Micron's sold-out memory meets a peak-cycle warning before earnings

Micron closes its fiscal year on Wednesday, and the report lands on a stock that has done what memory stocks are not supposed to do. It has compounded. The company now carries more than $1 trillion in market value, a level the market would not have granted a memory maker a year ago. The tape says one thing. Monday said something else, with shares down about 3.6% as the broad market slipped. Underneath the price sits a genuine argument over what Micron has become. Has memory turned into a contracted, multi-year business with the kind of forward visibility chip investors rarely get? Or is it the same cyclical commodity it has always been, now priced for a top that history says shows up on schedule?

Five-year contracts have replaced the spot market

The optimistic read starts with the customer list. On its June-quarter call, Micron said it holds 16 strategic agreements worth roughly $100 billion in minimum contracted revenue, most of them five-year take-or-pay deals covering about a fifth of its DRAM volume and a third of its NAND. Chief Executive Sanjay Mehrotra told investors that high-bandwidth memory, the premium chips stacked beside AI accelerators, is sold out through 2027. Fiscal third-quarter revenue hit a record $41.5 billion, and the company guided the fourth quarter to $50 billion with earnings of $30.73 a share. Pricing backs the story. The research firm TrendForce tracked conventional DRAM contract prices climbing 93% to 98% in the first calendar quarter and another 58% to 63% in the second, and it sees supply staying tight into 2028. For the bulls, that is years of booked demand, a visibility a chip cycle rarely offers.

SK Hynix just committed $38 billion to more fabs

The other read starts with the same demand and arrives at the opposite worry. High prices are an invitation. In August, SK Hynix approved about $38 billion in new plants, a DRAM fab in Yongin and a NAND site in Cheongju, and Chinese makers ChangXin and Yangtze are adding lines of their own. Every one of those wafers eventually competes for the same buyers. Citi analyst Atif Malik made the timing explicit this month. He cut his Micron target to $1,150 from $1,400 while keeping a buy rating, and mapped a path where quarter-over-quarter DRAM gains fade from 23% to 9% to 2% to zero over the coming year, peaking in the spring 2027 quarter before slipping in the back half. His reason was capacity. The skeptic’s point is not that the demand is fake. It is that memory has run this play before, and the boom has always ended when the new fabs come online. There is a wrinkle inside the current run, too. TrendForce has noted that HBM contract prices themselves eased during 2026 even as shipment volumes climbed, a hint that the premium tier could soften before the commodity one does.

What the data shows

Micron carries a 52-week high of $1,255, and shares changed hands just above $1,080 late last week before Monday’s pullback. The company’s market value has pushed past $1 trillion, a size almost unheard of for a business the market treated as deeply cyclical only a year ago. Monday’s decline came on a broadly lower session, so part of the move belonged to the market and not to the memory story. That size is itself part of the argument. It cuts both ways. A trillion-dollar valuation is pricing in durability, and a cyclical multiple would not; which one is correct depends on the next question.

What would settle it

Wednesday’s report is the short-term test. The line that matters most is the fiscal 2027 first-quarter guide, alongside any update on how far HBM4 has cleared customer qualification and what management says about pricing into next year. Past that, the calendar is specific. Citi’s thesis rests on the November, February, and May DRAM prints; if quarter-over-quarter gains hold up on that schedule, the peak call weakens, and if they roll over on cue, it strengthens. The capacity has dates too. SK Hynix has said its new Korean fabs are built for 2029 and beyond, which is when the current supply math actually changes. Until those milestones arrive, both the contract story and the cycle story fit the same set of facts.

The print settles the quarter just closed. The structural question waits on the pricing releases and the fab schedules, and those run well past this week.

This is a general-market summary for information only — not investment advice, and not a recommendation regarding any security.