Why we backed Yann Lecun's $1B seed round.
Why a seed funf would back a $1b seed round
Hey everybody, Guillermo here. Today we have Michael Stothard, investor at firstminute capital — one of Europe’s most active seed funds, backed by 130+ unicorn founders — sharing why his firm just backed the biggest seed round in European history.
In this article, you’ll learn:
Why a seed fund wrote a cheque into a $1.03bn round
The promise of Ai for the physical world
Why the next world-leading AI company might have their HQ in Paris
What a $100 trillion physical AI market looks like up close
Now, over to Michael.
—
By Michael Stothard, investor at firstminute capital, a $500m early stage fund backed by 130 unicorn founders. The fund has backed five AI unicorns at seed. Michael was previously a foreign correspondent for the Financial Times and founding editor of Sifted.
We are a pre-seed and seed fund. We normally back founders raising $2-3m for the first time. Not $1bn.
This week we participated in the $1.03bn seed round for AMI Labs — the AI research company founded by Yann LeCun and Alex LeBrun, aiming to build the generation of AI systems that comes after LLMs.
Why?
1. An incredible team we have know for years. At firstminute we backed Alex’s last company, Nabla, at pre seed. They have since gone on to raise Series C and built a product used by over 85,000 clinicians across 130 organizations. We’ve known Yann — one of the “godfathers” of AI and Turing award winner — for years as well through the Founders Forum network. In the end, venture is a people game and these are incredible people.
2. The ambition is genuinely European. It’s extraordinary to see a company of this scale and ambition come out of Europe. If AMI gets this right, it won’t be a regional champion — it will be up there with one of the most important technology companies in the world, built out of Europe. The prize is physical AI: systems that can understand and operate in the real world, powering the next wave of industrial automation, robotics, and autonomous systems. Nvidia estimates that market at $100 trillion.
3. There are things LLMs just can’t do. A house cat has roughly the same number of synapses as an LLM has parameters. And yet it can stalk a mouse, remember where it went, understand it still exists behind the sofa, and execute a perfectly-timed pounce. LLMs cannot do any of this! Yann’s answer is JEPA: instead of learning from text, these systems learn from video and sensor data, building compressed internal models of how the world actually works — cause and effect, physics, object permanence, planning. The result is the promise of an AI that can operate in the physical world, not just describe it. That’s what unlocks robotics, industrial automation, autonomous systems — markets that language models structurally cannot touch.
We don’t normally do rounds of this size. But we’ll always back the best founders we know swinging at the biggest problems in the world. We are honoured to be supporting with this one alongside many others like Cathay Innovation, Greycroft, Hiro Capital, HV Capital and more.
Hope you found this valuable. Talk soon.
Cheers, Guillermo