Interview: Nvidia CEO Jensen Huang About Accelerated Computing

• Source: Stratechery (Ben Thompson)
• Date: 2026-03-17
• Type: interview
• RDCO Relevance: Medium-High (AI compute architecture, inference optimization)

Interview with Jensen Huang following his GTC 2026 keynote. Key topics: Nvidia’s strategic pivot from single-GPU architecture to three distinct architectures (Vera Rubin GPUs, Groq LPUs, Vera CPUs), navigating both China and Washington, and Nvidia’s self-concept as a computing platform company. Huang discussed how Groq’s LPU architecture (acquired by Nvidia) handles specific inference workloads where deterministic, low-latency token generation matters, while Vera Rubin handles prefill and KV cache management. The interview explored Nvidia’s paranoid defense of its market position against ASICs and custom silicon from hyperscalers.