Machenta Raises $40M Funding to Teach Robots New Tasks From a Single Demonstration

Machenta, the London-based physical AI spinout from Imperial College London, has emerged from stealth with $40 million in seed funding—the largest seed round ever raised by a UK robotics startup. The round was co-led by Creandum and Union Square Ventures, with participation from IQ Capital and Hoxton Ventures .

The funding comes as Machenta unveils a robot foundation model that enables robots to learn new physical tasks from a single demonstration, with no further training or fine-tuning required. The model is pre-trained entirely on simulated data—without using “a single pixel of real-world data”—a contrarian bet against competitors spending hundreds of millions collecting human demonstration data .

Machenta Funding Overview By CREDX Media

Startup Name	Machenta
Founders	Professor Edward Johns (CEO), Vitalis Vosylius (CTO)
Founded	2025
HQ	London, UK (King’s Cross)
Round	Seed Funding
Amount Raised	$40M
Valuation	Not disclosed
Category	Physical AI, Robotics Foundation Models, Simulation
Lead Investors	Creandum, Union Square Ventures
Key Investors	IQ Capital, Hoxton Ventures
Total Funding Raised	$40M
Employees	~11
Key Technology	Instant Robot Learning, Vision-Demo-Action (VDA) model
Commercial Product	None yet (research focus)

💡 Use of Funds

• Team Expansion: Growing the research team at Machenta’s new office and lab in King’s Cross, London .
• Compute & Data: Scaling the computing power and simulated data used to train the model for more precise and complex tasks .
• Model Development: Extending the range and complexity of tasks that can be taught through single demonstrations .

Startup Overview: Machenta

What Machenta Is Building in Simple Words

Machenta has built a robot foundation model that learns how to learn. Instead of training a robot on thousands of demonstrations of a specific task, Machenta’s model is pre-trained in simulation to develop a general capability: the ability to infer a task from a single example and execute it immediately.

The core technical approach is called Vision-Demo-Action (VDA). The model receives a demonstration of the desired behaviour—which can be a human hand, a VR controller, or another robot—alongside camera observations, and generates robot actions in a closed loop . Once trained, the model can direct a robot to perform a new task after watching it done just once, without any task-specific fine-tuning or weight updates.

• Single Demonstration Learning: Robots learn tasks like packing, tidying, and assembly after one demonstration .
• Trained Entirely in Simulation: No real-world data is used for pre-training, sidestepping the cost and privacy constraints of collecting human demonstrations .
• Works Across Robots: The same model has been demonstrated on six different robot embodiments with different cameras and hardware, including humanoid hands .
• Physical Vibe Coding: CEO Edward Johns likens the approach to “physical vibe coding”—showing a robot what to do and having it execute immediately, without engineering expertise .

Why Machenta Matters

1. The Data Bottleneck Is Real—and Machenta Is Betting Against It The best-funded robotics labs are spending hundreds of millions collecting real-world demonstration data. Machenta’s thesis is that this approach is a chicken-and-egg problem: data only exists after robots are deployed, and every company collects its own data without sharing .
2. Human-Like Learning Speed Professor Johns frames the core insight: “Intelligence is not so much about what you can already do, it’s about how quickly you can learn to do something new. Humans don’t need to see the same thing over and over before they can do it” .
3. Proven Performance in Tests In company-run, one-hour tests on self-resetting tasks, robots completed 74 block placements from 74 attempts, 78 screwdriver handovers from 78 attempts, and 97 bottle-uprighting attempts from 99—with the two failures attributed to potential collision and kinematic constraints .
4. The UK’s Largest Robotics Seed Round At $40 million, this is the largest seed round raised by a UK robotics company, signaling investor conviction that Machenta’s simulation-first approach can compete with US rivals valued at $10 billion to $14 billion .

Founder Background: The Imperial Professor and His PhD Student

Professor Edward Johns (CEO) is a Professor of Robot Learning in the Department of Computing at Imperial College London. He has spent six years investigating whether robots could learn new physical skills as quickly as humans, leading the Robot Learning Lab before founding Machenta .

Vitalis Vosylius (CTO) was Johns’s PhD student and co-founded Machenta with him. Together, they developed the techniques—including Implicit Graph Alignment (CoRL 2023) and Instant Policy (ICLR 2025)—that became the foundation for Machenta’s model .

The founding insight came from years of academic research. As Johns put it: “We’ve spent years trying to understand what it would take for robots to learn at the speed of a human. That is what instant robot learning is about” .

Founder Intelligence: What Founders Can Learn from Machenta

• Bet against the consensus: While competitors race to collect real-world data, Machenta trained entirely in simulation. The contrarian approach attracted top-tier investors precisely because it sidesteps the industry’s most expensive bottleneck .
• Build on years of research, not months: Machenta’s technology stems from six years of academic work at Imperial. The company didn’t start with a theory—it started with proven results from the lab .
• Solve the “learning to learn” problem: Machenta isn’t teaching robots specific tasks. It’s teaching them how to learn new tasks quickly. That meta-capability is more valuable and more defensible than any single task library .
• Keep the team small and focused: With just 11 employees, Machenta remains research-focused with no near-term commercialization plans. In deep tech, density of talent matters more than headcount .

