Industry Notes

Industry Notes / Pharma / Chapter 14 of 20 · Pie

AI: real speed-up, often misplaced value

Speed-up is preclinical; cost and failure still sit in the clinic

What is still unclear?

Write what you still want explained. Questions go to the author and are not shown publicly.

Capitalized cost sketch
  • Target discovery12% · 12%
  • Molecule design12% · 12%
  • IND tox/CMC5% · 5%
  • Phase I–III61% · 61%
  • Approval & other10% · 10%

The pie is an illustrative capitalized-cost split: Phase I–III is still about three fifths, while AI today mostly touches targets and molecule design. Faster preclinical work cuts total cost only modestly; lifting II→III success would move NPV more—and that proof is still thin. Value more reliably flows to pick-and-shovel CXO; pure AI platforms are riskiest.

Source:研发成本结构为产业常用资本化成本拆分示意;AI 渗透判断见调研正文与公开案例(Exscientia、Recursion、英矽智能等)

Research notes, not investment advice.