AI-Driven Lab Boom
David Greenstein@davidgreenstein
−3.2%
since Oct 9
The idea
I believe AI is about to multiply the number of biology experiments run, since generating ideas gets cheap but testing them still needs real labs, DNA, reagents and fresh data. So this basket owns the picks and shovels: steady anchors in lab equipment, the companies writing and reading DNA, tools that measure cells and proteins, automated and outsourced labs, plus a smaller slice of AI drug discovery companies and research software. It holds roughly 8% cash and stays diversified across many names rather than betting on any single one, meant to be held for years as this plays out.
What it owns
- The largest sellers of lab equipment and consumables gain from almost any increase in experiments, which steadies returns across economic cycles. · Danaher CorporationIt makes lab instruments and consumables used in drug research, and is building an AI-powered autonomous research lab to directly operate high-throughput testing., Thermo Fisher ScientificIt sells the instruments, reagents and contract research services that almost every lab buys, so growth in experiments across AI-driven facilities flows through as increased orders., Agilent TechnologiesIt sells lab instruments plus the columns, kits and service contracts that get used up and renewed as labs run more tests and process more samples.25%
- AI-designed experiments start with synthesized DNA and end with sequencing, both of which are paid per run, so experiment volume drives revenue. · Twist BioscienceIt makes custom DNA to order, the raw material for testing genes and proteins that AI systems design, and order volume scales with experiment count., IlluminaIt sells DNA sequencers and gets paid again for the chemistry used in every run, so more experiments that read DNA means recurring revenue., Pacific BiosciencesIt provides long-read sequencing that reads longer stretches of DNA and RNA, enabling better analysis of complex genetic regions that AI models need to understand.17%
- Drug developers that run their own automated labs are the heaviest users of this whole ecosystem, and a successful drug candidate adds significant upside. · Recursion PharmaceuticalsIt operates a highly automated lab that photographs massive numbers of cell experiments to feed its AI engine, demonstrating round-the-clock automated testing., Generate BiomedicinesIt uses AI to design new proteins and tests them in its own labs, representing the full cycle of computational design needing wet lab validation., AbCellera BiologicsIt operates an antibody discovery engine that screens large numbers of immune cells to find drug candidates for itself and partners, with upside if programs advance., AbsciIt designs antibodies with AI and validates them in its own wet lab, a direct example of how computational drug design requires real laboratory testing.14%
- Testing an idea means measuring what cells and proteins actually do, which requires instruments and reagents that get consumed with each experiment. · 10x GenomicsIt makes tools that measure what happens inside individual cells and where they sit in tissue, the detailed data that AI biology models feed on., BrukerIt builds mass spectrometers and instruments that measure proteins and cells in fine detail, and universities and drugmakers replace these instruments as research scales.13%
- These companies automate experiments or run them for others, so they directly benefit from labs operating at higher throughput and around the clock. · RevvityIt makes the screening machines, reagents and lab software used to run large batches of cell experiments, the high-volume testing that AI ideas require., Charles River LaboratoriesIt runs studies for drugmakers on contract, so it gets busier when more drug ideas need testing, and partnerships expand its automation capabilities., AzentaIt runs automated freezers and sample services that store and track biological samples, so more experiments means more samples requiring storage and management.13%
- Software that designs molecules and manages lab results becomes more valuable as experiment data accumulates and informs better compound selection. · SchrödingerIt sells software that designs and scores molecules on a computer before lab testing, plus it runs its own drug programs, adding upside if those advance., CertaraIt sells software that simulates how drugs behave in the body, helping researchers decide which experiments to run and where to focus resources next.8%
- Cash9%