Machenta raises $40 million to teach robots a task from one demo

Machenta, a London robotics startup founded by Imperial College London researchers, raised a $40 million seed round co-led by Creandum and Union Square Ventures, with its valuation not disclosed. The company is building an AI model that teaches robots to perform new physical tasks after seeing them demonstrated just once, without further training.

CEO and co-founder Edward Johns, an Imperial professor, told Business Insider the model is trained entirely in computer-generated simulated environments that mimic how objects move and interact, using no real-world data. Once trained, the model can direct a robot to tidy a table or pack objects into a box after watching the task performed a single time. In a video shared with Business Insider, three robots pack and close a box on an assembly line. The 11-person company is research-focused with no near-term plans to commercialize, and says precise tasks and complex movements remain challenging.

Johns began researching fast robot learning at his Imperial lab six years ago, when early robots still needed about 30 minutes of extra data collection after a guided demonstration. He and former PhD student Vitalis Vosylius, now Machenta's CTO, worked to remove that step and founded the company in late 2025. The funding will expand the team and fund more compute and simulated data for harder, more precise tasks. Its rivals include Physical Intelligence, valued at about $11 billion, and Skild AI, valued at over $14 billion.

Why it matters

Learning from a single demonstration in pure simulation, if it generalizes, could sharply cut the cost of teaching robots new tasks.

Source: Business Insider