PARIS – A new artificial intelligence startup founded by former Meta chief AI scientist Yann LeCun has rapidly achieved “unicorn” status, securing over $1 billion in funding just months after its inception. Advanced Machine Intelligence (AMI), headquartered in Paris, is poised to challenge the dominance of large language models (LLMs) with its focus on “world models,” a more comprehensive approach to AI that seeks to understand and interact with the physical world.
The company announced Tuesday it raised $1.03 billion (approximately €890 million) at a $3.5 billion pre-money valuation, a figure significantly higher than the €500 million it was reportedly seeking in December 2025. This substantial investment underscores growing confidence in the potential of world models as the next frontier in artificial intelligence development. The funding round positions AMI among a select group of AI startups attracting significant capital, including SpAItial, which raised $13 million, and World Labs, led by Fei-Fei Li, which secured $1 billion last month, according to TechCrunch.
Beyond Language: The Promise of World Models
AMI’s core mission is to develop AI that learns from reality, rather than solely relying on vast datasets of text and code. This approach, known as world modeling, aims to create AI systems capable of reasoning about the physical world, making predictions, and interacting with their environment in a more nuanced and reliable way. This contrasts with the current generation of LLMs, such as OpenAI’s ChatGPT and Anthropic’s Claude, which excel at generating human-like text but can struggle with factual accuracy and real-world understanding – a phenomenon known as “hallucination.”
Yann LeCun, a Turing Award winner for his pioneering work in deep learning, co-founded AMI with six other former Meta researchers: Alexandre LeBrun, Laurent Solly, Pascale Fung, Saining Xie, and Michael Rabbat. LeCun’s research into Joint Embedding Predictive Architecture (JEPA), proposed in 2022, forms the foundation of AMI’s technological approach. JEPA focuses on learning representations of the world that allow AI systems to predict future states based on current observations. This predictive capability is crucial for building AI that can operate safely and effectively in complex environments.
Healthcare as a First Frontier
AMI is initially targeting the healthcare sector as a key application area for its world model technology. The company has partnered with Nabla, a digital health startup, to explore how world models can improve diagnostic accuracy and patient care. Alexandre LeBrun, CEO of AMI and also chairman of Nabla, explained that the limitations of LLMs – particularly the risk of hallucinations – are unacceptable in healthcare settings where incorrect information could have life-threatening consequences, as reported by Reuters.
“As CEO of Nabla, I reached the same conclusion as Yann on the limitations of large language models,” LeBrun stated. “Hallucinations could have life-threatening repercussions.” World models, by grounding AI in a more accurate understanding of reality, offer a potential solution to this critical challenge.
A Long-Term Vision
While the $1.03 billion funding round is substantial, AMI’s leaders acknowledge that bringing world models to commercial viability will be a long-term undertaking. “AMI Labs is a very ambitious project, because it starts with fundamental research,” LeBrun told TechCrunch. “It’s not your typical applied AI startup that can release a product in three months, have revenue in six months, and produce $10 million in [annual recurring revenue] in 12 months.” He anticipates it will take years for the technology to mature and find widespread application.
This longer development timeline differentiates AMI from many other AI startups focused on rapidly deploying LLM-based products. However, LeBrun believes that the potential rewards of building truly intelligent systems that understand the world are worth the investment. He even playfully predicted that “world models” will become the next buzzword in the AI industry, with many companies attempting to capitalize on the trend.
The Rise of Alternative AI Approaches
AMI’s success reflects a growing interest in alternative approaches to AI that move beyond the limitations of LLMs. While LLMs have demonstrated impressive capabilities in natural language processing, they often lack common sense reasoning and struggle with tasks that require understanding of the physical world. World models represent an attempt to address these shortcomings by building AI systems that can learn and reason in a more holistic way.
The substantial funding raised by AMI, World Labs, and SpAItial signals a broader shift in the AI landscape. Investors are increasingly recognizing the potential of these alternative approaches and are willing to provide the capital needed to support long-term research and development. According to The New York Times, Advanced Machine Intelligence Labs was founded by Yann LeCun and other ex-Meta researchers and has raised more than $1 billion from investors.
Key Takeaways
- Significant Funding: AMI secured $1.03 billion, valuing the company at $3.5 billion, demonstrating strong investor confidence in world model technology.
- Focus on Reality: Unlike LLMs, AMI aims to build AI that understands and interacts with the physical world, reducing the risk of inaccuracies and hallucinations.
- Healthcare Application: The company is initially targeting the healthcare sector, where reliable AI is crucial for accurate diagnosis and patient care.
- Long-Term Investment: AMI acknowledges that developing world models will require years of research and development, differentiating it from faster-moving LLM startups.
The development of world models is still in its early stages, but AMI’s substantial funding and the expertise of its founding team position it as a leading player in this emerging field. The company’s success could pave the way for a new generation of AI systems that are more robust, reliable, and capable of solving complex real-world problems.
AMI plans to continue its research and development efforts in Paris, building on the foundation laid by Yann LeCun’s pioneering work in deep learning. The company will be closely watched by the AI community as it seeks to translate its ambitious vision into tangible results. The next major milestone for AMI is expected to be the demonstration of a functional prototype of its world model technology, anticipated in late 2027, according to company statements.
What are your thoughts on the future of world models? Share your comments below and join the conversation.