Robotics freak-out
by Tom Nugent
Good morning John,
I don’t often hear the word “heretical” in interviews, but when I called Prototype's Andreas Klinger to chat about buzzy new physical AI lab Machenta, he threw it out there.
“It questions the axiom of ‘data is everything’ in robotics,” he told me. “For us they made it impossible to invest in any other model company [...] it’s heretical.”
Machenta, a spinout from Imperial College London, came out of stealth earlier this week with a $40m seed round led by Creandum and Union Square Ventures.
The company’s premise is this: a human demonstrates a task to a robot once, and said robot then completes that task autonomously.
Cofounder and CEO Edward Johns has been working with a team on this idea for years in his capacity as professor of robot learning at Imperial.
Its model — what he calls a vision demo action (VDA) model — “conditions on vision and a demonstration,” and then predicts the action needed. “It allows you to teach a robot with a single demonstration with no training afterwards,” he told me.
Machenta’s approach to model building flips a longheld narrative in robotics completely on its head. Billions of dollars have been invested into the sector on the hope that the scaling laws that led to ever more more capable LLMs (large language models) — more data plus more compute equals better performance — would hold in robotics. Companies have therefore been racing to collect reams of data on the understanding that the more they collect the better their models will be.
Now, “the constraint becomes imagining what is possible to teach [robots] versus data sets for them to learn how to do [a task]. It's effectively what vibe coding did to create software, but in the physical world,” Rebecca Kaden, a managing partner at Union Square Ventures, told me.
Like a lot of the new AI labs that have popped up in London recently, Machenta is pre-product and doesn’t yet have the details on what its business model will be. That’s a problem it’ll have to solve, shoulder-to-shoulder with its other buzzy peers in the UK capital.
“We’re just about to move to a new office in King’s Cross,” Johns said. “It feels like the place to be. I think it’s the centre of Europe in AI, and it’s certainly challenging Silicon Valley these days.”
Elsewhere today:
One in five companies are capturing nearly three-quarters of AI's economic gains
AI isn't rewarding effort evenly, it's rewarding a small number of companies disproportionately well. Sifted's new report, in partnership with London Business School, unpacks who's winning, who isn't and why. Download the report to find out which camp your company is in.
💰 Exclusive: Chip startup Fractile raised £540m in new Series C funding, company filings show
🌳 Departing UK entrepreneurship adviser Alex Depledge took aim at “whiny” founders and irrational economic policy following the Budget, telling founders: “There’s no magic money tree.”
ADVERTISEMENT
🔐 The European startups making agents secure — and useful.
🇫🇷 Is France’s startup nation falling out of love with Macron?
🦾 Why Creandum and USV just led a $40m seed round into a UK robotics lab with no product.
Talent tops European AI startups’ concerns as funding pressures mount
Debates about European AI tend to centre on compute, capital, and regulation. Ask the companies themselves what’s holding them back, however, and two challenges stand out: hiring great people and finding customers.
Sifted has once again surveyed companies from our latest AI 100 cohort, revisiting last year’s questions on growth challenges, fundraising, and international ambitions, alongside emerging issues around cloud providers, model strategies, and compute costs.
Some 78% put access to talent among their three biggest growth blockers, while 67% picked go-to-market challenges and competition — the same top two as last year, ahead of funding, compute, and regulation.
Elsewhere, the picture differs. Read the full breakdown.
Sifted AI 100: Meet Europe’s rising AI startups
Europe’s AI startups are taking on everything from drug discovery to the warehouse floor. Sponsored by N47, the second annual Sifted AI 100 spotlights 100 rising companies selected for their boldness, momentum and originality. Explore the ranking and meet the founders putting AI to work across the continent.
If you’d like to submit a deal, let us know.
Turin, Italy-based Argotec, which designs and builds small satellites of up to 350kg for civil, scientific and security missions, raised €50m in funding from Nuo Space.
The Hague, Netherlands-based Braid, which provides an API for businesses to add SIM and eSIM mobile connectivity to their apps, raised $38m in Series A funding. Octopus Ventures led the round and was joined by investors including KPN Ventures
Ghent, Belgium-based Ravical, which develops an AI platform that helps accounting and tax firms turn their experts' practices into standard workflows applied across clients, raised $17m in seed funding. MMC Ventures led the round and was joined by investors including Lakestar, Pitchdrive and Shapers.
Barcelona, Spain-based Allox, which develops an AI drug discovery platform to find small-molecule drugs for hard-to-target proteins, raised $10.5m in seed funding. Asabys led the round and was joined by investors including CDTI.
Hamburg, Germany-based MuRay, which develops imaging to inspect the inside of dense parts such as rocket engines and gas turbine components, raised €3.5m in funding. First Momentum Ventures and Caelora led the round and were joined by investors including Superangel, Mätch VC, Arcas VC, Tiny VC and Nordic Science Investments. It also raised €3.6m in public funding, including grant funding from the EU’s Horizon Europe programme.
Sifted Pro subscribers get exclusive access to all of the deals we track across Europe. Not subscribed yet? You can do that here.
Copyright © 2026 SIFTED (EU) LTD, All rights reserved.
Sifted EU Ltd, Thanet House, London, England, WC2R 1DA