Anthropic is running its own wet lab to test AI biology in the real world

Anthropic has quietly become a company that runs biology experiments at the bench. TechCrunch reports that the firm is operating a wet lab in the Bay Area, built around Coefficient Bio, a stealth biotech startup it bought in April 2026 for about $400 million. The reasoning is that you cannot claim AI is advancing medicine on the strength of computational models alone. Eric Kauderer-Abrams, who leads life sciences at Anthropic, puts it plainly: the final test still happens in real lab work.

The lab focuses on fundamental biology rather than chasing specific drugs, with work on protein design and biomolecular modeling, and it pairs internal research with partners including Novo Nordisk. Leadership talks about AI helping to cure major diseases within five to ten years. That framing sits awkwardly next to Anthropic's public warnings about catastrophic AI risk, a tension the investor Chamath Palihapitiya was quick to point out. This is a bet on capability, not a safety project.

Why it matters

An AI lab owning physical experiments changes what "AI for science" means: the model's predictions now get checked against real cells and proteins in-house, which is a higher bar than a benchmark. If you work in biotech, Anthropic is now a competitor and a possible partner at the same time, not just a supplier of a chatbot.