Investor Intelligence: Why This Round Gets Premium Capital

Creandum and Union Square Ventures co-led the round. Kathi Wilhelm, Partner at Creandum, framed the thesis: “Very few teams in robotics can point to a body of work like this, which has repeatedly changed what people thought was possible. The team at Machenta has spent years breaking through this extremely hard problem, long before there was a company around it” .

Rebecca Kaden, Managing Partner at Union Square Ventures, added: “The industry is pouring enormous amounts of capital into collecting real-world data and training robots one task at a time, but Machenta is changing this. Whether this is for a robot in the factory or at home, we need general-purpose AI which can adapt quickly in the same way that humans can” .

IQ Capital and Hoxton Ventures also participated, bringing UK deep tech and early-stage expertise .

The Road Ahead for Machenta

Expansion Targets

With $40 million in capital, Machenta will:

• Expand its team at its new King’s Cross office and lab .
• Increase computing power and simulated data to train the model on more precise and complex tasks .
• Extend the range of tasks that can be taught through single demonstrations, moving beyond simple tasks like packing and tidying toward more complex manipulation .

The Commercialization Question

Machenta has no near-term plans to commercialize its technology. The company remains focused on research, and its model can learn some tasks immediately but more precise tasks and complex movements remain challenging . The $40 million seed is a bet that the research will translate into a defensible foundation model for physical AI—but the path to revenue is not yet defined.

Competitive Response

Expect:

• Physical Intelligence (valued at ~$11 billion), **FieldAI** (~$10 billion), and Skild AI (~$14 billion) to continue scaling their real-world data collection approaches .
• Skild AI has already unveiled a model that can perform unseen tasks after a single video demonstration and reported $100 million in annual recurring revenue .
• Machenta’s moat: simulation-first training + single-demonstration learning + cross-embodiment generalization + six years of academic research.

Conclusion

Machenta’s $40 million seed funding validates the thesis that the robotics industry’s data collection race may be the wrong race entirely. By training a foundation model entirely in simulation and enabling robots to learn from a single demonstration, Machenta is betting that algorithmic innovation can beat brute-force data collection.

Market tailwinds favor Machenta: the UK’s push to become a leader in AI-enabled robotics, investor appetite for contrarian deep tech bets, and the growing recognition that physical AI needs new approaches to scale beyond controlled environments. Whether Machenta’s simulation-trained model can handle the messy complexity of real-world tasks at commercial scale remains to be proven—but the company has the talent, the capital, and the thesis to try.

For deep analysis of physical AI funding, robotics innovation, and companies reshaping how machines learn, explore CREDX Media for comprehensive coverage of startup funding trends, valuation intelligence, and founder playbooks.

Want the latest funding news and market insights delivered straight to your inbox? Subscribe to CredX Letters for verified deal terms, valuation benchmarks, and daily macro signals.

Frequently Asked Questions

1. What is Machenta’s total funding and latest round size?
Machenta has raised $40 million in seed funding—the largest seed round ever raised by a UK robotics startup .

2. How much did Machenta raise in its seed round and who led the investment?
Machenta raised $40 million co-led by Creandum and Union Square Ventures, with participation from IQ Capital and Hoxton Ventures .

3. What is the valuation of Machenta?
Machenta did not disclose its post-money valuation .

4. Who are the major investors in Machenta?
Major investors include Creandum and Union Square Ventures (co-leads), IQ Capital, and Hoxton Ventures .

5. Who are Machenta’s founders?
Machenta was founded in 2025 by Professor Edward Johns (CEO), a Professor of Robot Learning at Imperial College London, and Vitalis Vosylius (CTO), his former PhD student .

6. What is Machenta’s business model?
Machenta does not yet have a commercial product or business model. The company remains focused on research, with no near-term plans to commercialize its technology .

7. How does Machenta’s technology work?
Machenta’s Vision-Demo-Action (VDA) model is pre-trained entirely in simulation. At deployment, it receives a single demonstration of a new task alongside camera observations and generates robot actions immediately, without task-specific fine-tuning or weight updates .

8. What traction justifies Machenta’s $40 million seed funding?
Machenta has demonstrated its model on six robot embodiments, with company-run tests showing 74/74 block placements, 78/78 screwdriver handovers, and 97/99 bottle-uprighting attempts. The research behind the company won the Best Paper Award at ICLR 2025’s Robot Learning Workshop .

9. What are Machenta’s plans for the $40 million seed capital?
Machenta will expand its team at its new King’s Cross lab, increase computing power and simulated data, and extend the range of tasks its model can learn .

10. Who are Machenta’s main competitors?
Machenta competes with US robotics AI companies including Physical Intelligence (~$11B valuation), **FieldAI** (~$10B valuation), and Skild AI (~$14B valuation) .

11. Who is the CEO of Machenta?
Professor Edward Johns serves as CEO and co-founder. He is a Professor of Robot Learning at Imperial College London and has led the Robot Learning Lab for six years .