Foundry capacity

Foundry capacity is the physical factory space available to actually manufacture advanced chips — the front-end source of every piece of leading-edge silicon before it is ever packaged, fitted with memory, or turned into an accelerator. Nothing downstream in the AI stack exists until a wafer has been fabricated, which makes capacity the true origin of supply. It matters now because leading-edge demand keeps climbing while the ability to make these chips sits with only a handful of fabs, and because where those fabs are built has become a strategic and geopolitical question rather than a purely industrial one.

The model treats concentration itself as the risk: when so much advanced output depends on so few sites, any capacity crunch or geopolitical shock ripples through the entire chip supply at once. That leverage extends to the narrow set of firms the whole industry leans on — the fabs like TSMC and SMIC, the lithography and equipment makers such as ASML and Tokyo Electron, and materials suppliers like Shin-Etsu and ASM International. Each sits at a pinch point where scarcity, not competition, sets the terms.