⚠ The Cyclicality CurseHigh threat

The single most important risk in Micron — the one that governs all the others and that no amount of oligopoly or HBM can abolish — is cyclicality. Memory is the most violently cyclical major business in technology, and the numbers that make Micron look so extraordinary today are peak-cycle numbers that history says will, in some measure, revert. An investor who does not internalize this will misjudge everything about the company, because the same figures that seem to justify a trillion-dollar valuation are the product of a cycle at its zenith, and the cycle always turns.

The mechanism is inescapable and ancient. Memory demand surges and stalls with the technology cycle; memory supply is added in lumpy, multi-year, multibillion-dollar increments that reliably arrive at the wrong time. When demand outruns supply, prices soar, margins explode, and profits gush — as now, with gross margins near eighty-five percent and net income around fifty billion dollars¹. But those very profits draw the investment that creates the next oversupply, and when supply catches up or demand cools, prices collapse toward cash cost, margins crater, and profits evaporate — sometimes into losses. This is not a tail risk or a possibility; it is the fundamental, repeating rhythm of the business, and Micron has lived it over and over. As recently as fiscal 2023 — just two years before the current euphoria — Micron lost nearly six billion dollars in a memory down-cycle². The swing from a six-billion-dollar loss to fifty billion in profit in the space of a few years is the truest single portrait of what memory is.

The current peak is the highest in the industry's history, which cuts two ways. The bull case is that HBM and the structural, durable demand of AI have permanently raised memory's floor, so the next trough will be far shallower than past ones — that the industry has entered a new, higher, less-cyclical era. That case is serious and may prove partly right. The bear case is that a peak is a peak: the higher margins climb above their sustainable level, the further they have to fall; all three makers are building the capacity that ends booms; and the AI demand driving it all is an untested investment cycle. What is certain is that eighty-five-percent margins are not a baseline, that the capital being invested now will become tomorrow's supply, and that memory has never — not once — sustained a peak indefinitely. The oligopoly has made the busts shallower and the booms more profitable than in the fragmented past, a genuine improvement. But it has not repealed the cycle, and the cyclicality curse remains the master risk of the entire Micron story: these are the best of times, and the best of times, in memory, are precisely the times to remember the worst.