Yann LeCun says Meta is burning cash on dead end AI models
Yann LeCun breaks his silence on his exit from Meta, calling the company's $15B bet on Alexandr Wang and LLMs a "dead end" in a bombshell FT interview.
Mark Zuckerberg and Yann LeCun / Image Credits: Yasiru S / Pressvia
Meta just lit $15 billion on fire, and the man who built their AI division says the flames are only getting higher. When Yann LeCun, the Turing Award winner and former Chief AI Scientist at Meta, walked out of Menlo Park in November, the company framed it as a standard retirement, but a scorching new interview reveals a scientific mutiny that could derail Mark Zuckerberg’s entire artificial intelligence roadmap.
The $15 Billion Betrayal
In his exclusive sit-down with The Financial Times, LeCun dismantled the industry’s obsession with Large Language Models (LLMs), effectively calling Meta’s current strategy a scientific fraud. The breaking point wasn't just philosophical; it was financial. LeCun revealed that Zuckerberg’s decision to invest $14.9 billion in Scale AI and hire its 28-year-old CEO, Alexandr Wang, to lead the new Superintelligence Labs was the final insult.
"You don't tell a researcher like me what to do," LeCun told the FT, referencing his forced reporting line to Wang, whom he described as "inexperienced" in research protocols Futurism. While Zuckerberg pivoted the company to chase OpenAI, LeCun remained steadfast that LLMs are a "dead end" for achieving true superintelligence because they lack a fundamental understanding of the physical world Times of India.
Llama 4 Was The Warning Shot
The interview confirms what Wall Street had long suspected: the Llama 4 model, released in April 2025, was a flop. LeCun admitted that the benchmark results for the model were "fudged" to save face after the model failed to show significant reasoning improvements over its predecessors Implicator.ai.
This failure drives the wedge deeper. While Meta prepares to spend upwards of $40 billion in capital expenditures in 2026 to train Llama 5, LeCun argues that simply adding more compute to an LLM is like building a better ladder to reach the moon, it will never work. He has since launched his own startup, Advanced Machine Intelligence Labs, which is already targeting a $3 billion valuation to build "world models" that learn from video and physical interaction rather than text WebProNews.
The "LeCun Discount"
LeCun’s exit is a massive blow to Meta's credibility. For a decade, his presence reassured investors that Meta was a serious scientific player, not just an ad company. With him gone, and vocally betting against the company's core product—analysts are scrambling to re-evaluate the stock's long-term potential. If the "godfather of AI" says the underlying tech is broken, Zuckerberg's trillion-dollar pivot to the metaverse and AI might be built on sinking sand.
