At $572.50, the stock was still up 141% in 2026. But Q3 guidance implied the earnings reset had further to run, while a new $1 billion Nanya stake raised the cost of being early.
Sandisk entered 2026 as a newly independent storage company and became one of the market's fastest-moving AI infrastructure trades. The shares closed 2025 at $237.38. By March 19, they had reached $772.09, a gain of 225% in less than three months. Then the trade broke hard. Seven sessions later, Sandisk closed at $572.50, down 25.9% from the peak.
That pullback is the reason to study the stock now. The operating evidence had strengthened even as the share price corrected. Fiscal second-quarter revenue reached $3.03 billion, up 31% sequentially and 61% year over year. GAAP gross margin expanded to 50.9% from 29.8% one quarter earlier. Non-GAAP diluted earnings reached $6.20 per share, more than five times the prior quarter.
Management's fiscal third-quarter guidance implied another step-change. The midpoint called for $4.6 billion of revenue, 66% non-GAAP gross margin, and $13.00 of non-GAAP diluted earnings per share — 52% revenue growth and 110% EPS growth in a single quarter versus Q2.
01CHAPTER 01The Setup — A 159% Run and a 26% Drawdown
The original StockResearch weekly called Sandisk a buy on pullback. The timing deserves precision. At the March 27 close of $615.83, the shares were up 159.4% from December 31. On March 30, they fell another 7.0% to $572.50. That left the stock up 141.2% year to date and 25.9% below its March 19 peak. All valuation and recommendation analysis in this report uses $572.50.
The volatility was not random. Revenue had declined from $1.88 billion in Q2 FY25 to $1.70 billion in Q3 FY25. GAAP gross margin fell from 32.3% to 22.5%. By Q4 FY25, revenue had recovered to $1.90 billion and gross margin to 26.2%. Q1 FY26 improved again to $2.31 billion and 29.8%. Then Q2 FY26 broke the prior range: $3.03 billion of revenue and 50.9% GAAP gross margin.
The stock was therefore responding to a genuine change in the earnings function. Revenue grew, but margin grew much faster. That same mechanism makes the trade dangerous: if prices reverse while fixed costs remain, incremental profit can disappear just as quickly.
02CHAPTER 02Catalyst #1 — Pricing, Not Bits, Drove the Earnings Reset
NAND businesses are often described as unit-growth stories. That description misses the near-term economics. In fiscal Q2, Sandisk's exabytes sold increased 22% year over year. Average selling price per gigabyte increased 36%. Revenue increased 61%.
In Q2, ASP improved faster than cost per gigabyte, which drove a $935 million year-over-year increase in gross profit and a 1,900-basis-point expansion in GAAP gross margin. GAAP operating income rose from $176 million to $1.07 billion sequentially. This is the operating leverage investors were paying for.
Edge provided the clearest evidence that pricing was broad: Q2 Edge revenue reached $1.68 billion, up 63% year over year, with ASP per gigabyte up 55% on only 10% exabyte growth. Consumer revenue increased 52%. Datacenter behaved differently — revenue increased 76% year over year, but ASP per gigabyte actually declined 8% as exabytes increased 90%. That was a volume story, not yet a pricing story.
03CHAPTER 03Catalyst #2 — Datacenter Was Small but Becoming Strategic
Sandisk's Q2 datacenter revenue was $440 million — up 64% sequentially and 76% year over year, but only 14.5% of company revenue. The modest mix should make the thesis more disciplined, not less interesting: Sandisk did not need datacenter to explain Q2. It needed datacenter to improve the quality of future earnings.
By January, Sandisk had completed qualification of its PCIe Gen5 high-performance TLC drive at a second hyperscaler, with additional hyperscaler qualifications expected over the following quarters and BiCS8 TLC solutions to follow. These statements did not guarantee orders. They established a visible funnel.
The counterpoint is competition. Samsung, SK hynix and Solidigm, Micron, Kioxia, and other suppliers all target enterprise storage. Investors should monitor signed qualifications and revenue, not broad statements about AI.
04CHAPTER 04Catalyst #3 — Q3 Guidance Changed the Earnings Base
Management guided to $4.4–4.8 billion of revenue, 65%–67% non-GAAP gross margin, and $12–14 of non-GAAP diluted EPS. At the midpoint, revenue would increase 52% sequentially and EPS would increase 110%. Annualizing the $13 midpoint would produce $52 of EPS — not a forecast, but a sensitivity that shows why the stock could trade above $500 without requiring a technology-company multiple.
The valuation problem is normalization. A shortage quarter can generate profits far above the through-cycle level. Our scenarios therefore use normalized EPS below the simple annualized guide in the bear and base cases.
The quarter that broke the prior range
Q3 FY26 figures are management guidance published at the historical cutoff, not reported results.
