Micron Fell 13% While Memory Prices Soared: The Selloff Is a Bet on 2028, Not on the Shortage

Here is the puzzle at the heart of Micron's recent tape. The company makes the one input the AI build-out cannot get enough of, and the price of that input keeps climbing — Dell has been warning its salespeople about memory-driven PC price increases, and prices for the DRAM that feeds AI data centers have been rising more than 50% in a single quarter. Yet Micron shares are down about 13% over the past three months. A product's price surging while the maker's stock falls is not a contradiction. It is usually a sign that the stock market and the spot price have stopped talking about the same thing.

The memory surge itself is not in dispute. AI data centers consume high-bandwidth memory (HBM) and DRAM faster than the three incumbents — Micron, Samsung and SK Hynix — can supply it, and the shortage has spilled into consumer PCs: Dell has announced 10–30% price increases on commercial lines citing memory costs, and AI memory capacity is effectively sold out. Against that backdrop, Micron reported a blowout fiscal third quarter on June 26 and guided the next quarter to about $50 billion in revenue. The "looming" price surge of the headline is not looming; it is here, in the reported numbers.

So if the measure everyone watches — the price of a gigabyte of memory — keeps rising, why did the shares fall about 13%? Because the selloff was never really about the price per gigabyte. It was about the price investors are willing to pay for the earnings that per-gigabyte price produces.

Walk the sequence. A year ago Micron traded near $114; by late June it had run to a $1,255 peak and crossed a $1 trillion market capitalization in May, roughly a tenfold move in under a year. When the company then printed its blowout June quarter, the stock fell about 7% the next day — classic profit-taking after a run of triple-digit percentage returns — and a broad July retreat in semiconductor and memory names dragged the complex lower. None of those forces questioned whether memory prices were real. They repriced what investors would pay for the stock, not what a chip sells for.

That reframes the actual disagreement. The bulls and bears no longer dispute whether today's prices are genuine. They dispute whether those prices can last. On the bull side, Micron has signed 16 "strategic customer agreements" that lock in floor-priced memory through about 2030, covering roughly 40% of revenue at gross margins the company says are "well above" any prior cycle's peak. That slice of the business is contracted scarcity — for it, memory has genuinely stopped behaving like the boom-and-bust commodity it was for three decades.

The other roughly 60% is not contracted, and that is where the bear case lives. Memory's history is that every shortage seeds the overcapacity that ends it, because the rational response to scarcity is massive capacity spending. Micron is already spending about $25 billion a year on capacity, and China's CXMT — now the world's fourth-largest DRAM maker with roughly 8% global share — raised $8.6 billion in Asia's largest IPO of 2026 to expand, and is selling DRAM near half of world prices. If that capacity ramps as advertised, the new marginal unit of supply arrives on the order of 2028, and spot prices eventually turn down — hitting the roughly 60% of Micron's revenue that sits outside contract floors.

That is the third possibility the A-versus-B framing misses. The debate is not "memory has structurally changed forever" against "memory crashes next quarter." It is a question of timing and coverage: how much of Micron's profit is shielded by price floors, and when does the next marginal unit of supply break spot. The 13% decline is the market's partial answer — it has decided those record earnings are closer to a peak than to a new equilibrium, and it has compressed the multiple accordingly, even while the contract book keeps the reported numbers real.

For a reader deciding what to make of the pullback, the useful question shifts from "are memory prices collapsing?" — there is no evidence of that; they are rising — to "at what price will the spot market clear, and how much of Micron rides on it?" The forward observables are spot and contract DRAM prices rolling over, CXMT's capacity ramp, and Micron disclosing how far beyond 40% the floor-price coverage extends. The day spot prices turn, Micron's thesis changes; until then, a three-month decline says more about the multiple the market is willing to place on cyclical-peak earnings than about the health of the shortage that got the stock here.