⚠ HBM Is a Differentiator, Not an Escape from CyclicalityHigh threat

The most important thing to understand about the HBM franchise — the source of all the excitement — is what it is not: it is a genuine differentiator that has improved memory's economics, but it is not an escape from the cyclicality that defines the business. HBM is, at bottom, a form of DRAM, made by the same three companies, subject to the same fundamental dynamic in which supply, added in lumpy multi-year increments, chases demand that surges and stalls. It is more differentiated than commodity memory — harder to make, contract-priced, scarcer — and that differentiation is real and valuable and may durably raise the floor. But it does not repeal the cycle; it merely, for now, sits at the cycle's most favorable point.

All three memory makers see HBM's extraordinary margins and are racing to expand capacity, which is precisely the mechanism by which memory booms end: the profit that scarcity creates funds the capacity that destroys the scarcity. Meanwhile the demand side rests on the AI-infrastructure buildout — an investment cycle of historic scale but untested durability, concentrated in a few buyers. If HBM capacity outruns AI demand, or if the AI-investment cycle cools, HBM's premium economics would compress, and while HBM's genuine manufacturing difficulty might keep its margins above commodity memory's, they would not stay at eighty-five percent. The bull case — that HBM and AI have structurally transformed memory into a higher-floor, less-cyclical business — is plausible and may prove partly right. But the prudent reading is that HBM has made Micron a better cyclical business, not a non-cyclical one, and that the current numbers capture a differentiated product at a cyclical peak — the best of both worlds — an 84.6% gross margin on memory¹ — which is exactly the combination least likely to persist.