THE ESSENTIALS IN 30 SECONDS

  • Machenta, spun out of Imperial College London, raised $40 million in seed funding led by Creandum and Union Square Ventures.
  • The company is building a robot foundation model pre-trained entirely on simulated data, aimed at learning new tasks from a single demonstration.
  • The money will fund scaling the model and a larger team at a new office and lab in King's Cross, London.

Machenta, a physical AI company spun out of Imperial College London, has raised $40 million in seed funding. The round, announced on October 27, 2026, was led by Creandum and Union Square Ventures. Imperial describes it as the largest seed round ever raised by a UK robotics startup; no other investors or valuation were disclosed.

What Machenta is building

The startup is developing a robot foundation model intended to let a robot pick up a new task after being shown it once. According to the university, today's robots often need many demonstrations collected over several days before they can reliably perform a new job. Machenta's approach is to pre-train its model entirely on simulated data, rather than on large collections of real-world or human demonstration recordings.

The company says a single model has already learned varied tasks such as packing and tidying, and that it can run on different kinds of robots, including humanoids. Imperial also suggests that relying on simulation could reduce the need for physical data collection and sidestep some of the practical and privacy issues that come with recording people. These capabilities are described by the company and the university; no independent evaluation has been published alongside the announcement.

Who is behind it

Machenta is led by co-founder and CEO Edward Johns, Professor of Robot Learning in Imperial's Department of Computing. The research started about six years ago in his Robot Learning Lab, where he worked on it with PhD student Vitalis Vosylius and other lab members. Johns frames the company's goal around speed of learning, arguing that intelligence is about how quickly a system can learn something new.

What the money is for

The funding will go toward scaling the model across more robots and tasks and toward expanding the team, which will be based at a new office and laboratory in King's Cross, London. Neither source gives hiring targets or a timeline for commercial products.

Machenta's bet is that simulation combined with one-shot learning can cut the amount of real-world data a robot needs before it becomes useful. That claim will need to be tested on customer sites, outside the controlled conditions of a research lab.