'You're so rich now': Jensen Huang toasts Taiwan supply chain partners as Vera Rubin ramp begins

NVIDIA’s aggressive AI supercomputer capacity doubling in Taiwan, China reveals not just surging demand but deep entrenchment in CoWoS and advanced packaging ecosystems. This move pressures TSMC and ASE to further invest in 2.5D/3D integration, spurring upstream demand for ABF substrates, high-speed SerDes IP, and liquid cooling. Yet, amid U.S. CHIPS Act restrictions and decoupling from mainland China, such geographic concentration inflates compliance risk and supply chain fragility. Competitors like AMD and Intel will exploit this by pushing chiplet-based heterogeneous integration to reduce reliance on Taiwan, China’s packaging dominance. Over the next 18 months, the AI hardware landscape will shift toward 'decentralized redundancy'—customers diversifying across U.S., Japan, Korea, and Southeast Asia—to mitigate single-point failures, reshaping capex strategies and accelerating modular fab adoption.