Micron: The Market Is Pricing a Memory Bust. The Contracts Say Otherwise

Micron: The Market Is Pricing a Memory Bust. The Contracts Say Otherwise

Micron just reported the best quarter in its history, and the stock is lower than it was. Revenue for fiscal Q3 came in at $41.46 billion, up 346% from a year earlier. Earnings were $25.11 a share, up more than 1,200%. Gross margin hit 84.9%, a company record, versus 39% a year ago. Management then guided the quarter it is now finishing to about $50 billion in revenue and near $31 a share in earnings, so the numbers keep climbing. And yet the stock, after touching a record around $1,255 in late June, trades near $967 today, about 23% off the high. A price that falls while a business prints records is not a contradiction. It is the market placing its bet on the most famous pattern in technology.

That pattern is the memory cycle. DRAM and NAND — the chips that hold data in phones, servers, and now AI accelerators — swing between shortage and glut because new plant takes years to build, and in a shortage every producer builds at once. The last bust, two years ago, cut Micron's revenue roughly in half and pushed it to losses. Investors who lived through that price memory companies on the bust they expect after the boom, not on the boom itself. So the honest reading of the pullback is not that something broke — every disclosed number, including next quarter's guidance, still points up. The pullback is investors asking the only question that has ever mattered here: how long can this last?

The market shows its answer in the multiples. Micron costs about 21 times its trailing year's earnings — the past four reported quarters sum to roughly $45 a share — which makes it look cheap beside Nvidia above 30 times or Broadcom near 60. But the forward P/E is quoted near 150 times, because the earnings models behind it expect next year's profit near six dollars a share: about a quarter of what the company just earned in a single quarter. That gap — 21 times trailing, roughly 150 times forward — is the entire debate in two numbers. The current quarter's guidance annualizes to about $124 a share, which would put Micron at about eight times the earnings run rate management just guided. Pay the trailing or run-rate multiple and you are buying strength. Pay the forward multiple and you are buying the bust. You cannot hold both views; the multiple you choose is the thesis you hold.

The bear case deserves respect, because it is the historical default. Margins near 85% are scarcity prices, and scarcity prices summon supply: Samsung, SK Hynix, and Micron are all pouring billions into new fabs, and Micron's own CEO says industry supply starts to improve "gradually" in 2028 — a date on the calendar for when the super-profits normalize. Chinese rival YMTC has, by one count cited by CNBC, passed Micron in NAND shipment share. The macro backdrop helps the bears too: Micron is up about 240% this year and roughly 700% over twelve months while the broad market flashes the same bubble warnings that have preceded semiconductor selloffs before. Buy a memory stock at peak margins and you have historically bought the top, full stop.

But this cycle has produced something no previous one did, and it is a contract, not a narrative. Micron has signed 16 take-or-pay agreements that run through calendar 2030, lock in cumulative minimum revenue of roughly $100 billion, and carry $22 billion of customer cash deposits on its books. Take-or-pay means the customer must buy the committed volume whether it needs it or not. On top of that, high-bandwidth memory — the specialized memory stacked into AI accelerators — is fully booked through calendar 2027, with demand reaching into 2028. Data center memory alone just cleared $25 billion in a single quarter, an annualized run rate above $100 billion. And management says about 40% of total revenue now carries fixed prices or price ceilings. A disclosed floor of roughly $100 billion of contracted revenue through 2030 is a different object from "memory always turns." It does not prove the bust is gone, but it raises the bar for the bears: to be right they now have to show these contracts being renegotiated or DRAM pricing cracking, not just point at a calendar.

Notice what Micron gave up to get that floor. Fixed prices and price ceilings on about 40% of revenue mean some upside is capped if memory prices spike again. That is the tell that management is trading the peak for stability — flattening the cycle on purpose rather than riding it to the top — and it is why this is not an Nvidia-like story to buy or sell on momentum. The design is a steadier compounder. For an investor who wants durability, that is the point. It is also a reminder that even the strongest structural story here caps how much of the boom you own.

The market's positioning agrees the ramp is real. AInvest's aggregate signal still labels the stock a Buy, and sell-side targets sit well above the price — New Street's Pierre Ferragu upgraded to Buy with a $1,250 target in mid-August, and several research houses carried $1,500-to-$2,000 targets after the June report. Treat that as what it is: positioning that bets the ramp continues, not evidence that it does. The balance sheet funds the build-out regardless — $26 billion of trailing free cash flow and roughly $25 billion of cash on hand.

I am not going to claim $967 is the low or that the shakeout is finished. This stock's average daily move is near $65, it has already cut more than 30% off its high into late July before snapping back to chop around its 50-day average, and the tape is in a regime where anything with "semiconductor" attached gets sold and rebought inside a week. What I can give you is the test that separates a genuine shakeout from a top, and it lands in a few weeks when Micron reports the quarter it is finishing now. Management guided revenue of about $50 billion. If the outlook for the quarter after that keeps climbing — no guide implying revenue down sequentially, gross margin holding near the 86% it just forecast, HBM bookings extending beyond 2027 — the market is still under-modeling the ramp and the selloff was a shakeout by the numbers. The specific invalidation is just as clear: guidance implying sequential revenue decline, margins rolling over, HBM bookings slipping, or those take-or-pay contracts bending. Run that list when the report lands. Do not run it on the argument that memory is always cyclical — the disclosed contracts were built to make that argument weaker this time.

"Don't get shaken out" is advice; this is the reasoning underneath it. The market is pricing a memory bust at roughly 150 times forward earnings while a company guiding to eight times its current run rate sits on $100 billion of take-or-pay revenue and a sold-out HBM line through 2027. If you believe those contracts survive the calendar — if you would still want the memory business at 60% gross margins instead of 85% — then a 23% pullback in the strongest quarter the company has ever reported is the kind of shakeout a thesis holder is meant to ride through, with the September guide as the checkpoint. If you would not own memory at normal margins, no contract language changes that answer. The market has told you which bet it is placing. The only question is whether you know which one you are placing.