Nvidia’s Jensen Huang maps a full-stack AI future: GPUs, CPUs, and Groq

Stratechery’s interview with Nvidia CEO Jensen Huang outlines Nvidia’s “accelerated computing” full-stack vision: CUDA, software ecosystems, and AI factories are crucial, with CPUs remaining essential for tool-use and single-thread performance. Huang defends the Groq acquisition for disaggregated inference, discusses Dynamo and open-source models, and weighs China policy, supply chains, and energy constraints as Nvidia seeks to lead across five AI layers rather than rely on a single stack.