Analysis · Semiconductors & Memory · August 17, 2026
Micron’s Gross Margin Is Already 85%. SanDisk’s 2030 Target Is 80%.
Earlier today I wrote that SanDisk’s margin expansion was price, not volume, and that its fiscal 2030 model — roughly 80% gross margin, roughly 75% operating margin — is a bet on a price level holding for four more years. The obvious next question is whether that is a SanDisk story or a memory story. Micron is the second-largest memory position in this book, at 5.38%, and it reports the same industry from the other side, with better disclosure. Its most recent quarterly filing answers the question cleanly. It is a memory story, and the numbers are more extreme.
SanDisk is fiscal 2026 versus fiscal 2025, from its 10-K. Micron is the quarter ended May 28, 2026 versus the same quarter a year earlier, from its 10-Q. Different periods and different fiscal calendars — but the same shape, and the shape is the point.
Quarter ended May 28, 2026 against the same quarter a year earlier. The company states these as ranges — “low-260% range,” “mid-310%,” “low-20% range,” “low-double-digit” — and the bars plot the midpoint of each, labelled with the company’s own wording underneath.
Micron’s consolidated gross margin percentage, as stated in its 10-Q. The dashed line is SanDisk’s roughly 80% non-GAAP gross-margin target for fiscal 2030 — a different company, a different metric basis, and shown only to make the level comparison visible.
In the quarter ended May 28, 2026, Micron reported revenue of $41.46 billion against $9.30 billion a year earlier — a 346% increase. Cost of goods sold rose from $5.79 billion to $6.40 billion, or 10.5%. Gross margin was 84.6% of revenue, which the company rounds to 85%. Operating margin was 80.4%.
Read that against the piece from this morning. SanDisk’s fiscal 2030 targets — the ones that struck me as demanding — are roughly 80% gross and 75% operating margin. Micron is above both of those numbers right now. Which changes the shape of the question considerably: the SanDisk target is not implausible as a level. It is a description of where this industry sits at the top of a cycle. What it asks you to believe is that the top of the cycle lasts until 2030.
1. The same signature, twice
The tell in the SanDisk filing was that revenue and cost of revenue moved in completely different directions of magnitude: revenue up 175%, cost up 12.3%. Micron shows the identical pattern, only wider. Revenue up 346%, cost of goods sold up 10.5%. In both cases the cost of physically making the product barely moved. In both cases the money did.
2. Micron actually shows you the split
SanDisk tells you volume grew “mid-teens percent” and leaves you to infer the rest. Micron does the arithmetic for you, separating average selling price from bit shipments for each product line. The result is the most direct evidence available that this is a pricing event.
DRAM selling prices roughly tripled while the number of bits shipped rose about 20%. NAND selling prices rose more than fourfold on low-double-digit bit growth. Over the nine-month period the pattern is the same, if less extreme: DRAM prices up about 140% on roughly 30% more bits, NAND prices up about 130% on low-20% bit growth.
3. What this does to the SanDisk question
I want to be careful here, because the comparison cuts in SanDisk’s favour as much as against it.
On the one hand, this weakens the objection that SanDisk’s 80% target is fanciful. A direct competitor cleared it this year on GAAP figures. Nobody has to imagine what an 80% gross margin looks like in memory; it is on file.
On the other hand, it sharpens the real objection considerably. Micron did not arrive at 85% by improving anything. It got there because DRAM prices roughly tripled and NAND prices more than quadrupled in twelve months. A margin produced that way is not a property of the company. It is a property of the market clearing price, and it can be given back on the same timetable it was gained — Micron’s own margin was 37.7% four quarters ago, and 74% one quarter ago. That is a series moving fast in one direction, and series that move that fast have historically moved fast in the other direction too.
4. What it means for the book
The uncomfortable finding is not about either company individually. It is that I have been treating these as two positions.
| Position | Weight | Exposure | Latest disclosed driver |
|---|---|---|---|
| SanDisk (SNDK) | 7.94% | NAND pure play | Revenue +175% on mid-teens % volume |
| Micron (MU) | 5.38% | DRAM + NAND + HBM | Revenue +346% on ~20% DRAM bit growth |
| Combined | 13.32% | Both above AEA’s 10% single-position cap when treated as one exposure | |
On paper these are two names in different sub-industries, sized 7.94% and 5.38%, each comfortably inside the 10% single-position limit in the Investment Policy Statement. In practice, the thing that produced the last year of returns at both is one variable: memory average selling prices. Combined, that is 13.32% of the book — above the single-position cap, if the cap were applied to the exposure rather than to the ticker.
This is precisely the failure mode The Concentration Premium was written about, and the one the Risk X-Ray tries to catch: positions that look independent in a table and are not independent in reality. Micron does have a genuine diversifier that SanDisk lacks — high-bandwidth memory for AI accelerators is a distinct demand curve, and NAND is only part of its business. But DRAM and NAND priced in the same direction, at the same time, for the same reason this year. The diversification was smaller than the labels suggest.
What I am not claiming
Not that memory prices are about to fall — I have no ability to forecast that, and the supply-discipline case laid out in The NAND Supercycle is credible. Not that either company has done anything wrong; every number here is their own, disclosed clearly, and Micron’s price-versus-bits breakout is better disclosure than most companies give. And not that I am selling: nothing here is a trade, and per the IPS a concentration observation is a sizing question, not a thesis break. The claim is narrower — that 13.32% of this book is one bet wearing two tickers, and I had not been describing it that way.
What would change my read
WatchThe disclosure that settles it
Micron publishes average selling price and bit shipment changes every quarter, separately, for DRAM and NAND. That single disclosure is the cleanest read on this cycle available from any company I hold — better than SanDisk’s, better than any third-party price index, because it is audited and specific. A quarter in which Micron’s bit shipments rise and average selling prices fall is the turn. It would show up in Micron’s filing before it shows up in SanDisk’s results, and well before it shows up in either share price. Micron’s fiscal fourth quarter is the next opportunity to see it.
Two pieces today have now asked the same question of two different holdings: not whether the reported number is accurate, but which variable it is standing on. In both cases the answer was the same variable. That is the part I did not know this morning.