What we know after Jensen Huang resigns as Nvidia’s CEO - AP News

Huang's departure fractures the "founder moat" that anchored CUDA's ecosystem for two decades. The successor's core question isn't architectural—it's directional: sustain general-purpose GPU lock-in, or pivot toward the custom-silicon model Broadcom executes for hyperscalers? The technical transmission is immediate. CUDA's iteration cadence—PyTorch integration, TensorRT optimization, NVLink topology—was inseparable from Huang's personal engineering judgment. A 12-month strategic ambiguity propagates directly into HBM procurement (SK Hynix, Samsung) and CoWoS-L packaging allocation at TSMC (Taiwan, China), compressing order-book visibility across the advanced-node supply chain. Competitively, AMD's MI350 and Intel's Gaudi 3 are positioned to exploit any window. The 1990s TI precedent is instructive: after losing founder-led direction, its analog share eroded 12 points within three years. If Nvidia's developer-community velocity dips, hyperscaler in-silicon budgets (Microsoft Maia, Meta MTIA) accelerate. The 12–24-month tell: CUDA community commit frequency, Blackwell Ultra yield curves, and whether Nvidia initiates M&A targeting Groq or SambaNova. Deterioration on all three signals the structural end of the "one superpower" AI compute era.