TSMC's CEO said demand from AI customers has outpaced what the company can supply, even with new US fabs under construction.
Speaking after a shareholder meeting, CEO C.C. Wei put it plainly: "Customer demand is so high, and we can only support so much." TSMC is mid-buildout on US manufacturing capacity — part of a broader effort to reduce dependence on Taiwanese production — but new supply isn't arriving fast enough to absorb AI-driven orders. Wei added that the company is working hard to ensure it does not "become a bottleneck." The crunch extends beyond logic chips: the memory industry is also buckling under AI workloads, with shortages of RAM and NAND Flash expected to persist for years.
TSMC fabricates the silicon inside nearly every high-end AI accelerator on the market, which means its ceiling is effectively the AI hardware industry's ceiling. A public acknowledgment from the CEO that demand is outrunning supply isn't reassurance — it's a warning that the constraints upstream of every AI product launch are structural, not temporary.
The US CHIPS Act was designed to reduce exactly this kind of concentrated dependency. So far, it has mostly made Arizona a very busy construction site.