📌 AI Supply Tightness Spreads to the Manufacturing Chain
Supply constraints for AI chips are extending from advanced packaging to broader segments such as substrates and wafer manufacturing. This means that the orders and pricing pressure brought by AI capital expenditures are no longer concentrated only in a small number of packaging capacities; however, whether each segment can turn tightness into sustained revenue and profit still depends on the implementation of capacity expansion, yields, and whether demand continues.
TSMC is expanding outsourced production of key CoW packaging, directly reflecting that internal capacity in a single segment can no longer fully absorb demand. Outsourcing can increase available capacity, but it also means that orders and part of the manufacturing revenue will be diverted to external partners. At the same time, chip dies are getting larger, driving demand for large-size ABF substrates higher. Customers are not only signing long-term agreements, but are also directly funding the acceleration of capacity expansions originally scheduled to begin in 2029 to 2030, indicating that contracts alone are no longer sufficient to secure supply.
Tightness has also not remained confined to the high-end chain. GlobalWafers said that capacity for 12-inch, 8-inch, and 6-inch silicon wafers is all close to full utilization, with demand now covering wafer supply across different sizes. Taiwanese wafer foundries expect AI-related demand to continue pushing up mature-process prices, with the increase in 2027 potentially exceeding this year’s. If supply remains tight, the prices and revenue of relevant manufacturers may benefit; if new capacity is brought online in a concentrated manner, or AI demand slows, the sustainability of price pass-through will need to be reassessed.
▌ Sources