§ 01 Business Overview
SanDisk makes NAND flash memory. That is the storage technology inside solid state drives, data centers, smartphones, and laptops. For nearly a decade it was a division of Western Digital. In February 2025 it spun off as its own public company, and that spinoff is where this story starts.
The business sells into three markets. Datacenter is the one everyone is buying the stock for, because AI infrastructure runs on flash storage and SanDisk is one of the few pure-play ways to bet on that demand. Edge covers client devices like laptops and gaming consoles. Consumer covers retail products like SD cards and USB drives. Datacenter revenue grew 103% year over year in Q3 of fiscal 2025, and that number is what sent the stock from $36 at its first trade to over $700 at its peak.
Their newest technology is BiCS8, their latest generation 3D NAND chip, which delivers more storage per physical chip because data centers want density above almost everything else. They are also developing High Bandwidth Flash, aimed at AI inference workloads. It is early and unproven, but it points to where management wants to take the business.
The stock closed around $610 today. Twelve months ago it was trading at $27.89.
§ 02 Competitive Moat · Weak
Here is the uncomfortable part of this analysis, and it is worth saying directly: SanDisk does not have a durable moat, because NAND flash is a commodity.
Morningstar assigns it no economic moat because memory chips are fungible. One manufacturer's NAND is largely interchangeable with another's. When supply is tight, every NAND maker gets pricing power. When supply loosens, they all compete on price and margins collapse. That cycle has repeated itself consistently for two decades, and there is no structural reason SanDisk breaks it.
SanDisk is the fifth-largest NAND producer globally, behind Samsung, SK Hynix, Micron, and Kioxia. Its gross margins run about six percentage points below Micron's, not because SanDisk makes inferior products, but because in a commoditized market, no one has lasting pricing power over anyone else.
The bulls point to BiCS8 and proprietary firmware as differentiation. Some of that is real. But Samsung has a larger R&D budget, more manufacturing scale, and a longer enterprise track record. When supply loosens and pricing falls, technical differentiation does not protect margins. Nothing does.
The current pricing power SanDisk enjoys exists because 2026 NAND supply is effectively sold out. That is a market condition, not a moat. Market conditions change.
§ 03 Financial Snapshot
| Period | Revenue | Gross Margin |
|---|---|---|
| FY 2023 | $6.09B | Negative |
| FY 2024 | $6.66B | Low single digits |
| FY 2025 | $7.36B | 26% in Q4 |
| Q1 FY2026 | $2.31B | Rising toward 40%+ |
Revenue grew 10% in FY2025, which sounds modest. The quarterly trajectory tells a faster story: Q1 FY2026 came in at $2.31B, up 21% sequentially, with management guiding Q2 to $2.55B to $2.65B and gross margins of 41 to 43%.
Gross margin is the number that matters most here, because in a commodity business, when NAND prices rise, margins expand fast while production costs stay relatively fixed. That is the mechanism behind the recent numbers. ROIC on a trailing basis is negative because a $1.83 billion goodwill impairment charge in Q3 FY2025 wiped out prior profitability, but looking at the trend direction, Q1 FY2026 produced $112 million in GAAP net income. The business is recovering quarter over quarter.
This is where the moat section connects directly to the financials: every dollar of that margin expansion comes from NAND price increases in a supply-constrained market, not from anything SanDisk does better than its competitors. Micron's margins at the peak of the last NAND cycle collapsed to near zero within two years. That is the pattern this business runs on, and there is no reason to expect SanDisk to be exempt from it.
§ 04 Risk Rating
The risk is higher here than in either Palantir or ServiceNow, for a straightforward reason: SanDisk's entire earnings trajectory depends on a supply cycle staying tight, and supply cycles always turn.
Samsung alone has the capacity to flood the NAND market if it runs its fabs at full utilization. It has done exactly that in prior cycles, and SanDisk has no pricing power that would protect it when that happens.
A new risk emerged this week. Google unveiled a compression algorithm called TurboQuant that reduces memory requirements for AI models, and SanDisk stock fell 11% in a single session on the news. That reaction is worth studying, because it shows exactly how narrative-driven this stock is. The entire bull case assumes AI keeps consuming more memory. Google just demonstrated that assumption is not guaranteed.
