Recursion Pharmaceuticals — AI supply-chain exposure

Recursion is an AI-first drug-discovery company, and in the value chain it is a consumer of compute rather than a supplier of it. It runs NVIDIA-built BioHive supercomputers over its own phenomics dataset — vast libraries of cell images — to search for drug candidates, turning raw accelerator power into biological hypotheses. That places it on the application layer as an integrator of GPUs and inference serving, and a provider of agent-style research workflows.

Its structural hook is that it represents the demand side of the build-out: the reason all that silicon and power gets bought is to run workloads like this. Recursion's own moat is not the chips, which it rents or houses, but the proprietary experimental data it feeds them — a dataset a competitor cannot easily replicate. The model weights it as a downstream name whose exposure rises and falls with both compute availability and the value the market assigns to AI-native science.