AMI Labs founder Yann LeCun thinks the AI boom is built on shaky economics and Musk’s xAI is already showing signs of wobbling.
LeCun, one of the so-called godfathers of AI and Meta’s former chief AI scientist, took aim at the current AI frenzy in a CNBC interview.
He was not buying the business model or the technology behind today’s leading AI outfits. He suggested the industry may be heading for a correction once investors get bored with subsidising everyone’s chatbot habit.
“xAI is kind of a failure, frankly, because the founding team has” departed, LeCun said.
Several co-founders have left since xAI launched, leaving the outfit competing for scarce AI talent while trying to pretend churn is just another growth metric.
“Elon is now in a position that is very, very difficult for him to kind of hire top people in AI, because he’s kind of, you know, not behaved in sort of very good ways toward the … previous team,” LeCun said.
That is not a great look for xAI boss Elon Musk, who has been scaling the company with the subtlety of a rocket launch.
Earlier this year, Musk merged xAI with SpaceX in a deal valuing the combined operation at $1.25 trillion. The plan leans heavily on vast computing infrastructure, including the Colossus 1 and Colossus 2 data centres in Memphis.
Those facilities were built for large-scale AI training, but LeCun said they are now moonlighting as rented compute.
xAI has “huge infrastructure” that it rents out to other companies, because that’s the only way he [Musk] can recoup the cost.”
Google and Anthropic have reportedly tapped that capacity, which says plenty about how scarce and expensive AI compute has become.
The numbers remain ugly. In the first quarter, SpaceX’s AI segment, which includes xAI, posted a $2.5 billion operating loss.
That is not just a Musk problem. It points to the wider AI industry’s awkward little habit of burning cash faster than it can invent pricing tiers.
“The prices of those AI services are going up, but the cost of running them is going down, but not nearly fast enough. And so all of those companies are losing money, and basically, the use for most people is funded by the investors. That can’t go on for a very long right?” he said.
LeCun thinks the bill is coming due.
“Labs like OpenAI and Anthropic are going to have to increase prices, they’re going to have to cut costs, or there’s going to be a big bubble explosion.”
His technical complaint is just as sharp. Most frontier systems are built on large language models, which are good at text, coding and structured reasoning, but not necessarily at understanding the real world.
LeCun has been pushing what he calls world models. These systems are meant to grasp how environments function, including cause and effect, physical interactions and context.
“I personally don’t think we’re going to have generalised reliable agentic systems until they’re based on world models,” he said.
That puts him at odds with the current herd, where OpenAI and Anthropic are trying to build more capable agents on top of large language model foundations.
LeCun is not dismissing those systems outright. He is questioning whether they can scale economically when users do not want to pay what they cost to run.
AMI Labs is betting on the alternative route. The company raised about $1.03 billion earlier this year at a reported $3.5 billion pre-money valuation to build world model-based systems.