The third risk is the valuation itself. SanDisk carries a $91 billion market cap against $7.36 billion in annual revenue and still-negative trailing earnings. That math only works if the cycle continues and margins expand to levels this business has never sustained for long.
§ 05 Bull vs. Bear
Bull case: The NAND supercycle is not over. AI data centers need exponentially more storage for every new model generation, and the entire 2026 NAND manufacturing supply is effectively already sold out. As the only pure-play publicly traded NAND company in the U.S., SanDisk captures the full benefit of that without the drag of a diversified business. BiCS8 is ramping toward the majority of production by end of FY2026, which drives further margin expansion. Management has guided for 65 to 67% gross margins near the peak of the cycle. If that happens, the earnings power of this business looks completely different from what trailing numbers show.
Bear case: Smart money is already leaving. Cliff Asness of AQR Capital cut his SanDisk position by 22% in Q4 while adding to Nvidia. Steven Schonfeld did the same, trimming 27% of his SanDisk stake while tripling his Nvidia position. The difference between those two trades is a moat. Nvidia has one. SanDisk does not. When Samsung brings new NAND capacity online, pricing will compress, margins will fall, and this stock will re-rate hard. Bear fair value on conservative assumptions sits around $157, which is a 75% decline from today's price.
SanDisk is a real business riding a real wave. That is not the problem. The problem is $91 billion for a company with negative trailing earnings and no moat in a market that has never sustained commodity memory valuations at this multiple for long.
The quarterly numbers are genuinely improving. If you caught this at $36, the thesis was sound and the execution delivered. If you are looking at it now at $610, the easy money is gone and what is left is a bet on a cycle that history says will turn.
Verdict: Avoid at current prices. My price to get interested: $200 to $250. That would still represent a significant premium to pre-AI levels, but it would leave room for the cycle to turn without being wiped out.
Three things that would change my view: (1) Samsung or SK Hynix announcing capacity cuts rather than expansions. (2) Gross margins sustaining above 50% for three or more consecutive quarters. (3) TurboQuant-style memory compression failing to gain adoption at other hyperscalers.
I picked SanDisk because it was the hottest stock of the year and I wanted to understand why. What I found is that the why is actually pretty simple, and simple is not always a good sign when a stock is up 1,600%.
The most important thing I learned this week is the difference between a market condition and a moat. Going in, I thought strong pricing power meant a strong business. What I did not understand is that pricing power can come from two completely different places. It can come from something a company built, like Palantir's ontology or ServiceNow's switching costs - things competitors cannot easily copy. Or it can come from a supply shortage that nobody planned and nobody controls. SanDisk has the second kind. When supply tightens across an entire industry, every producer in that industry gets pricing power at the same time. That is not a competitive advantage. It is just good timing.
The goodwill impairment charge confused me at first. A $1.83 billion write-down in a single quarter made the trailing financials look terrible, and I almost dismissed the business because of it. Then I understood what a goodwill impairment actually means: it is an accounting adjustment for an asset that was acquired at a price that turned out to be too high. It is a backward-looking charge, not a forward-looking one. Once I separated that from the operating performance, the quarterly trend became a lot clearer. That is a lesson I will carry into every analysis going forward: read the footnotes before you react to the headline number.
The Google TurboQuant news hit while I was working on this analysis, and watching the stock drop 11% in one session because of an algorithm announcement was genuinely interesting to observe in real time. It showed me something about how narrative-driven stocks behave. When a stock's valuation is built almost entirely on a future story, any news that questions that story creates an outsized reaction, even if the long-term picture has not actually changed. That is the tax you pay for buying at the front of a cycle.
The thing I am still working through is how to value a cyclical business. With Palantir and ServiceNow, I could point to recurring subscription revenue, renewal rates, and compounding growth. None of those apply here. SanDisk's earnings in two years could be dramatically higher than today, or dramatically lower, depending entirely on a supply dynamic that no analyst can predict with confidence. I do not yet have a clean mental model for how to assign a fair value to that kind of business. That is next on my list.