London robotics startup Machenta has raised $40 million in seed funding to build an AI model that teaches robots a new task from a single demonstration. Creandum and Union Square Ventures co-led the round, announced on October 27, 2026.
Machenta spun out of Imperial College London. The university called it the largest seed round ever raised by a UK robotics startup.
The company trains its model entirely in simulation, without real-world data. Machenta did not disclose a valuation.
| Detail | Machenta |
|---|---|
| Headquarters | London, United Kingdom |
| Founded | 2025 |
| Founders | Edward Johns (CEO), Vitalis Vosylius |
| Round | Seed |
| Amount | $40 million (about £30 million) |
| Valuation | Not disclosed |
| Co-lead investors | Creandum, Union Square Ventures |
| Team size | 11 people |
| Origin | Spinout from Imperial College London |
| Announced | October 27, 2026 |
Machenta seed funding round and investors
Swedish venture firm Creandum and New York-based Union Square Ventures co-led the $40 million round. No other investors were named in the Imperial College London announcement.
Kathi Wilhelm, a partner at Creandum, said the team had worked on the problem for years before forming a company. She pointed to its track record of research results as the reason for the bet.
The headline figure differs from the company’s latest UK filing. A Companies House return dated September 30 records seed preferred shares worth about £22.7 million, Konsulteer reported. The reason for the gap has not been confirmed.
How Machenta’s robot model learns from one demonstration
Most AI-powered robots learn a task after people demonstrate it many times. That training can take days, and adding a new skill usually means developers must update the model.
Machenta’s model takes a different approach. A user shows the robot a task once, and the robot can repeat it with no further training.
The model relies on in-context learning, a technique Johns and Vosylius developed for robots. It is pretrained only on simulated data, and Johns told Business Insider it uses no real-world images at all.
Machenta says one model has learned packing, assembly and tidying tasks across several robot types, including humanoids. In a video shared with Business Insider, three robots packed and closed a box on an assembly line.
Training on simulated data also avoids the privacy limits of collecting large amounts of human data, Imperial said.
Who founded Machenta
Edward Johns is Machenta’s co-founder and CEO. He is Professor of Robot Learning in Imperial’s Department of Computing and directs the university’s Robot Learning Lab.
Johns began studying one-shot robot learning in 2020. In March 2026, his lab taught a robot 1,000 tasks in 24 hours, some from a single human demonstration, Forbes reported.
Co-founder Vitalis Vosylius is Johns’ former PhD student. The pair founded Machenta late last year with colleagues from the lab.
“Intelligence is not so much about what you can already do,” Johns said in the announcement. He argued the real measure is how fast a robot learns something new.
How Machenta’s $40 million seed compares
Machenta’s round is about 13 times the median AI seed round of $3.1 million between October 2025 and September 2026, according to FundingWatcher data.
It is also larger than recent early rounds for other robotics startups. Zurich-based Enact Intelligence raised a $30 million pre-seed on October 22, and simulation startup Antioch raised $32 million in September.
For other European AI companies that raised large rounds this year, see our list of Europe’s top-funded AI startups.
What Machenta will do with the funding
Machenta will use the money to scale its model and grow its team at a new office and lab in King’s Cross, London.
Much of the spending will go to computing power and larger simulation training runs, Tech Funding News reported. The company has not said how many people it plans to hire.
Machenta has no near-term plans to sell a product, according to Business Insider. The team will focus on proving the approach works on harder tasks and over longer periods.
Simulation-trained robot models draw investor money
Many companies building general-purpose robot models collect large volumes of real-world data, often from human demonstrations. Machenta is betting that simulation can replace most of that work.
The bet is still unproven. Moving skills learned in simulation onto physical machines remains a hard technical problem, and Machenta has shown its results only in demonstrations so far.
Investors are backing both approaches. Seed-stage firms are writing larger checks for robotics teams with deep research records, and our list of the most active seed investors in AI tracks who leads those rounds.
FAQ
How much did Machenta raise?
Machenta raised $40 million, about £30 million, in a seed round announced on October 27, 2026. Imperial College London called it the largest seed round ever raised by a UK robotics startup.
Who invested in Machenta?
Creandum and Union Square Ventures co-led Machenta’s $40 million seed round. The company did not name other investors or disclose a valuation.
What does Machenta build?
Machenta builds a robot foundation model that learns a new task from one demonstration without further training. The model is pretrained entirely on simulated data.
Who founded Machenta?
Imperial College London professor Edward Johns and his former PhD student Vitalis Vosylius founded Machenta in 2025. Johns is the CEO and directs Imperial’s Robot Learning Lab.
Is Machenta selling its robot model yet?
No. Machenta has no near-term plans to commercialize, Business Insider reported. The 11-person company is focused on research and scaling its simulation-trained model.
