Machenta
An Imperial College London spinout training a robot model entirely in simulation so that robots learn a task from one demonstration.
Europe's most interesting startup stories. Every week.
No spam. Unsubscribe anytime.
Podcast Episode
๐๏ธ Deep Dive
Machenta, Atlas Bio, Agora Labs โ $40M for robots, $10M for virtual trials and โฌ4.3M for clinical data โ Daily News October 27, 2026
Listen to the episodeAbout Machenta
Machenta comes out of Imperial College Londonโs Robot Learning Lab, led by Professor Edward Johns (CEO, who previously led robot learning at Dyson), with his former PhD student Vitalis Vosylius, first author of the Instant Policy research, and other lab colleagues; the company was incorporated in London in July 2025, after six years of research. It develops a foundation model pre-trained entirely on simulated data, with no real-world data or human demonstrations, which after a single demonstration performs the task and adapts to variations without further training; the team has shown the same model learning tasks such as packing and tidying across different robots, including humanoids. On October 27, 2026 it raised $40 million in a seed round co-led by Creandum and Union Square Ventures, described as the largest seed ever for a UK robotics startup, to scale the model and grow the team at its new Kingโs Cross office and lab. Customers and the commercial model have not been disclosed.
The Story
Incorporated in London in July 2025 by Edward Johns and Vitalis Vosylius, after six years of research in Imperial College Londonโs Robot Learning Lab.
How Machenta works
Business Model
B2B technology for robot makers and users; commercial model and pricing undisclosed.
Products & Services
Key Products
- Modello di fondazione per il robot learning (Instant Robot Learning)
Core Use Cases
- One-demonstration robot learning
- Automation of variable tasks such as packing and tidying
- Simulation-based pre-training
Market & Clients
Key Customers
Robot makers and users; no customer named.
Geographic Presence
How Machenta competes
Competitive Advantages
- Model trained without real-world data
- Immediate transfer to new tasks and different robots
- Backed by Imperial College research
How Machenta grows
Growth Strategy
Scale the model and grow the team at the Kingโs Cross lab.
Distribution Model
B2B integrations with robot makers and users.
Moat (Defensibility)
Proprietary research, synthetic data and simulation pipelines, still to be proven at commercial scale.
Regulatory Context
Workplace robot safety and liability for autonomous actions.
Key Risks
- Technology still at research stage
- Performance not proven at scale
- Dependence on compute capacity
- Customers and revenue undisclosed
Strategic Insights
- Simulation moves the bottleneck from real-world data toward compute and validation quality.
Funding & Investors
Total Funding
Funding Rounds
- 2026 - Seed: $40M
Lead: Creandum, Union Square Ventures
Key Investors
Founding Team
Founders
Key Metrics
Lessons from Machenta
- A contrarian data strategy can become an advantage when backed by reproducible results.
Similar Startups
Nscale
AI cloud and data centres with a registered office in London and a campus in Norway. 2026 Series C at a $14.6 billion post-money valuation; IPO only proposed.
Read AnalysisIsembard
London company founded in 2024 that makes precision components for aerospace, defence, energy and robotics through a network of owned and franchised factories. $50M Series A in 03/2026.
Read AnalysisFractile
Fractile builds chips and systems dedicated to AI inference, aiming to reduce cost and latency of advanced models and make agentic AI economically sustainable at scale.
Read AnalysisSources
- Imperial AI spinout Machenta raises $40 million to make robots instant learners — Imperial College London
- Backing Machenta: the era of Instant Robot Learning — Creandum
- Companies House โ MACHENTA LTD (16594535) — Companies House
- An 11-person startup raised $40 million to challenge robotics' data-hungry giants — Business Insider