−8.6%
since Oct 6
The idea
I believe AI will accelerate life sciences the way it did math and coding, but biology can't verify answers in software, every hypothesis still needs a physical lab. So I'm positioned in the bottleneck that gets paid on every attempt: consumables and supply chain, sequencing and data generation, lab automation, AI-for-bio software, and lab real estate. I hold no single name above roughly 8%, keep a cash buffer, and plan to hold for years. If wet-lab work itself gets automated away as the limiting step, I'd rethink the whole basket.
What it owns
- Every experiment burns through reagents and consumables regardless of who designs it, so these companies get paid on each attempt. · Thermo Fisher ScientificIt supplies the instruments and consumables that every bench-top experiment requires, generating revenue per attempt regardless of source., DanaherIt owns the bioprocessing tools that convert validated ideas into actual drug material, a step no computational model can replace., AvantorIt supplies reagents and basic laboratory chemicals that get consumed on every experiment, regardless of discovery method., RepligenIt makes filtration and chromatography consumables that biologics manufacturing uses on every batch.26%
- AI models in biology are only as good as the experimental data they train on, and sequencing and synthesis are how that data gets created. · IlluminaIt reads DNA sequences, the primary method for generating the biological training data that AI models depend on., Twist BioscienceIt synthesizes DNA sequences to physically test what computational models propose, completing the write side of the validation loop., Ginkgo BioworksIt operates outsourced laboratory capacity, letting companies with computational ideas but no bench run experiments at scale., 10x GenomicsIt generates single-cell resolution data, the level of granularity AI biology models increasingly demand for meaningful insights., Pacific BiosciencesIts long-read sequencing captures complete genomic information, providing richer data that AI models increasingly value as they mature.21%
- If AI multiplies the number of hypotheses to test, robots that run experiments continuously become the rate-limiting layer. · AzentaIt automates sample storage and handling, the essential infrastructure labs need once experiment volume outgrows human capacity., TeradyneIts collaborative robots perform experiments around the clock, providing the automation layer that lets labs scale throughput., Agilent TechnologiesIt sells automated workflows and analytical instruments that increase throughput, letting labs run more experiments per scientist.18%
- Software that translates AI capability into actual experiment designs sits at the front of the pipeline and steers wet-lab spending. · SchrodingerIt builds physics-based computational models to predict which molecules are worth testing, serving as the decision layer before wet-lab spending., Recursion PharmaceuticalsIt runs the closed loop directly: computational models propose experiments, its automated labs test them, results retrain the models., CertaraIt simulates drug behavior and trial outcomes, helping companies allocate limited physical testing budgets to the most promising candidates., Tempus AIIt curates clinical and molecular datasets to serve as training fuel for AI biology, creating value from data aggregation.16%
- Physical lab space and building capacity, not ideas, become the constraint when AI floods the system with experiments to run. · Alexandria Real Estate EquitiesIt owns laboratory buildings and facilities, the physical infrastructure that must expand if experiment volume truly multiplies., Jacobs SolutionsIt designs and constructs laboratory and biomanufacturing facilities, providing the physical infrastructure for new capacity build-out.9%
- Cash10%