AI Drug Discovery Bottleneck
Alvin Hsia@alvin
−2.0%
since Sep 30
The idea
I believe AI labs are about to turn their best models loose on drug discovery, generating far more candidate molecules than the physical world can currently test. Since none of that can be simulated away, I own the bottleneck: preclinical CROs first, then lab instruments, synthesis and sequencing tools, bioprocessing, and trial infrastructure behind them, spread across many names rather than one bet. I hold some cash to add on volatility, and I'm underwriting the next year or two, when that molecule surge should first show up in bookings.
What it owns
- More candidate molecules means more wet-lab experiments, and these firms sell the machines and consumables that run them. · Thermo Fisher ScientificIt sells the instruments, reagents, and services that pharma labs run on, making it a broad way to own rising wet-lab activity if AI-designed candidates flood the discovery pipeline., DanaherIt operates in both lab instruments and bioprocessing, collecting revenue when molecules are studied and again when they're manufactured at scale., Mettler-Toledo InternationalIt sells precision weighing and measurement instruments that every synthesis and quality-control step depends on, making it a picks-and-shovels way to own rising lab activity in drug discovery., Waters CorporationIt makes the mass spectrometry and chromatography tools that characterize new molecules, so instrument orders rise as a wave of AI-designed candidates moves from computational design to lab validation., Agilent TechnologiesIts analytical instruments confirm that a designed molecule is actually the molecule that was synthesized, and more candidates means more characterization runs and consumables demand., AzentaIt automates sample handling and storage, the operational backbone that scales when candidate numbers explode and labs need to process vastly more experiments., Bruker CorporationIts structural biology and imaging tools verify what an AI-designed molecule actually looks like, which matters as much or more when a computational model dreamed it up.31%
- Every AI-designed molecule has to be tested in animals and then humans, and these companies run those studies for hire. · Charles River LaboratoriesIt runs preclinical animal studies for every AI-designed molecule before human testing begins, making its bookings an early signal of whether frontier labs are deploying models at scale into drug discovery., Medpace HoldingsIt runs trials for small biotechs most likely to sprint AI-originated candidates into the clinic, so backlog growth reflects whether a surge of new molecules is reaching human testing., IQVIA HoldingsIt owns the data networks and clinical sites that trials depend on, so it collects on nearly any AI-originated molecule that reaches humans as long as sponsors keep routing studies through its platform., ICON plcIt operates global capacity to run human trials at scale, a bottleneck that cannot be automated and tightens as more AI-designed candidates enter clinical testing.26%
- A molecule designed in software still has to be physically made and read, which is these companies' whole business. · Twist BioscienceIt synthesizes custom DNA to order, and AI-driven discovery consumes designed DNA at scale to test computational predictions in the lab., IlluminaIts sequencers read out the biology in every AI-designed experiment that touches cells, so sequencing volumes grow as discovery activity accelerates and candidates move through validation., Ginkgo BioworksIt operates as an outsourced wet lab where AI-driven discovery customers can run physical experiments that models cannot, earning revenue only if it lands customers before cash constraints bite.12%
- Drugs that advance need to be manufactured and put into vials, a capacity bottleneck no model can replace. · RepligenIt supplies the filters and consumables used in every biologic manufacturing run, so demand recurs with each batch as AI-originated molecules scale from trials toward production., West Pharmaceutical ServicesIt makes the stoppers and seals on nearly every injectable vial, a manufacturing layer that scales with more drugs reaching trials and production regardless of how they were discovered.11%
- A wave of new clinical programs runs on the same few software platforms that manage trials and regulatory filings. · Veeva SystemsIts software is the default platform pharma runs trials and regulatory filings on, so a surge of new candidates means more seats and studies flowing through its systems., CertaraIt simulates trial design and dosing optimization, work that sponsors purchase more of when candidate counts jump and they must choose which molecules to advance.8%
- Cash12%