this architecture cannot support a truly general AI system. The "world model" route that he advocates is completely different, aiming to enable AI to understand the way the real or simulated world operates, covering objects, causal relationships, and action logic. "I personally believe that without a world model, we cannot have a general and reliable intelligent system," LeCun said. Currently, from Anthropic to OpenAI, major AI companies are heavily betting on AI agents - systems that can autonomously perform complex tasks. LeCun acknowledges the practical value of LLM in programming, mathematics, and other fields, but emphasizes that "the cost of running these systems at this level of performance is much higher than the amount users are willing to pay," and this contradiction is at the core of the industry's bubble risk. AMI Labs' $1 billion financing completed in March is a major bet by LeCun on the world model route, and a new footnote to his divergence from the current mainstream AI narrative. This article was reprinted from the WeChat public account "Wall Street Seehearings," author: Bu Shuqing; GMTEight editor: Song Zhiying.