A Quiet Rewrite of the World's Semiconductor Map

The fabs rise slowly, which is why the story is easy to miss. A leading-edge semiconductor plant takes four to five years from ground-breaking to volume production, and so the decisions made in the subsidy race of the early 2020s are only now hardening into concrete, ultrapure water systems, and the most precise machines human beings have ever built.

Walk the construction sites — in the Arizona desert, outside Dresden, in Kumamoto prefecture, on the outskirts of Ahmedabad — and the same pattern repeats: a core of veteran engineers rotated in from the old heartlands, a ring of local suppliers learning tolerances they did not know existed, and a training pipeline that reaches down into regional universities rewriting their curricula in real time.

The economics remain brutal and the concentration at the very leading edge remains extreme; nothing under construction today displaces the Taiwanese and Korean champions from the frontier. What is changing is the shape of the second tier. Mature-node capacity — the unglamorous chips in cars, turbines, and medical devices — is dispersing across a dozen new sites, and with it the bargaining position of everyone who depends on them.

Industrial policy's critics predicted subsidy tourism: plants that exist on paper to harvest incentives. Some of that happened. But the surprise of the programme era is how much genuinely moved — not the frontier, but the foundation. Supply chains that were single points of failure in 2021 are becoming redundant in the engineering sense: capable of absorbing a shock without transmitting it.

Maps of chip capacity are usually drawn as national scorecards. The more accurate picture is a lattice — one that now has more nodes, shorter edges, and, for the first time in thirty years, meaningful slack. Whether that slack survives the next downturn's consolidation is the question the whole experiment now turns on.