Semiconductors get told as a technology story. Their economics are a capacity story, and capacity runs on a clock the technology narrative rarely mentions.
The technology is genuinely remarkable, and it is what draws attention to the sector. The economics, though, are governed less by what the chips can do than by what it takes to make them. Capital is the governing variable, and once that is understood, most of the industry's characteristic behaviour stops looking mysterious, including the violence of its cycles.
Start with what capacity costs. A modern fabrication plant consumes an enormous sum and several years before it produces anything at all. That gap sits between the decision to expand and the arrival of the expansion, which means the decision has to rest on demand expected years out. Expectations get formed under whatever conditions happen to prevail at the moment of deciding. So the industry commits to building when demand is strong, and the capacity shows up after the conditions that justified it have moved on.
That lag drives the cycle. Strong demand lifts prices and profits, which justifies new capacity, which arrives some years later into a market that has frequently cooled. The added supply pushes prices down, which discourages further investment, which eventually produces scarcity once demand recovers. Then it runs again. This is not incompetence. The lag is intrinsic to the physics and the construction schedule, and the cycle follows from it.
The cost structure makes the swings worse. A fabrication plant carries very high fixed costs and comparatively little cost for each additional unit, so profitability depends heavily on how full it runs. At the limit it is highly profitable. Well below the limit it can lose money badly, because the fixed costs arrive regardless of output. Modest movements in demand therefore produce dramatic movements in profit, and sector earnings swing far more violently than sector revenues do.
Nor is the industry a single kind of business. Some firms design chips and manufacture nothing. Some manufacture chips designed elsewhere. Some do both. A smaller group builds the equipment that the manufacturing depends on. A designer carries far lighter capital requirements and a different set of risks than a manufacturer does, and reading the sector as one homogeneous block hides distinctions that matter a great deal.
Concentration is the other defining feature. The technical and capital demands at the leading edge have reduced the number of firms able to operate there to a very small group, and that cuts in two directions. Those who remain enjoy considerable advantages, since the barriers to joining them are now formidable. It also leaves the entire electronics economy dependent on a handful of facilities in a handful of places, a vulnerability that has attracted a great deal of political attention.
For anyone trying to understand this sector, then, the capital cycle is the more useful frame. How much capacity is being added, when it will actually arrive, how fully the existing plants are running, where the industry currently sits in its investment cycle: these are questions about supply, and they explain the sector's behaviour more reliably than any account of what the technology might one day enable. What they do not do is establish what any company's shares are worth. They describe the terrain on which such a question would have to be asked, and that terrain is usually missing from a discussion conducted entirely in the language of technological possibility.