Methodology: layers follow the closed Design · Make · Test · Learn loop of a self-driving
lab — AI designs, robots synthesize, instruments measure, data trains the next cycle. The
bottleneck score ranks how supply-constrained / hard-to-replicate each layer is, not company quality.
The reframe: compute is abundant and the AI model is commoditizing, but high-throughput measurement
— the layer that generates the experimental data models train on — is the moat, which is why it is the one 5/5 layer.
Company lists are representative, not exhaustive. All layer framing, scores, and takes are BEP Research.
DISCLOSURE — Long NVDA, LITE, CRDO, TSEM, LSCC, ALAB, WOLF, ORCL (2027 LEAPS), BE.
Angel / private holdings in Terray Therapeutics and VistaPath Bio (flagged on the map).
This is research, not investment advice. Do your own work.