AMI LABS a 3.5 billion dollars valuation for an EU world model startup

Large language models were built to predict the next word, not the next action within a physical environment. Because they are trained on text, these models often struggle with physical intuition, misinterpret sensor data, and lack the temporal planning required by industrial or medical systems. This gap creates a fundamental ceiling on AI reliability wherever digital logic meets the real world.

AMI Labs, headquartered in Paris, was founded to overcome that ceiling. Executive Chair Yann LeCun (Turing Award, former Meta Chief AI Scientist), CEO Alexandre LeBrun (former Nabla), and COO Laurent Solly (former VP Meta Europe) are betting onWorld Models. By training on video and spatial sensor data rather than text, they believe these models will outperform LLMs in any environment where physical consequences matter.

The architecture generates action-conditioned predictions within representation space, enabling persistent memory, environmental reasoning, and reliable planning. While no formal benchmarks have been disclosed yet, due to complexity and novelty of the approach, the target sectors are broad: robotics, industrial automation, wearables, healthcare. A direct implementation is already visible with Nabla, the company´s first partner, applying the models in clinical workflows.

AMI's $1.03B seed round at a $3.5B valuation, puts AMI Labs at the top of Europe's AI founding rounds, and is backed by Bezos Expeditions, NVIDIA, Temasek, and Toyota Ventures. The bet is structural: the next wave of AI value lives not in language, but in the physical world.

Sources: Trending Topics / AMI Labs