In February we published a piece on Micron's HBM4 sellout that made a structural claim, not a price call:
"Memory has always been a brutal cyclical business — boom, bust, repeat. But HBM operates differently. Long-term contracts, locked pricing, and capacity that sells out before it's built. That's closer to TSMC's foundry model than the old DRAM spot market."
Then we leaned on it. Our six-month capex retrospective reaches back for it as established context — "Micron sold out its entire 2026 HBM capacity back in February" — and builds its Apple section on the memory shortage that follows from it. We spent the credibility of that call without ever going back to see whether the filings agreed.
They do, and by a margin that is hard to overstate.
One thing to know before the numbers: our own series does not include this quarter. Our stored Micron data ends at the quarter ended 2026-02-26. Everything below comes from Micron's 10-Q for the quarter ended 2026-05-28, SEC XBRL, CIK 0000723125. If you click through to /stocks/mu you will see a company whose most recent gross margin is 74.4% — that page is not wrong, it is one quarter behind.
The two columns
| quarter ended 2025-05-29 | quarter ended 2026-05-28 | change | |
|---|---|---|---|
| Revenue | $9,301M | $41,456M | 4.46× |
| Cost of goods sold | $5,793M | $6,400M | +10.5% |
| Gross margin | 37.7% | 84.6% | +46.9 pts |
| Diluted EPS | $1.68 | $24.67 | 14.7× |
Revenue rose four and a half times. Cost of goods rose a tenth.
That single pairing is the entire argument. A commodity manufacturer cannot do this: in a spot market, selling 4.5× the volume means buying 4.5× the inputs, and margin barely moves. Micron sold far more product without buying proportionally more to make it, which is what happens when the constraint is capacity that was contracted before it existed and the price was locked at the top of a shortage. 37.7% is a commodity business. 84.6% is not. That is the February claim, tested against an income statement rather than argued.
The sequential step says the same thing over three months rather than twelve: revenue $23,860M → $41,456M, gross margin 74.4% → 84.6%, EPS $12.07 → $24.67.
The metric breaks here, and it breaks upward
Run our own formula on the May quarter and it produces a number we are not going to call a score:
- Revenue growth, year over year: +345.7%
- Free-cash-flow margin: +42.4% — operating cash flow $25,388M less capex $7,826M, against $41,456M of revenue
Those two halves sum to 388. As a Rule of 40 that figure is meaningless, and the reason is structural rather than a data problem: a year-over-year growth term measures where in a cycle you happen to sample, not how efficiently the business is run.
Worth being precise about that, because the tempting version of this sentence is wrong. The year-ago quarter was not the bottom — at 37.7% gross margin it was already well into a recovery. Our own series puts Micron's actual trough two years earlier, at −32.7% for the quarter ended 2023-03-02. So the growth term here is not peak-against-trough; it is peak-against-midpoint, and it would have been even larger measured from the real low. That is the problem with the composite in one line: its value depends on which point of the cycle sits twelve months back.
Our Rule of 40 explainer already carries a section for scores that are category errors rather than verdicts — its own table bottoms out at IREN on −884. Micron at +388 is the same failure with the sign reversed, and in a sector that section never flagged. The low-side cases are lenders and banks, where the metric quietly understates. This is the loud version: a cyclical manufacturer at the top of its cycle, where the growth term does all the work and the composite tells you nothing you could act on. Read the two halves. Ignore the sum.
It also means the number on our homepage is doubly worth ignoring: MU currently renders R40 219 · 2026 Q2, which reconciles exactly to the quarter ended 2026-02-26 (+196.29 growth, +23.12 cash margin). It is correctly labelled with the quarter it came from — and that quarter has already been superseded by a filing showing a much larger business.
What February got right, and what it did not settle
The February piece priced its own caution: "At $410, the stock prices in a lot of good news." We are not going to convert that into a percentage move here. The price stored in our data carries no timestamp, and a claim about how far the stock has travelled needs a quoted price and the moment it was quoted, which this article does not have. The $410 is published text from February and stands as that.
For the same reason there is no P/E in this piece. Our stored ratio ends at 12.8 as of 2025-11-27, which is not a current figure and would be badly misleading beside an EPS line that has gone from $4.60 to $24.67 since that date. What we can state from the filings is trailing-twelve-month diluted EPS through 2026-05-28: $44.17 ($2.83 + $4.60 + $12.07 + $24.67). Anyone wanting a multiple should divide by a price they can date.
The bear case from February is not resolved, and one half of it cannot be resolved by this print. We named two risks:
- Double-ordering. "When capacity is this tight, customers sometimes book more than they need to secure supply." A sold-out year is precisely the condition under which double-ordering is invisible — it looks identical to real demand right up until orders are pulled. Record revenue is not evidence against it.
- Competitive supply. "Samsung and SK Hynix are both ramping their own HBM4 production. By late 2026 or 2027, the supply picture could look very different." Nothing in a Micron revenue line speaks to this either way.
So the honest verdict is that the structural claim is confirmed so far. The pricing regime the February piece described is visibly real in the gross margin. Whether it is durable is a question about supply, and supply is the thing neither quarter has tested.
What to watch
- Whether 84.6% holds in the August quarter. A pricing regime shows up as a plateau; a shortage shows up as a peak. One more quarter near this level is far stronger evidence than this one.
- Cost of goods, not revenue. If the input line starts rising with volume, the foundry comparison is breaking and the spot market is reasserting itself. That row is the tell, and it is the one to read first.
- The first HBM4 volume from Samsung or SK Hynix. That is the event February said would change the picture, and it has not happened yet.
Micron figures for the quarters ended 2026-05-28 and 2025-05-29 are from SEC XBRL, CIK 0000723125, 10-Q for the quarter ended 2026-05-28: revenue and cost of goods tagged directly for each period, diluted EPS tagged directly, and operating cash flow and capital expenditure derived by differencing the nine-month and six-month year-to-date cumulatives ($45,702M − $20,314M and $19,602M − $11,776M). Gross margin and free-cash-flow margin are derived from those figures. The derivation was checked against a quarter our own data does hold: applied to the quarter ended 2026-02-26 it gives a gross margin of 74.41%, which is exactly the value in our stored Micron data. Stored figures quoted here — the 219 score, its component halves, the 12.8 P/E as of 2025-11-27, and the EPS components $2.83, $4.60 and $12.07 — are from that file. The Rule of 40 definition is the one in our explainer.