Machenta raises $40m seed to build a robot model that learns a task from one demonstration
Imperial College London spinout Machenta closed a $40m seed led by Creandum and Union Square Ventures, pre-trained only on simulated data.
Machenta, a London robotics company spun out of Imperial College London, announced a $40 million seed round on 27 October 2026. Creandum and Union Square Ventures led it. Tech.eu and Imperial both describe the round as the largest seed raise by a UK robotics startup to date.
The company is building a robot foundation model trained once and then pointed at many tasks. The distinctive choice is the training data: the model is pre-trained entirely on simulated data, with no real world or human demonstration footage, which the company says sidesteps the privacy and data access problems that slow its rivals. Machenta says a robot can carry out a task immediately after seeing it demonstrated once, and that one or two real demonstrations are enough per new task. The team has shown a single model handling everyday work such as packing, assembly and tidying across several different robot bodies, humanoids among them. Precise tasks and complex motions remain hard.
The founders are Professor Edward Johns, co-founder and chief executive and Professor of Robot Learning in Imperial's Department of Computing, and Vitalis Vosylius, a doctoral student at Imperial. Imperial says further colleagues from its robot learning laboratory also co-founded the company. The money goes into scaling the model, buying more compute, generating more simulated data and filling a new office and laboratory at King's Cross in London. Notably, Machenta still describes itself as a research company with no near term plan to commercialise.
The round lands in a well funded corner of deep tech. Tech Funding News, citing Crunchbase, counts $47.4 billion of global physical AI venture funding across 521 deals in the first half of 2026. "Intelligence is not so much about what you can already do, it's about how quickly you can learn to do something new," said Johns. Kathi Wilhelm, a partner at Creandum, said robots become far more useful once they can pick up a new task and start work straight away, and added that very few teams in robotics can point to a comparable body of work. Rebecca Kaden, managing partner at Union Square Ventures, framed the need as general purpose AI that adapts as quickly as a person does. Professor Mary Ryan, Imperial's Vice-Provost for Research and Enterprise, said the company's early success signals an opportunity for the UK to lead in AI enabled robotics.
Why it matters for AgTech founders: agricultural robotics has been held back less by hardware than by the cost of teaching a machine each new job. Every crop, variety, trellis and field layout is effectively a new task, and the window to collect real examples opens once a year, for a few weeks, in weather nobody controls. That is the constraint this round is funding a way around. Three things follow. First, if simulation-first pre-training holds up outside the lab, the moat in ag robotics stops being the proprietary field dataset that most pitch decks still lead with, and starts being the simulator and the task library. Second, check honestly whether your roadmap assumes thousands of real demonstrations per crop, because that assumption is now a liability rather than a defence. Third, note where the money went. A pre-commercial university spinout raised $40m for the model layer, while ag-specific application layers are raising single digit millions, which argues for being the fastest adopter of someone else's foundation model rather than training your own. Ask which parts of your stack you would rather rent in two years than own.
Machenta raises $40m seed to build a robot model that learns a task from one demonstration
Creandum and Union Square Ventures led the round for the Imperial College London spinout whose model is pre-trained entirely in simulation, with no real world or human demonstration data.
Machenta, a London robotics company spun out of Imperial College London, announced a $40 million seed round on 27 October 2026. Creandum and Union Square Ventures led it. Tech.eu and Imperial both describe the round as the largest seed raise by a UK robotics startup to date.
The company is building a robot foundation model trained once and then pointed at many tasks. The distinctive choice is the training data: the model is pre-trained entirely on simulated data, with no real world or human demonstration footage, which the company says sidesteps the privacy and data access problems that slow its rivals. Machenta says a robot can carry out a task immediately after seeing it demonstrated once, and that one or two real demonstrations are enough per new task. The team has shown a single model handling everyday work such as packing, assembly and tidying across several different robot bodies, humanoids among them. Precise tasks and complex motions remain hard.
The founders are Professor Edward Johns, co-founder and chief executive and Professor of Robot Learning in Imperial's Department of Computing, and Vitalis Vosylius, a doctoral student at Imperial. Imperial says further colleagues from its robot learning laboratory also co-founded the company. The money goes into scaling the model, buying more compute, generating more simulated data and filling a new office and laboratory at King's Cross in London. Notably, Machenta still describes itself as a research company with no near term plan to commercialise.
The round lands in a well funded corner of deep tech. Tech Funding News, citing Crunchbase, counts $47.4 billion of global physical AI venture funding across 521 deals in the first half of 2026. "Intelligence is not so much about what you can already do, it's about how quickly you can learn to do something new," said Johns. Kathi Wilhelm, a partner at Creandum, said robots become far more useful once they can pick up a new task and start work straight away, and added that very few teams in robotics can point to a comparable body of work. Rebecca Kaden, managing partner at Union Square Ventures, framed the need as general purpose AI that adapts as quickly as a person does. Professor Mary Ryan, Imperial's Vice-Provost for Research and Enterprise, said the company's early success signals an opportunity for the UK to lead in AI enabled robotics.
Why it matters for AgTech founders: agricultural robotics has been held back less by hardware than by the cost of teaching a machine each new job. Every crop, variety, trellis and field layout is effectively a new task, and the window to collect real examples opens once a year, for a few weeks, in weather nobody controls. That is the constraint this round is funding a way around. Three things follow. First, if simulation-first pre-training holds up outside the lab, the moat in ag robotics stops being the proprietary field dataset that most pitch decks still lead with, and starts being the simulator and the task library. Second, check honestly whether your roadmap assumes thousands of real demonstrations per crop, because that assumption is now a liability rather than a defence. Third, note where the money went. A pre-commercial university spinout raised $40m for the model layer, while ag-specific application layers are raising single digit millions, which argues for being the fastest adopter of someone else's foundation model rather than training your own. Ask which parts of your stack you would rather rent in two years than own.