Analysis · Semiconductors & Memory · August 17, 2026
SanDisk Sold Mid-Teens Percent More Memory and Made $12.3 Billion More Gross Profit
SanDisk filed its fiscal 2026 annual report on Monday. Revenue rose 175%, to $20.25 billion. Gross margin went from 30.1% to 71.5%. Operating margin went from minus 18.7% to plus 61.3% — a swing of 80 percentage points in a single fiscal year. Four days earlier, at its investor day, management laid out a long-term model calling for roughly 80% gross margins and roughly 75% operating margins sustained through fiscal 2030. SanDisk is the largest position in this book, at 7.94%. So the question I actually have to answer is not whether the year was extraordinary. It is what produced it, and whether that thing recurs.
The 10-K answers the first half plainly. Buried in the segment discussion is the sentence that reframes the entire result: “Total products sold increased by mid-teens percent on an exabyte basis.” Revenue nearly tripled. Volume grew mid-teens percent. Cost of revenue rose 12.3%, from $5.14 billion to $5.78 billion — barely moving while revenue added $12.9 billion.
Which means almost none of this was operating leverage in the ordinary sense, and almost all of it was price.
FY2024–FY2026 are GAAP figures from SanDisk’s own income statement. The FY2030 bars are management’s stated long-term model and are non-GAAP — not directly comparable to the three years to their left, and shown outlined to mark that difference.
1. The scissors
Put the three growth rates next to each other and the mechanism is unmistakable. Revenue grew 175.3%. Volume grew mid-teens percent. Cost of revenue grew 12.3%. A business that sells roughly 15% more of something, spends roughly 12% more making it, and collects 175% more money for it has not become more efficient. It has been repriced.
Revenue and cost of revenue are exact, from the income statement. Volume is the company’s stated “mid-teens percent” on an exabyte basis, plotted at 15% as the midpoint of that range and labelled accordingly.
The company does not dispute any of this. Its own explanation of the gross-profit line reads: gross profit “increased $12,260 million in 2026 compared to 2025, primarily due to higher sales and higher pricing,” and gross margin “increased 4,100 basis points… primarily due to higher sales and higher pricing.” Pricing is named in both sentences.
2. Where the price came from
The segment detail is more revealing than the consolidated number, because it shows the pricing move was not confined to the AI-adjacent part of the business.
In Datacenter, volume and price both roughly doubled. In Edge, volume barely moved and price did almost all of the work — a mix that looks far more like a shortage than like demand-led growth.
| Segment | Revenue change | Volume (exabytes) | Revenue per gigabyte |
|---|---|---|---|
| Datacenter | +437% (+$4,193m) | almost +120% | almost +150% |
| Edge | +195% (+$8,033m) | high single digits | almost +180% |
| Consumer | +29% (+$667m) | lower (not quantified) | higher (not quantified) |
Edge is the tell. Revenue there rose 195% — more dollars than Datacenter added — on high-single-digit volume growth and a roughly 180% increase in revenue per gigabyte. Consumer sold fewer units and still grew revenue 29% on price alone. Those are not the signatures of a demand boom in those segments. They are the signature of capacity being pulled toward the highest bidder, which is exactly the reallocation described in The NAND Supercycle.
3. What the fiscal 2030 model requires
Against that backdrop, the long-term model is a strong claim. As reported from the August 13 investor day, it calls for mid-to-high-teens revenue growth in line with bit growth, non-GAAP gross margin around 80%, non-GAAP operating margin around 75%, operating expenses around 5% of revenue, and adjusted free-cash-flow margin around 50%, with 100% of excess free cash flow returned to shareholders.
Three observations, stated as arithmetic rather than opinion.
It requires peak pricing to become permanent
FY2026’s 71.5% gross margin was produced by revenue per gigabyte rising 150–180% on mid-teens volume growth. Holding roughly 80% gross margin through FY2030 while volume grows mid-to-high teens means today’s prices are not a peak to be given back but a new floor to build on.
The 5% opex assumption is below FY2026’s R&D line alone
R&D was 6.6% of revenue in FY2026 and SG&A a further 3.3%, for 10.2% of total operating expenses. A ~5% opex ratio requires revenue to keep growing materially faster than spending, indefinitely — it is a statement about the denominator as much as the numerator.
Part of the gap is definitional, not operational
The targets are non-GAAP; FY2026’s 61.3% operating margin is GAAP. Some of the distance between 61.3% and ~75% is stock-based compensation and similar add-backs rather than business improvement. The two numbers should not be read as a straight-line trajectory.
4. The case that it holds
I hold this stock, it is my largest position, and it has returned 151% against my cost. That is a reason for more scrutiny, not less — but it also obliges me to put the other side properly.
The bull case is not that memory stopped being cyclical. It is that this cycle is supply-constrained in a way prior ones were not: fab capacity takes years, the industry is far more consolidated than in previous downturns, and AI-driven enterprise storage is a genuinely new demand curve rather than a reallocation of an old one. If that holds, elevated prices persist longer than the historical cycle would suggest, and capacity discipline — not demand — becomes the thing to watch. SanDisk also now has a pure-play NAND balance sheet, having earned $11.4 billion of net income in one year against a business that lost money in the two prior years, which buys a great deal of room.
It is worth being precise about what the company itself flags. Among its named risk factors is: “Competitive conditions, including declining average selling prices, volatile demand, technological change, industry consolidation, lengthy product qualifications, and supply constraints, in our industry can negatively impact our business.” Declining average selling prices is listed first. The company is not claiming the cycle has been repealed.
5. What this means for the book
SanDisk is 7.94% of the portfolio, against a 10% single-position cap, at an average cost of $483.21. Its status in the Coverage Book is already Under Review, and nothing here changes that to a sell — but it sharpens what the review is about.
The thing I own is not a company that discovered an 80% gross margin. It is a company holding a commodity at an extraordinary price, with an unusually good balance sheet and a genuinely improved competitive position. Those are different assets with different risks, and the second one is worth less than the first if pricing normalises. Read alongside Micron at 5.38%, the book carries 13.3% in two names exposed to the same memory-pricing variable — which is the concentration point The Concentration Premium makes about correlated positions that look independent on a position-by-position view.
What I am not claiming
Nothing here suggests SanDisk’s reported numbers are wrong or that its targets are dishonest. Every figure above is the company’s own, from a filing made this week, and the price-driven explanation of the margin expansion is the company’s explanation, not mine. I am also not forecasting NAND prices — I have no ability to do that, and the bull case for sustained pricing is credible. The narrower point is that the FY2030 model is a bet on the durability of a price level, that this is a different bet from a bet on execution, and that an investor should know which one they are making.
What would change my read
WatchFour things, in order of how much they would move me
1. Revenue per gigabyte. The single number that matters. SanDisk discloses it by segment; a sequential decline in Datacenter or Edge revenue per gigabyte is the first hard evidence the repricing is unwinding. 2. The volume/price mix. If future growth comes from exabytes rather than price, the model gets far more defensible, and the opposite if it does not. 3. Industry capacity announcements. The bull case rests on supply discipline; large new fab commitments from any major producer weaken it directly. 4. Opex ratio. Whether the 5% assumption is approached by growing revenue or by cutting R&D — those imply very different businesses in 2030.
The useful-life piece published earlier today made a point about an assumption sitting underneath reported earnings at four AI-infrastructure companies. This is the same kind of question asked of the book’s largest holding: not whether the printed number is accurate, but which variable it is standing on.