⚠ Capital Intensity Meets a Cyclical IndustryHigh threat

The technology-and-capital barrier that protects Micron carries a matching vulnerability that is the source of memory's notorious pain: Micron must sink vast, fixed, up-front capital into a business whose demand is violently cyclical, and getting the timing wrong — which the industry reliably does — is enormously costly. Fabs take years to build and cost over ten billion dollars each¹, so Micron must commit capacity long before it knows whether the demand will be there. Memory's entire history is of makers investing into booms that turn to gluts, leaving expensive new capacity idle exactly when prices collapse — the mechanism that produces the industry's brutal down-cycles and, in the worst of them, its multibillion-dollar losses.

The current moment is the classic setup. Flush with the record profits of the AI-memory boom, all three makers are pouring capital into new capacity on the assumption that HBM and AI demand persist. If that assumption holds, the investment will be richly rewarded; if it does not — if the AI cycle digests or the makers collectively overbuild — Micron will have committed tens of billions into a fading boom, and the fixed costs of that capacity will crush margins as the cycle turns. This is not a risk Micron can engineer away, because the capital intensity that keeps competitors out is the same capital intensity that chains it to the cycle: it must keep spending to stay competitive, and the spending is irreversible once committed. The barrier and the burden are the same wall of capital, and the wall is at its most dangerous precisely now, at the top of the cycle, when the profits are highest and the temptation to overbuild greatest — the eternal memory trap, which the oligopoly has softened but never abolished.