SanDisk was spun out of Western Digital in February 2025 at roughly $48 a share. It now trades around $1,400. That is a gain of more than 2,800% in about fifteen months, and it happened without a new product category, a new market, or a new technology. It happened because of price.
The cycle in numbers
| Input | Reading |
|---|---|
| SNDK, spinoff price (Feb 2025) | ~$48 |
| SNDK, Q3 2026 | ~$1,400 |
| Move | >2,800% in ~15 months |
| Q3 2026 revenue | $5.95B |
| Q2 2026 revenue | $3.03B |
| Sequential revenue growth | +96% |
| AI data center share of 2026 NAND demand | ~44% |
| WDC, year to date 2026 | ~+321% |
Revenue nearly doubling sequentially is the giveaway. No storage company doubles unit shipments in ninety days. That is pricing.
What is actually happening
AI data centers went from a marginal NAND customer to roughly 44% of total demand. Supply did not respond, because adding NAND capacity takes years and the last cycle taught the industry that overbuilding is how you destroy your own margins. So the industry did the disciplined thing, and the disciplined thing produced a shortage.
In a shortage, contract pricing resets rather than drifts. Reports have suggested SanDisk could roughly double pricing on certain 3D NAND used in enterprise SSDs. When you double price on a product whose marginal cost barely moves, every incremental dollar lands close to the operating line. That is why the revenue growth and the stock move are on completely different scales from anything in the underlying volume data.
The same force lifted the whole shelf — Micron in DRAM and HBM, Western Digital in high-capacity drives, SK Hynix, Seagate. This is not a company story. It is a commodity story wearing a semiconductor logo.
The part I'd be careful about
Every memory cycle in history has ended the same way: the price signal works, capacity gets added, and the same operating leverage that produced these earnings runs in reverse. The math that turns a doubling in price into a triple in earnings turns a 30% price decline into something considerably worse.
The specific thing I watch is not demand — AI demand is real and probably durable. It is capacity announcements. The moment the industry starts guiding capex up meaningfully, the clock on this cycle starts, and it starts roughly two years before the supply actually arrives. Memory stocks have historically topped long before earnings did, because the market prices the capacity announcement, not the wafer.
The second risk is subtler: a lot of this demand is AI infrastructure buildout, which is itself a capex cycle. A commodity cycle riding on top of a capex cycle is levered twice to the same thing.
My take
I don't short cycles that are working and I don't chase them at 2,800%. Both of those are ways of expressing an opinion about timing that nobody has.
What I would say is that anyone holding this needs to be honest about what they own. This is not a compounder. It is a superbly executed position in a commodity at a moment of extreme scarcity, and the correct behavior with such a position is to have a plan for the exit written before the exit is obvious. A trailing stop on a name that has gone up 28-fold is not timidity, it is arithmetic.
For anyone looking at the sector now, the more interesting question is which of these businesses keeps some pricing power after the cycle turns. High-capacity drives and HBM have structural moats that commodity NAND does not. That is where I'd rather be as the cycle matures.
Bottom line: the shortage is real and the earnings are real. But a 2,800% move on pricing is a cycle, not a compounding machine — and cycles are ended by the very capex their prices invite.
This is analysis, not investment advice.