London robotics startup Machenta raises $40 million for a model that teaches robots a task from a single demonstration

London-based Machenta announced a $40 million seed round on October 27, co-led by Creandum and Union Square Ventures. The company declined to disclose its valuation. According to Tech.eu, Machenta says it is the largest seed round raised by a UK robotics company to date. It was founded late last year by Edward Johns, a professor of robot learning at Imperial College London, and his Ph.D. student Vitalis Vosylius. The startup has 11 employees.

Machenta is building a foundation model meant to let a robot perform a new task after watching it once, without further training. Johns told Business Insider the model is trained in computer-generated environments without a “single pixel of real-world data.” The company says it has shown this with packaging, assembly and tidying tasks, and that one model works across different robots, including humanoids. Most rivals instead gather hundreds or thousands of real-world demonstrations and train a model for each task.

The company remains focused on research, with no near-term plans to commercialize. Its model can learn some tasks immediately, but precise tasks and complex movements remain hard. The funding will go toward expanding the team, more computing power and more simulated data. Rivals include Physical Intelligence, valued at roughly $11 billion, and FieldAI. Skild AI, valued at more than $14 billion, unveiled a model in August that it said could perform unseen tasks after a single video demonstration.