Biological Data Flywheel
Amit@amit
−7.8%
since Oct 7
The idea
I believe cheap AI reasoning makes proprietary biological data the real moat, so I'm backing companies that generate real experiments and real measurements faster than anyone else, letting better data train better models that design better experiments. The basket splits across the data-rich compounders already spinning this flywheel, a decade-long core of tools labs depend on to read, write and automate biology, and large instrument makers as ballast, with no single name dominating.
What it owns
- Large, profitable instrument and automation makers that supply every lab, steady enough to hold through a long biological data expansion. · DanaherIts bioprocess and tools businesses get paid when biology moves from model to manufacturing, the physical end of the flywheel., Thermo Fisher ScientificIf labs everywhere are generating more biological data, they buy more of everything it sells, and its guidance points to broadening demand., Agilent TechnologiesIt sells the instruments and lab automation that make generating biological truth faster and cheaper, the exact capability this take says wins., BrukerIt makes the high-end measurement tools, mass spec and imaging, that produce data nobody can get any other way., RevvityIt sells multiomics tools and reagents across the whole read-and-measure stack, a broad way to own rising biological data volume., AzentaIt automates sample storage and handling, the physical plumbing that every data flywheel needs once experiment volume scales.34%
- Companies already sitting on proprietary biological datasets that are visibly improving their AI and their products right now. · Tempus AIIt keeps collecting patient genomic and outcomes data no one else has, and that data trains models that make its reports more valuable, which brings in more data., Recursion PharmaceuticalsIt runs the industrial wet-lab loop where automated experiments generate proprietary biological measurements, models learn from them, and the next round gets smarter., NateraEvery test it runs adds to a circulating tumor DNA dataset that improves the test itself, the compounding data advantage in its purest form., Guardant HealthIts liquid biopsy database grows with every blood draw, and recent studies show that database proving out in new cancers., SchrodingerIt pairs physics simulation with machine learning, so each drug program generates computed and measured data that feeds back into the platform., AbCellera BiologicsIts antibody discovery engine produces proprietary measurements on millions of antibodies per program, data that compounds across every partner deal it signs.32%
- Sequencing and measurement companies that get paid every time anyone, anywhere, generates biological data. · IlluminaIt gets paid almost every time anyone generates biological truth, since most sequencing still runs on its machines., QiagenIt sells the sample prep and testing kits that turn physical specimens into usable data, the first step of every flywheel in this take., 10x GenomicsIt sells single-cell and spatial tools that produce the richest biological measurements available, exactly the proprietary physical feedback this take says AI makes valuable.13%
- The compute that keeps making biological reasoning cheaper, the force driving the whole take. · NVIDIAIt supplies the compute that makes biological reasoning cheap enough for the physical data to become the scarce input.7%
- DNA synthesis and engineered biology, the tools for running the experiments that AI designs. · Twist BioscienceIt collects rent on writing DNA, the step where an AI-designed experiment becomes a physical one., Ginkgo BioworksIt is the outsourced foundry where companies run the physical experiments their models design.6%
- Cash8%