Yann LeCun is now working on his own startup, but he’s still beefing with Elon Musk.

The AI pioneer — who left Meta last year to co-found AMI Labs, a Paris-based research startup building “world models” as an alternative to large language models — used a recent CNBC interview to deliver a blunt verdict on Elon Musk’s AI company. “XAI is kind of a failure, frankly,” LeCun said, and he didn’t stop there.

His specific charge: the founding team has almost entirely walked out, either through departures or firings, and Musk is now struggling to attract serious AI talent as a result. “Elon is now in a position that it’s very, very difficult for him to kind of hire top people in AI because he’s kind of, you know, not behaving in sort of very good ways toward the previous team,” LeCun said. The upshot, in his telling, is that xAI is left with an enormous GPU infrastructure it can’t fully utilize internally — and so Musk rents it out. “He’s got this huge infrastructure which he rents to other people because that’s the only way he can recoup the cost,” LeCun said. Asked directly whether he believes xAI can compete on the frontier, he was unequivocal: “No, I don’t.”

The numbers behind the co-founder exodus give LeCun’s critique some grounding. By early 2026, all 11 of xAI’s original co-founders had left the company, a cascade that accelerated sharply after SpaceX’s acquisition of xAI in February 2026. The departures included Jimmy Ba, who co-authored the most-cited paper in AI history, and Igor Babuschkin, the chief engineer who came from Google DeepMind. Musk himself acknowledged on X that xAI was “not built right the first time around” and was being “rebuilt from the foundations up” — an unusual admission from a founder whose company had just been valued at $250 billion.

This isn’t the first time LeCun has gone after Musk publicly. Their feud stretches back at least to Memorial Day weekend 2024, when LeCun responded to an xAI job posting by accusing Musk of claiming to want a “maximally rigorous pursuit of the truth” while simultaneously spreading conspiracy theories on his own platform. LeCun took issue with Musk’s AI doom rhetoric too — Musk has long warned that AI could “kill everyone,” a position LeCun considers overblown. Musk’s counter-attack was to question LeCun’s scientific output: “What ‘science’ have you done in the past 5 years?” LeCun came back with a link to his Google Scholar page, citing over 80 papers published since January 2022. Musk’s reply: “That’s nothing, you’re going soft. Try harder!”

The exchange spiraled into a debate about whether scientific publishing even constitutes science, drawing in thousands of commentators. When LeCun described himself as “a scientist, not a business or product person,” Musk shot back that he was “just following orders” — a remark LeCun met with: “You don’t seem to understand how research works.” It was, at least, honest about what kind of argument it was.

LeCun also had a dig at Musk’s management style in 2024, and Musk’s response was essentially to ask who he was — which, given that LeCun’s research on convolutional neural networks underpins virtually every modern computer vision system including Tesla’s own self-driving stack, was a strange move. The two men disagree on almost everything: the risks of AI, the role of academic research, what counts as meaningful scientific contribution, and now, apparently, whether xAI is a viable company at all.

Meanwhile, LeCun has his own problems to solve. AMI Labs raised $1.03 billion in seed funding at a $3.5 billion valuation in March 2026, backed by the likes of Nvidia, Samsung, Eric Schmidt, and Jeff Bezos. The company is building on LeCun’s JEPA architecture — a fundamentally different approach to AI that learns abstract representations of reality rather than predicting the next token. LeCun serves as executive chairman, with Alexandre LeBrun as CEO, and the startup’s own commercial timeline is measured in years rather than quarters.

The irony isn’t lost: the man calling xAI a failure runs a company that, by its own CEO’s admission, won’t have anything ready to use for at least a year. LeCun had also been at the helm of Meta’s AI efforts, which spiralled into irrelevancy last year before Mark Zuckerberg stepped in to rejig the team. But LeCun’s pitch is that he’s betting on a different paradigm entirely, one that doesn’t involve scaling up LLMs indefinitely. Whether that bet lands is an open question. But when it comes to xAI, his read on the co-founder departures tracks with what’s actually happened — whatever Musk is rebuilding, he’s doing it largely without the people who built it the first time.