Latent Labs' $50M bet: ex-DeepMind scientist uses AI to make biology programmable

Builds AI foundation models that generate and optimize proteins — enzymes, antibodies — for biotech and pharma partners, via direct model access or project partnerships.

Generative models trained on biology can design therapeutic proteins on demand, shrinking drug discovery from years of wet-lab iteration to computation.

$10M seed plus $40M Series A (February 2025), $50M total

Simon Kohl was a senior research scientist on DeepMind's AlphaFold2 team, co-led the protein design team and set up DeepMind's wet lab at the Francis Crick Institute. He left DeepMind at the end of 2022, incorporated Latent Labs in London in mid-2023 and hired a small team to focus on generative models for protein design.

In stealth, Latent Labs raised a $10M seed from Kindred Capital, 8VC and Pillar VC and recruited roughly 15 people from DeepMind, Google and Microsoft across London and San Francisco. In February 2025 it exited stealth with a $40M Series A co-led by Radical Ventures and Sofinnova Partners, with angels including Google chief scientist Jeff Dean, Cohere founder Aidan Gomez and ElevenLabs founder Mati Staniszewski.

The company deliberately avoids developing its own drug candidates; revenue is meant to come from partners. As of the February 2025 announcement it was spending the round on compute, UK and US hiring, and building the capacity to service partnerships — no named pharma deals or products yet.

AlphaFold showed machine learning could predict protein structure at scale, opening the door to generative protein design.

Protein design is vast, unconverged white space where a focused team can out-run organizations doing many things at once.

A wet lab alongside the model team gives real-world feedback, so predictions get validated instead of trusted blindly.

Skipping in-house drug development keeps the company aligned to one thing: selling the model and the partnerships around it.

When a large lab spreads across many frontiers, the focused spin-out can own the white space — and pairing frontier models with a small wet lab closes the validation loop most model shops lack.

As of September 2026 the only published facts are from the February 2025 launch: Latent Labs was spending the $50M on compute for larger models, growing its London and San Francisco teams, and building commercial capacity for partner programs. No commercial products, named pharma partnerships or therapeutic candidates had been disclosed in the sources reviewed.