TSMC has warned customers to expect annual price increases on advanced nodes from 2026 to 2029.

Initial guidance for 2026 points to hikes of roughly 3% to 10% depending on node maturity and volume.

Four consecutive years of increases signals a structural view inside TSMC that demand will stay ahead of effective supply even as new fabs come online. Pricing becomes both a capacity allocation mechanism and a margin defense, not a short term squeeze.

Advanced nodes like 3 nm and 2 nm are expensive because cost scales non linearly. EUV tool time, mask counts, design complexity, and yield learning curves all rise sharply at each shrink. Once a foundry is selling every qualified wafer start, pricing power shifts decisively to the supplier.

In theory, customers can dual source with Samsung’s 2 nm gate all around process. In practice, switching foundries takes 12 to 24 months and brings yield risk, schedule risk, and redesign cost. That limits real price elasticity.

The pressure concentrates on AI accelerators and premium mobile SoCs. These designs monetize leading edge performance per watt. Older nodes are not substitutes, which makes them the least price sensitive customers and the first to absorb sustained increases.