Science Is Cooked
Andrew Chang@achang
The idea
I believe AI is about to take over the grunt work of scientific discovery, running experiments and designing molecules with less human hand-holding. So I weighted this toward the lab instruments, automation, and robots that physically run those experiments, alongside the compute and simulation companies every AI scientist depends on. A smaller slice sits in AI-first discovery firms that could pay off if their pipelines keep working, plus the larger drugmakers folding AI into their own research.
What it owns
- Every experiment an AI designs still has to be run on real instruments with real supplies, and these companies sell them regardless of which AI tools win. · Thermo Fisher ScientificIt supplies the reagents, consumables and instruments that run lab experiments at scale, making it a beneficiary of any increase in experimental workload from AI-driven research., Danaher CorporationIt announced plans for AI-powered autonomous labs that run experiments designed by AI, turning software discovery into physical hardware at scale., Agilent TechnologiesIts analytical instruments and software power the measurement workflows embedded in automated chemistry and biology, positioning it to capture volume growth if AI research generates more experiments to run., Waters CorporationEvery molecule an AI proposes has to be measured and checked, and its mass spectrometry and chromatography machines do that work., IlluminaAI scientists in biology learn from huge amounts of genetic data, and its sequencers produce most of the data used in research., Twist BioscienceWhen an AI designs a gene or protein, someone has to make the DNA to test it, and it sells custom DNA fast and cheap to researchers., Bruker CorporationIt makes high-end instruments for seeing molecular structure, the kind of hard data AI models need to check and improve their predictions.35%
- AI models doing science, from protein folding to lab design, train and run on specialized chips and simulation software, so these companies get paid as discovery moves onto computers. · NVIDIA CorporationNearly every AI model doing science trains and runs on its chips, from protein folding to lab robotics., Alphabet Inc.It owns DeepMind and Isomorphic Labs, which are trying to discover drugs with AI and publish real scientific results., SynopsysIt sells simulation software that lets scientists test designs on computers before building them, the same simulate-first approach AI science runs on., Cadence Design SystemsIts software simulates both chips and molecules, and its molecular modeling tools are used in drug design by researchers.24%
- Large drugmakers using AI in their own labs turn cheaper, faster discovery into more products and higher profits without needing to sell the tools. · Eli Lilly and CompanyIt is spending heavily to build AI into its research, and as one of the largest drugmakers can turn cheaper, faster discovery into substantial profit., AstraZenecaIt has built automated and AI-guided labs into its own research, so it gets the payoff from AI science without needing to sell the tools., SanofiIt sources AI discovery through partners and recently selected a drug candidate from its collaboration work to develop further.16%
- These companies are built around AI finding new drugs, so they gain the most if the approach works and lose the most if it fails. · Recursion PharmaceuticalsIt is a pure bet that AI-found drugs work, with early candidates from its AI collaboration showing progress in human trials., SchrödingerIt uses physics-based computer models to design drugs and sells that software to other drugmakers, so it gets paid either way AI discovery wins., AbCellera BiologicsIt uses AI and automated screening to find antibody drugs, a direct test of whether machines can do discovery faster than people., Absci CorporationIt uses generative AI to design proteins from scratch, treating AI-driven molecular design as the core of drug discovery.13%
- Cash12%