Q3 2026 Convergence Report

Molecules got cheap. Proof got a passport.

In Q3 2026, teams using general-purpose AI models produced lab-confirmed protein binders at up to 3.5 times the industry hit rate, and big pharma signed to deploy agentic design tools at scale. Proving a drug works in humans stayed as slow and as expensive as before. Then Washington started pricing where that proof gets generated, in the same quarter the Phase III of the best-known AI-discovered drug started entirely at Chinese sites.

  1. 354Lab-confirmed binders from 1,320 AI designs, 14 of 15 targets, 22.6–35.1% hit rates vs a 10–15% normAnthropic / Adaptyv data via eWeek, Aug 2026
  2. 47 / 47Rentosertib Phase III sites located in China · First patient dosed Sep 10News-Medical, Sep 2026
  3. $110BChinese innovative-drug out-licensing, 81 deals in H1 2026, a recordCCTV via The Standard, Jul 2026
  4. 50%Proposed FDA application-fee cut for data with a US-anchored Phase 1PDUFA VIII letter, Aug 2026
Members