a16z's Contrarian Take: AI Devalues Drug Discovery, Shifts Value to Clinical Trials

FinanceYF5 · x · 2026-07-04

a16z presents a contrarian view: AI has led to an oversupply of drugs, making each individual one less valuable. Over the past decade, the drug pipeline has doubled, but the number of annual FDA approvals remains stuck at around 50, indicating the bottleneck lies in clinical development rather than discovery.

An increasing number of drugs are targeting the same mechanisms—two decades ago, a single target had 2-3 drugs, compared to 7-16 today. PD-1 and GLP-1 each already have over 100 projects, diluting the value of individual discovery assets and continually lowering preclinical valuations. The maturation of three frontiers is staggered: discovery (with hourly feedback) is commoditized first, toxicity next, and clinical efficacy (requiring years for endpoints) is the hardest and matures last.

The value inflection point hinges on Phase 2 readouts: a drug with $1 billion in peak sales sees a 4x valuation difference before and after Phase 2. The recommendation is to model based on the 30% industry average, compile 10 independent Phase 2 projects (a mere 2.8% chance of zero return), and bet on the bottleneck rather than the path of model advancement. The winner is not the one who discovers the most drugs, but the one who first determines which drug is worth developing.

Original post →

More from Venture

Venture channel →