LeCun's AMI Raises $1.03 Billion to Challenge LLM Orthodoxy with 'World Models'

Advanced Machine Intelligence, the startup founded by former Meta chief AI scientist Yann LeCun, has raised $1.03 billion at a $3.5 billion pre-money valuation to commercialize artificial intelligence systems built around reasoning, planning, and what the company calls "world models."

The financing round, co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, positions AMI as a direct test of LeCun's thesis that current large language models fall fundamentally short of human-level reasoning and autonomy. Rather than predicting the next word in a sequence, AMI's approach aims to build systems that learn abstract representations of reality, simulate environments, anticipate consequences, and plan sequential actions based on cause and effect.

The distinction matters most in domains requiring deterministic reasoning under real-world constraints. While LLMs excel at clinical documentation and knowledge retrieval, they remain probabilistic text generators that struggle with continuous multimodal data streams—vital signs, imaging, sensor feeds—and long-term planning in complex operational environments. Healthcare applications have emerged as a proving ground for the limitations: LLMs cannot reliably model the cascading effects of prescribing a specific medication to a patient with particular comorbidities, nor can they manage trade-offs between individual patient optimization and system-wide resource constraints.

(LeCun joined Meta in 2013 to found Facebook AI Research, later known as FAIR, and became one of the company's most prominent AI leaders before departing at the end of 2025. Meta has since reorganized its AI efforts under Meta Superintelligence Labs, led by former Scale AI CEO Alexandr Wang.)

The fundraise arrives as evidence mounts that LLM adoption, while rapid, remains far from theoretical capability. Research tracking actual workplace usage of Anthropic's Claude found that the system covers just 33 percent of all tasks in computer and mathematics categories that it could theoretically handle. In healthcare specifically, LLM-related studies surged from 757 in 2023 to 2,787 in 2024, yet researchers increasingly advocate hybrid strategies: employing LLMs for generation and interaction while relying on traditional methods for structured information extraction and standardized concept representation.

The competitive landscape has consolidated around a handful of consumer-facing platforms, with OpenAI's ChatGPT maintaining dominance and Google's Gemini leveraging distribution advantages through its installed base. AMI's bet is that the next phase of AI value creation will come not from better text prediction, but from systems capable of operating autonomously in physical and digital environments where actions have irreversible consequences and resources are finite.