Nvidia Is Bigger Than Jensen Huang - The Dispatch

Nvidia's quiet de-personalization is more structurally significant than any quarterly print. Once CUDA, NVLink, and InfiniBand form a closed loop, the valuation anchor shifts from chipmaker to compute operating system. Jensen Huang's personal narrative is yielding to platform lock-in — the moat has migrated from the individual to the architecture. Upstream, CoWoS advanced packaging (Taiwan, China capacity), HBM3E (SK Hynix/Samsung), and 800G optical interconnect are deeply coupled. Export controls paradoxically reinforce this coupling: restricted markets cannot access equivalent process nodes or packaging resources, so Nvidia's scarcity premium is inadvertently amplified by policy. Competitive threats from AMD MI300 or custom ASICs (TPU, Trainium) are systematically overstated. The real barrier isn't silicon — it's the migration friction of millions of CUDA code lines. Over the next 12–24 months, Nvidia will accelerate its pivot from selling GPUs to compute-as-a-service (DGX Cloud), shifting valuation from hardware cyclicality to platform compounding. De-founding isn't a risk; it's the inevitable cost of scale.