The AI chip rush is crowding the same narrow pipeline - Axios

The rapid advancement of artificial intelligence technology has created unprecedented capacity pressure in the semiconductor industry, as demand for AI chips surges. Major chip manufacturers like TSMC... Read original →

Surging AI chip demand is pushing 3nm and sub-3nm nodes to their physical and capacity limits. TSMC, the sole foundry capable of high-volume 3nm production with High-NA EUV, has its capacity pre-allocated by NVIDIA and peers, forcing smaller fabless firms into 18+ month wafer queue delays—accelerating adoption of RISC-V and chiplet-based designs. Geopolitical friction and U.S. CHIPS Act export controls compel multinationals to build redundant fabs across the U.S., Japan, and Europe, inflating capex by 15–20%. In the near term, NVIDIA will likely lock in TSMC’s CoWoS advanced packaging to secure supply, while AMD and Intel pivot to 2.5D integration to bypass wafer shortages. Over the next 12–24 months, advanced packaging—not just transistor scaling—will become the primary performance lever, while delayed EUV ecosystem development outside Taiwan, China, will exacerbate supply chain fragility.