TSMC: AI chip shortage to persist for years despite capacity expansion - Bits&Chips

Despite ongoing capacity expansion efforts, including new fabs in the United States, TSMC's CEO CC Wei stated that global demand for AI chips will continue to outpace supply for years. Speaking at the... Read original →

TSMC’s capacity expansions won’t alleviate AI chip shortages because the bottleneck lies in high-NA EUV availability and geopolitical fragmentation of manufacturing. Advanced nodes below 3nm hinge on ASML’s delayed high-NA tools, slowing yield ramp and throttling AI accelerator innovation cycles. New fabs in the U.S., Japan, and Europe serve political goals but initially handle only mature nodes—irrelevant for cutting-edge AI chips. Compliance with CHIPS Act localization mandates inflates wafer costs by 15–20%. Samsung may counter with HBM-integrated packaging for edge AI, while Intel pushes its IFS foundry alliance. Over the next 18 months, chronic scarcity will spur 'capacity futures' contracts, forcing hyperscalers to pre-book 2027 output and cementing TSMC’s role as AI infrastructure backbone.