Jensen Huang lands in Taiwan, calls Vera Rubin biggest product ramp in computer history

Huang’s declaration of 'Vera Rubin' as the largest product ramp in computing history is, in essence, a full-throttle bet on TSMC’s CoWoS advanced packaging capacity. This move will trigger acute shortages across ABF substrates, HBM3e memory, and liquid cooling modules, forcing AMD and Intel to fast-track MI300X and Gaudi3 alternatives. Geopolitically, with U.S. export controls tightening, Taiwan—home to the world’s only high-volume CoWoS line—has become a single point of failure for global AI infrastructure. Over the next 12 months, NVIDIA will likely lock in billions in prepayments to secure capacity, sidelining smaller players. Even with SMIC’s N+4 process, Chinese firms remain bottlenecked by HBM and packaging constraints. The long tail: a TSMC-NVIDIA duopoly will dominate AI hardware, and non-U.S. cloud providers failing to adopt chiplet-based heterogeneous integration by 2027 will forfeit compute sovereignty.