Robotics startup Machenta raises $40 million to train robots from single demonstrations in simulation

Imperial College spinout takes on better-funded rivals with a model trained without real-world data

By LineZotpaper AI staff

Published27 October 2026

Read Time2 min

Illustration: Robotics startup Machenta raises $40 million to train robots from single demonstrations in simulation

London-based robotics startup Machenta has raised $40 million in seed funding to build an AI model that can teach robots new physical tasks after a single demonstration, using a model trained entirely in simulation. The round was co-led by venture capital firms Creandum and Union Square Ventures.

Machenta, founded by researchers from Imperial College London, is taking a contrarian approach to rivals that are amassing real-world data for general-purpose robot models. Edward Johns, the startup's CEO and cofounder and an Imperial professor, said the company trains its model in computer-generated environments that mimic how objects move and interact, without using a "single pixel of real-world data." After watching a task performed once, such as tidying a table or putting objects into a box, the model can direct a robot to perform it without further training, Johns said. In a video shared with Business Insider, three robots pack and close a box on an assembly line, a task Johns said they learned this way.

The startup has 11 employees and remains focused on research, with no near-term plans to commercialize its technology. Its model can learn some tasks immediately, but more precise tasks and complex movements remain challenging.

Machenta faces rivals with multibillion-dollar valuations in one of robotics' most competitive fields. California-based competitors include Physical Intelligence, valued at roughly $11 billion, and FieldAI, which is raising funding at a $10 billion valuation. Another rival, Skild AI, valued at over $14 billion, is already selling technology with similar capabilities. In August, Skild unveiled a model that it said could perform previously unseen tasks lasting up to 10 minutes after a single video demonstration, and last month said it had reached $100 million in annual recurring revenue from more than 60 customers.

Johns began investigating whether robots could learn as quickly as humans at his Imperial College lab six years ago. In early experiments, researchers physically guided a robot through a task, and the robot then spent about 30 minutes collecting additional data before performing it alone. Johns and former PhD student Vitalis Vosylius, now cofounder, spent years working to remove that step, founding Machenta late last year. The company plans to use the funding to expand its team and train its model with more computing power and simulated data, hoping to teach robots more precise and complex tasks.

Machenta declined to disclose its valuation.

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

Analysis

Why This Matters

Machenta's simulation-only approach challenges the dominant strategy of collecting large amounts of real-world data, potentially lowering a key barrier to general-purpose robots.
The seed round is an early bet on a research-stage team competing against rivals with multibillion-dollar valuations.
If the approach works for precise and complex tasks, it could change how robot models are trained and scaled.

Background

Robotics companies have increasingly bought and collected data to teach robots to interact with the real world, but questions remain about how much training data will be needed to make general-purpose robots a reality. Machenta is instead trying to design algorithms that generate synthetic data useful for teaching physical tasks, a harder challenge in Johns' view. The company emerged from years of research at Imperial College London, where early experiments required a robot to collect about 30 minutes of additional data after being physically guided before it could perform a task.

Key Perspectives

Machenta CEO Edward Johns: says collecting more real-world data is not the answer, and that training entirely in simulation lets robots learn tasks after a single demonstration. Skild AI: says it has a model that can perform previously unseen tasks lasting up to 10 minutes after a single video demonstration, and that it has reached $100 million in annual recurring revenue from more than 60 customers.

What to Watch

Whether Machenta can overcome the challenge of precise and complex movements, which it says remain difficult for its model.
Whether the company shifts from research toward commercialization after expanding its team and training capacity.
How Skild AI's claims of commercial traction affect the competitive picture as Machenta scales.

Sources

An 11-person startup raised $40 million to challenge robotics' data-hungry giants · Business Insider

Checked

12 claims in this story checked against 1 source: 10 supported.

Show each claim

SupportedMachenta raised $40 million to build a model that teaches robots physical tasks after one demonstration, using simulation.All Content from Business Insider: “Machenta, founded by researchers from Imperial College London, has raised a $40 million seed round to build an AI model that can teach robots new physical tasks after seeing them performed just once”
Not in the sourcesMachenta is an Imperial College spinout.All Content from Business Insider: “Machenta, founded by researchers from Imperial College London”
Not in the sourcesMachenta is taking on better-funded rivals.All Content from Business Insider: “A London-based robotics startup is betting it can teach robots new tricks without collecting the mountains of real-world data its much bigger rivals are chasing.”
SupportedCreandum and Union Square Ventures co-led the funding round.All Content from Business Insider: “Venture capital firms Creandum and Union Square Ventures co-led the round.”
SupportedJohns said the model is trained without a “single pixel of real-world data.”All Content from Business Insider: “Edward Johns, the startup's CEO and cofounder, and an Imperial professor, said the company trains its model in computer-generated environments that mimic how objects move and interact, without using a "single pixel of real-world data."”
SupportedAfter seeing a task once, the model can direct a robot to do it without further training.All Content from Business Insider: “Once trained, the model can direct a robot to perform a new task, such as tidying a table or putting objects into a box, after watching it done just once, without further training, Johns said.”
SupportedJohns said three robots in a Business Insider video learned to pack and close a box on an assembly line this way.All Content from Business Insider: “In a video shared with Business Insider, three robots pack and close a box on an assembly line, a task Johns said they learned this way.”
SupportedMachenta has 11 employees.All Content from Business Insider: “The startup has 11 employees and remains focused on research, with no near-term plans to commercialize its technology.”
SupportedMachenta has no near-term plans to commercialize its technology.All Content from Business Insider: “The startup has 11 employees and remains focused on research, with no near-term plans to commercialize its technology.”
SupportedMore precise tasks and complex movements remain challenging for Machenta’s model.All Content from Business Insider: “Its model can learn some tasks immediately, but more precise tasks and complex movements remain challenging.”
SupportedPhysical Intelligence is valued at roughly $11 billion.All Content from Business Insider: “Its California-based competitors include Physical Intelligence, valued at roughly $11 billion”
SupportedFieldAI is raising funding at a $10 billion valuation.All Content from Business Insider: “FieldAI, which is raising funding at a $10 billion valuation”