March 10, 2026 France/European Economic News - From Language to 'World Models': French Startup AMI Raises 890 Million Euros to Accelerate Next-Generation AI Development

Every day, I would like to pick out the most noteworthy piece of economic news from France.
If you are short on time, please read the first three lines of the summary.
If you like it, I would be happy if you could 'like' the post.


From Language to 'World Models': French Startup AMI Raises 890 Million Euros to Accelerate Next-Generation AI Development

Summary

Yann LeCun's new company, AMI, has launched with a valuation of 3 billion euros. To overcome the reasoning capabilities lacking in existing AI, it will focus on developing 'world models' using JEPA and other technologies. With 890 million euros in funding, it aims for general-purpose industrial applications.

Meta's former head of AI research, Yann LeCun, is the chairman of the new company, AMI (Advanced Machine Intelligence), which has now launched. Although the company is a French startup based in Paris, it has raised 890 million euros from investors in Europe, the US, and Asia, reaching a corporate valuation of 3 billion euros.

The Challenge of 'World Models' Beyond Conventional AI
AMI's core goal is to break through the limitations of language-centric AI models currently led by companies like OpenAI and Google. LeCun points out that existing AI lacks even the reasoning capabilities of a 'six-year-old' or a 'cat.' In response, AMI is focusing on developing a new form of AI called 'world models' (world models), which learn physical and conceptual abstract representations of the real world.

Technical Strategy and Future Outlook
The company's technical features and plans are as follows.

• JEPA Architecture: Based on this structure, which LeCun proposed at Meta, the company will conduct learning using video data over the next approximately 12 months.

• Industrial Application: Aiming to revolutionize current robotics, which only function in specific environments, and create a general-purpose model capable of adapting to unknown situations.

• Support for Open Source: The company plans to release much of its research and code, arguing that preventing AI monopolies is key to risk management.

Currently, AMI has about 10 employees, but plans to expand to a scale of 30 to 50 people within six months. The company aims to build the 'most general-purpose world model' that will bring true value to the industry over the next few years.