The head of TSMC, the world’s largest contract chipmaker, has warned that global chip supply will not keep up with AI‑driven demand for several years, even as new factories come online. Chief executive C.C. Wei told shareholders in Hsinchu on 4 June that advanced‑node capacity remains tight and that TSMC will “struggle to meet demand led by American customers” despite expansion efforts in the US and elsewhere.
Wei’s comments put numbers behind what AI buyers already feel in pricing. Industry summaries of his remarks suggest that demand for the most advanced process nodes is running 25–30% ahead of supply this year, and that the imbalance is unlikely to ease before 2027. TSMC’s May revenue underlines the boom: the company reported consolidated sales of NT$416.98bn (about US$13.9bn) for the month, up 30.1% year‑on‑year on the back of high‑end AI and data‑centre orders.
For Scottish and UK readers trying to make sense of AI compute pricing, this is the upstream bottleneck. Sovereign‑compute plans, AI Growth Zones and data‑centre projects still sit on top of a global foundry market where a handful of firms like TSMC, Samsung and Intel control advanced production. Until capacity catches up, the cost and availability of GPUs and AI accelerators – and the timelines for rolling out new “AI factories” in places like Korea by 2027 – will be set more by wafer supply than by government funding announcements.