05CHAPTER 05The Manufacturing Model — Leverage Without Full Ownership
Substantially all of Sandisk's flash memory wafers come from Flash Ventures, three joint ventures with Kioxia. Sandisk owns 49.9% and Kioxia owns 50.1% and the fabrication facilities. Sandisk generally purchases half of Flash Ventures' output at cost plus a small markup, pays its share of variable costs based on a rolling forecast, and pays half of fixed costs regardless of how much output it chooses to purchase. Orders placed for up to three months are binding.
As of January 2, Sandisk disclosed $3.10 billion of maximum estimable loss exposure related to Flash Ventures — a ceiling, not an expected loss, but a useful measure of the economic commitments sitting around the reported balance sheet. The joint ventures create operating leverage in both directions: valuable during a shortage, but fixed-cost obligations and binding forecasts can produce inventory and underutilization charges during an oversupply.
06CHAPTER 06The Nanya Investment — Supply Insurance at a High Price
The March 25 agreement with Nanya arrived during the share-price correction. Sandisk agreed to purchase approximately 139 million newly issued Nanya shares for about $1.0 billion, equal to roughly 3.9% of Nanya on a fully diluted basis, at a 15% discount to the 30-day average price, with a three-year statutory lockup. A concurrent multi-year arrangement makes Nanya a DRAM supplier to Sandisk.
At January 2, Sandisk held $1.54 billion of cash and $650 million of debt. The transaction's purchase price was larger than net cash and equal to about 65% of gross cash. The equity component is debatable: Sandisk could have signed a supply agreement without becoming a large shareholder. The investment converts liquid cash into a minority stake that cannot be sold for three years, near the top of a memory cycle.
07CHAPTER 07Competition and Market Structure
Sandisk competes with vertically integrated NAND suppliers including Samsung, SK hynix, Micron, Kioxia, and Yangtze Memory Technologies. The competitive landscape has three layers: NAND technology and cost, product qualification, and supply discipline. A rational industry can maintain pricing even when unit costs fall; a race for share can destroy the economics of every producer.
Sandisk's strengths were its broad product portfolio, recognizable consumer brand, controller and firmware capability, and long-running Kioxia partnership. Its weaknesses: it did not control the fabrication facilities underlying nearly all wafer supply, its datacenter revenue base was smaller than the market narrative implied, and it was operating with only a year of standalone public-company history.
08CHAPTER 08Valuation & Technical Structure
At $572.50 and approximately 148 million shares, Sandisk's market capitalization was about $84.7 billion — not justifiable by fiscal 2025 results, only by the Q3 guide representing a durable new earnings base. We use a normalized-EPS framework: what the business could earn across a less extreme year, times a multiple reflecting growth, cycle risk, product mix, and balance-sheet quality.
Normalized EPS $24 at 14x. NAND supply catches up faster than expected, Edge pricing weakens, datacenter converts slowly. ~41% downside.
Normalized EPS $38 at 18x. Supply stays tight through most of 2026, datacenter grows faster than the company. ~19% upside.
Normalized EPS $52 at 20x. The Q3 earnings rate proves durable, qualifications convert, BiCS8 supports cost competitiveness, industry stays disciplined. ~82% upside.
With probabilities of 25% bear, 50% base, and 25% bull, the weighted value is $686 per share — 19.8% above the March 30 close. The distribution matters more than the average: a 20% expected upside paired with 41% bear-case downside justifies a staged entry after a 26% drawdown, not chasing the stock after a rebound.
09CHAPTER 09Risks & Counter-Thesis
- This is a cyclical shortage being valued like a secular compounder. Competitors have the same incentive to increase output when returns rise, and customer inventory can accumulate.
- Datacenter may not close the gap. Qualifications can take longer than expected and hyperscalers can negotiate down prices once supply improves.
- Flash Ventures can amplify a downturn. Binding purchases based on forecasts mean the company can carry expensive inventory while still funding capacity.
- The Nanya transaction adds capital-allocation risk. A minority stake with a three-year lockup may not provide control, immediate earnings, or liquidity.
- The standalone history remains short, and competition from Samsung, SK hynix, Micron, Kioxia, and Chinese suppliers is intense.
- The valuation leaves room for disappointment. At an $84.7 billion market cap, Sandisk must earn far more than it did in fiscal 2025.
Sandisk was not cheap in March 2026. It was becoming much more profitable. The pullback from $772.09 to $572.50 created a better entry without requiring the thesis to change — but the $1 billion Nanya investment made the decision less comfortable, consuming liquidity at the top of the cycle. The appropriate historical rating was Buy on Pullback, with a staged position and explicit operating thresholds: Sandisk did not need NAND pricing to stay at peak levels forever. It needed the shortage to last long enough for datacenter mix, BiCS8 conversion, and normalized earnings to establish a higher floor.