AI generates thousands of molecular designs overnight. Every single one needs wet lab validation. The bottleneck just shifted from algorithms to the infrastructure providers.
Twist Bioscience ($TWST): AI-driven orders up triple digits for two consecutive years. FY26-27 projections show AI pharma discovery orders jumping from $25M → $50M → $100M. Over one-third of FY25 order growth already comes from AI clients.
GenScript: AI drug discovery revenue doubled YoY in H1 2026, marking three straight half-year periods of rapid growth. Channel checks indicate their experimental validation capacity hit ~8,000 designs per day, targeting 16,000 by year-end. Orders are accelerating into execution.
Why the bottleneck moved: AI made molecular design nearly free at the margin, but every design still requires physical wet lab verification. Failed experiments matter just as much—AI models need all results fed back as training data to iterate smarter. Every success and every failure becomes a paid data point.
The value chain breaks down: Model generates design → DNA synthesis → protein production → experimental validation → preclinical testing. DNA synthesis, protein production, and CRO services are capital-heavy, time-intensive, and impossible for AI to replace. Compute is abundant. The constraint is now data generation and validation.
In this AI pharma wave, the shovel sellers get paid first.
Twist Bioscience ($TWST): AI-driven orders up triple digits for two consecutive years. FY26-27 projections show AI pharma discovery orders jumping from $25M → $50M → $100M. Over one-third of FY25 order growth already comes from AI clients.
GenScript: AI drug discovery revenue doubled YoY in H1 2026, marking three straight half-year periods of rapid growth. Channel checks indicate their experimental validation capacity hit ~8,000 designs per day, targeting 16,000 by year-end. Orders are accelerating into execution.
Why the bottleneck moved: AI made molecular design nearly free at the margin, but every design still requires physical wet lab verification. Failed experiments matter just as much—AI models need all results fed back as training data to iterate smarter. Every success and every failure becomes a paid data point.
The value chain breaks down: Model generates design → DNA synthesis → protein production → experimental validation → preclinical testing. DNA synthesis, protein production, and CRO services are capital-heavy, time-intensive, and impossible for AI to replace. Compute is abundant. The constraint is now data generation and validation.
In this AI pharma wave, the shovel sellers get paid first.