Anthropic Quietly Builds Wet Lab, Advancing AI Drug Discovery with Focus on Rare Diseases
nashnova research
Anthropic has secretly set up a wet lab in the San Francisco Bay Area, moving its AI drug-discovery effort from pure computer simulation into real physical experiments — this means the company, eyeing an IPO at a valuation of up to $2 trillion, is extending its bet from 'smarter models' to 'models that run experiments.'
What is a wet lab, and why does Anthropic need one?
A wet lab — a physical laboratory with test tubes, petri dishes, and real biological samples, as opposed to a "dry lab" that only runs computer simulations — is now operational in the Bay Area.
Eric Kauderer-Abrams, Anthropic's head of life sciences, confirmed: some experiments run in-house, others with external partners.
This means → Anthropic is no longer content to predict molecular structures with AI alone; it wants those predictions tested in real experiments. Closing the loop from computation to bench work is the critical step for AI drug discovery to move from papers to products.
Letting Claude directly command lab robots — is that realistic?
One goal: have the flagship Claude model direct robotic units to execute scientific experiments with limited human intervention.
Kauderer-Abrams calls this "an area with the potential to meaningfully accelerate many different workflows," but stresses it remains at a "very early stage."
In plain terms = the ambition is for AI not just to advise but to physically act — yet the company simultaneously insists human oversight is essential for safety, signaling that a fully autonomous lab is nowhere close.
Why rare diseases specifically, and why stop before clinical trials?
Kauderer-Abrams drew a clear line: focus on rare diseases, stop at preclinical research, do not enter clinical trials.
This means → Anthropic picked a gap that traditional pharma has long ignored due to insufficient commercial returns, while deliberately staying off the battlefield where it would compete head-on with Pfizer or Novartis.
This reflects a calculated positioning: use AI's speed advantage to tackle work Big Pharma won't, and use the "no clinical trials" boundary to manage regulatory and commercial risk.
A flurry of supporting moves — money, people, and product landing together?
Anthropic launched a dedicated software product, Claude Science; added Novartis CEO Vas Narasimhan to its board; and acquired startup Coefficient Bio for roughly $400 million in stock.
In plain terms = software is the tool, a board seat is the resource bridge, the acquisition is the team — three moves made simultaneously, signaling a systematic bet, not a toe in the water.
All of this comes as Anthropic prepares an IPO at a potential valuation of $2 trillion; the life-sciences division is clearly being positioned as a key piece of the IPO narrative.
Safety controversy and personal motivation — what is the other side of the coin?
In the past two weeks, some Anthropic researchers publicly warned that AI could cause human extinction; the company itself disclosed cases where its systems could potentially be used to develop bioweapons.
Kauderer-Abrams acknowledged the company is constantly balancing "making models better to advance drug discovery" against "deploying cautiously to control risk."
CEO Dario Amodei has shared that his father died from a disease whose cure arrived only years later — that experience is one of his personal motivations for pushing AI into drug discovery. Capability and risk coexist; ultimate results remain to be proven.
市场有风险，内容仅供研究参考，不构成投资建议。