AI drug design is flooding Chinese labs with wet-lab work
Anthropic's Claude designed 1,320 binding proteins against a dozen targets, which Adaptyv Bio and Twist synthesized and tested; 354 were confirmed to bind, a hit rate near 30% versus the traditional 10%–15%. The article argues that as AI pushes the marginal cost of molecular design toward zero, the first orders flow to lab-service providers: GenScript's AIDD revenue doubled year over year in H1 with gross margins about 20 points above its traditional protein-expression business, while XtalPi posted 803 million yuan in 2025 revenue, tripled and its first full-year profit. Meanwhile Washington is trying to pull early-stage R&D and clinical capacity back onshore via HHS's TrialBlazer plan and the BIOSECURE Act.
It suggests the binding constraint on AI drug discovery is shifting from algorithms to wet-lab and clinical capacity, which is concentrated in China and is now a focus of US-China policy.