Yann LeCun, the French AI pioneer often called one of the “godfathers” of deep learning, has turned a new page in his life. After twelve years at Meta, where he led research and represented FAIR, he is leaving to create his own company. And not just any company: Advanced Machine Intelligence Labs, or AMI Labs, which will officially launch in January 2026.
AMI Labs: A Head-Turning Fundraise
He doesn’t do things by halves. According to circulating information (confirmed by several sources like the Financial Times), Yann LeCun is in discussions to raise approximately 500 million euros. This would value his startup at 3 billion euros from the outset, even before it releases a product.
In the 2025 AI world, where valuations for ideas on paper are exploding, this is huge, but not entirely crazy when you see what others like Ilya Sutskever have raised recently.
Also read on this topic: Meta: Yann LeCun Leaves the Company Due to an Overly “Business-Oriented” Vision of AI?
A Team Strengthened by Key Recruitment
To lead the daily operations, Yann LeCun convinced Alexandre Lebrun, the CEO and founder of Nabla (the French startup excelling in AI for healthcare), to join. Lebrun will become CEO of AMI Labs, while retaining a role at Nabla as chairman and chief AI scientist.
The two companies have even announced a strategic research partnership, with Nabla gaining priority access to AMI’s technologies. Smart move, as it allows testing ideas in the real world, especially in healthcare where reliability is crucial.
Why He Says No to Large Generative Models
What makes AMI Labs truly interesting is Yann LeCun’s vision. He has never been a fan of the mega LLMs that everyone loves to scale at OpenAI, Google, or even Meta now. For him, these generative models are limited: they churn out text or images, but don’t truly understand the physical world.
He’s betting everything on “world models,” systems that simulate reality, learn from videos and spatial data, predict the consequences of actions, and ultimately reason and plan like a human (or almost).
One can imagine robots that truly anticipate what will happen, or AI for autonomous driving that grasps physics without hallucinating.
That’s the ambition: smarter, more autonomous AI, applicable everywhere, from robotics to transport.
Yann LeCun himself said it: the Valley is obsessed with generatives, so we need to work elsewhere, like in Paris. AMI Labs will be global, with a strong base in Europe. Meta, for its part, will remain a technical partner (without investing money), to gain access to advancements.
In short, with his golden resume (Turing Award included) and this approach that defies the mainstream, Yann LeCun could once again shake up AI. 2026 promises to be exciting, and we can’t wait to see if these world models will truly change the game!
Source: Financial Times
