On June 18, Stifel's Brian Chin raised his price target on MicronMU-- from $550 to $1,500. It was not an outlier: across roughly 49 analysts polled by S&P Global, the consensus target now sits near $1,513. The stock has rallied more than 240% this year to the mid-$900s, a $1.1 trillion company, and a chorus of notes now frames the whole debate as one number — can it make $1,500 after the fiscal fourth-quarter report on September 30?
That is the wrong question, and it is worth being wrong about. A price target is an output, not a driver. It is what a model spits out after you feed it assumptions about revenue, margin, and how long they last. The argument about $1,500 is really an argument about one variable dressed up in a different form: how long memory prices can keep going up, and what happens when they stop. Micron's numbers this year were extraordinary. Whether they mark an opportunity or a classic cycle trap depends almost entirely on that one thing — and it is a question September 30 will sharpen but cannot answer.
The revenue quadrupled on price, not on chips
Start with what the business actually did. In the fiscal third quarter, which ended May 28, Micron reported revenue of $41.46 billion — up 346% from $9.30 billion a year earlier — with a non-GAAP gross margin of 84.9%, up from 39% in the same quarter last year, and GAAP net income of $28.24 billion against $1.89 billion.
Quick Backtesting Tool
| FQ3 2025 (year ago) | FQ3 2026 | Change |
|---|---|---|
| Revenue: $9.30B | Revenue: $41.46B | +346% |
| Non-GAAP gross margin: 39% | Non-GAAP gross margin: 84.9% | +45.9 pts |
| GAAP net income: $1.89B | GAAP net income: $28.24B | ~15x |
In a normal memory upcycle, a revenue jump of that size means a surge in the physical number of chips sold, with price rising along with it as a bonus. That is the demand-led boom the industry ran for decades. This is not what the data shows. The dominant move was price. In the first quarter of 2026, TrendForce measured conventional DRAM contract prices up roughly 93% to 98% in a single quarter — a near-doubling of price, not a near-doubling of volume. Micron's gross margin, which went from 39% to 85% in one year, is the signature of a price story: the same wafers, far more money on each.
Why can price run like that? The constraint is physical, and it sits in the product the whole AI buildout depends on. High-bandwidth memory — the stacked, high-speed memory bolted onto AI accelerators such as Nvidia's GPUs — consumes roughly three times the wafer capacity of a standard DRAM chip to make the same die area. As the industry shifts wafers toward HBM to feed AI servers, it is deliberately pulling capacity away from the ordinary memory that goes into phones, laptops, and SSDs. The supply of everything tightens at once. And the three firms that can make HBM at scale have, by their own accounts, very little spare 2026 capacity to offer; Micron says its HBM is sold out for the year.
That is the mechanism, and it is the reason the consensus is so loud. The stock is not up because someone is selling more chips. It is up because the industry is rationing them, and the few firms that can make the most in-demand kind are setting the price.
The contracts are new, and they change the rules
There is one structural difference between this cycle and the last, and it is the part most price-target debates skip. Memory is famous for boom-and-bust: demand surges, everyone expands, supply floods, price collapses, margins vanish. The 2018 boom did exactly that.
This time, at the same report, Micron disclosed 16 long-term supply contracts with customers — take-or-pay agreements, meaning the customer is obliged to buy a committed volume at a floor price, whether the market price later ends up above or below it. The guaranteed baseline across 14 of the 16 is about $100 billion through 2030, backstopped by roughly $22 billion in customer deposits and credit commitments. Management said the floor prices will produce margins "well above our peak quarterly margins in any past cycle." That is a striking claim from a commodity supplier: locking in pricing that beats its own historical best.
The contracts convert a slice of the business from spot-pricing to something closer to a fee-for-service arrangement, and that is genuinely new for memory. It is also bounded. The agreements cover roughly 20% of Micron's DRAM volume and a third of its NAND volume over the period. That is a meaningful floor under the business — and at the current price, a very high one — but it is not the whole building. A large share of revenue remains exposed to whatever the spot market does.
Why 6x is a warning, not a gift
Here is where the beginner and the professional disagree, and where the $1,500 question collapses. At its current price and fiscal 2027 consensus earnings of about $155 a share, Micron trades at roughly 6x expected earnings. The retail read is obvious: a trillion-dollar company at 6x earnings is a bargain, and the fact that Nvidia trades near 27x confirms the discount.
The professional read is the opposite, and it is the one the low number is actually encoding. A multiple this low on a company posting 85% gross margins is not the market underpricing a durable business. It is the market telling you, in one number, how much of that margin it expects to evaporate. The entire earnings base is built on prices that are roughly 90% above where they stood a year ago. When supply catches up, that base does not grow slower — it contracts. Memory has done this every cycle for forty years, and every "this time is different" has ended the same way.
The tell is in Micron's own spending. The company is committing more than $250 billion to new U.S. fabs through 2035 and roughly $27 billion in near-term capital expenditure, betting the orders hold. CEO Sanjay Mehrotra has said the shortage has "no line of sight" to a fix and expects industry supply to improve "gradually in 2028". That is not a forecast of the shortage ending; it is a forecast of the shortage's own supply response arriving. The capex that makes the contracts credible is the same capex that, in a normal cycle, would flood the very market the contracts are protecting. The risk is not that demand disappears; it is that the industry the demand attracts does the one thing memory has always done.
The skeptics are not quiet. Michael Burry has taken a short position arguing an AI-driven oversupply emerges by 2028, and an Intel-backed memory entrant surfaced just before the September 11 selloff that knocked about $57 billion off the stock's value. Add a customer-concentration reality — a single AI chipmaker is the center of gravity for the HBM driving these margins — and you have a business whose peak is unusually high and whose downside is unusually old.
What September 30 will actually test
The fourth-quarter print itself is not the event. Management already guided to roughly $50 billion in revenue at an 86% margin, and consensus is there; a beat is the base case, and the stock has actually averaged a small drop in the week after each of its recent reports because the bar is set so high.
What the report will tell you is whether the price engine is still running into 2027. Three readings carry the thesis: the trajectory of DRAM and HBM average selling prices (the number that does most of the work, and the one that can reverse without a single customer leaving); the fiscal first-quarter 2027 guide (where a company signals whether it believes in the second half of the cycle or the end of it); and any HBM4 volume confirmation (the next generation, where Micron's wafer discipline is supposed to protect its pricing against the Korean rivals).
The question is not whether Micron makes $50 billion this quarter. It almost certainly will. The question is whether the price that makes $50 billion possible can survive the success it caused — the capacity the whole industry is now building toward the 2028 date its own CEO named. A stock at 6x is not cheap because it is safe. It is priced for the cycle to end. September 30 will not decide where it ends, but it will tell you whether the clock is still running, or whether the industry has just started winding it down.
