Nvidia to expand capacity through next year: CEO - Taipei Times

NVIDIA CEO Jensen Huang announced plans to expand production capacity from the second half of this year through 2025 to meet strong demand. Huang highlighted that supply chain constraints stem from ra... Read original →

Huang’s capacity expansion reveals acute fragility in the AI chip supply chain: 3nm EUV output is overwhelmingly concentrated at TSMC in Taiwan, China, while Grace Blackwell and Vera Rubin platforms intensify reliance on advanced packaging like CoWoS—shifting bottlenecks from transistor scaling to heterogeneous integration. U.S. CHIPS Act delays and export controls have forced NVIDIA to pre-commit 2025 wafer allocations, inflating capex. Intel and Qualcomm will likely accelerate in-house AI accelerators paired with foundry diversification; Intel may leverage IFS to capture mid-tier inference markets. Over the next 18 months, AI agents will trigger a device-level architectural overhaul, elevating EMS players like Hon Hai from assemblers to co-definers of AI hardware. This demand-driven capacity arms race is, in essence, the opening salvo in the geopolitical rebalancing of global semiconductor manufacturing.