TSMC's leading-edge capacity just got scarcer, which should speed up its pricing power and drive a re-rating. Etched's $1B-plus of customer contracts is real wafer demand hitting a full queue, and its inference chip works on TSMC N4P after first-pass success. Inference scales with usage, not capex cycles, so more AI adoption means more N4P/N3 wafer starts. TSMC's 5-10% yearly advanced-node price rises are about to accelerate, which the Street is not modeling. x.comÂ

