Operational Excellence (OPEX) Insight – Tuesday - October 13, 2026: Building a Factory for a World That Does Not Exist Yet.

Welcome To Operational Excellence (OPEX) Insight Article For The Paid Subscriber-Only Edition.

The story of a long distance runner

The hardest decision is a decision about something that has no customers yet

Picture a room in which a group of people must sign a decision worth many billions of dollars for a factory that does not exist yet, to serve customers who also do not exist yet. That factory will take years to build. By the time it runs its first batch, the world will have changed, demand will have changed, and no one in the room today knows for certain which way. They cannot build half a factory to test whether there are customers, because a half built factory produces nothing. They also can almost never retreat, because most of the money has already gone into concrete, machinery, and things that cannot be resold to anyone. They must decide firmly, right now, about the size of something usable only years from now. That is the hardest decision in operations, and it has a humble name: capacity planning.

In October 2026, the semiconductor industry association SEMI restated its outlook for equipment spending at large wafer fabs, and the numbers tell exactly the story of that room, at the scale of the whole world. According to SEMI’s outlook, equipment spending for 300mm wafer fabs will rise from about 133 billion USD in 2026, up eighteen percent, to about 172 billion USD by 2029, with double digit growth in both 2026 and 2027. The logic and microprocessor segment alone is forecast to draw about 228 billion USD cumulatively over the three years 2027 to 2029, led by foundry demand for sub 2nm processes, while the memory segment draws about 175 billion. The biggest driver is AI: model training pulls demand for high bandwidth memory, and inference pulls demand for storage in data centers. In short, the whole industry is pouring money into building capacity for an AI hunger that largely still lies in the future.

One thing must be said at once to keep the standard of honesty. This is the forecast of an industry association, that is, an indicator of direction and expectation, not the data of a future that has already happened. Those numbers can be wrong, and the history of the semiconductor industry is full of forecasts that missed. But even if the specific figures slip, the phenomenon behind them is real and worth dissecting: an industry committing, all at once, enormous amounts of capacity, taking years to build, for an uncertain demand. And that is the classic problem of the single operations tool this piece will dig deep into, the tool of capacity planning, together with a central paradox that every business, from a trillion dollar chip maker to a one owner cafe, faces in its own way: you must decide about future capability today, when the future has not yet arrived.