NVIDIA
NVDAcompute· NASDAQNVIDIA – AI compute infrastructure: Custom graphics processing units, CUDA platform, and NVLink interconnects.
Why it matters
The company sells a working scale-up system, a software stack, a supply-chain coordination layer, and a seeded ecosystem of domain foundation models (Nemotron, BioNemo, Alpamayo, Groot, climate science). Any single-layer competitor is competing against a decade of joint optimisation.
The read
Levered to AI compute demand and rack-scale system supply (NVL72 → Vera Rubin → Feynman).
Bull case: Sells a working rack-scale system plus a decade-deep CUDA ecosystem; any single-chip rival competes against the joint optimization, not just the GPU.