Sandisk Corporation
Equity Research Report
Report Date: June 11, 2026 | Data as of: Q3 FY2026 Earnings (April 30, 2026) + market data June 10, 2026 | Sector: Semiconductors — NAND Flash Memory | HQ: Milpitas, California, USA
Business Model — What Does Sandisk Actually Do?
Sandisk makes NAND flash memory — the chips that store data persistently in SSDs, smartphones, memory cards, and USB drives. If DRAM (made by Micron, Samsung, SK Hynix) is a computer’s short-term working memory, NAND is its long-term storage: where files, models, photos, and databases actually live when the power is off.
The company has a long, winding history: founded in 1988, acquired by Western Digital in 2016, and spun off as an independent company in February 2025 — beginning trading on February 24, 2025 (Sandisk 10-Q, SEC filing). What followed was one of the most extraordinary first years of any public company in recent memory: the stock rose from a 52-week low of ~$39 to over $1,800, as the AI data center boom collided with a global NAND shortage.
Sandisk’s manufacturing runs through Flash Ventures — a decades-old joint venture with Japan’s Kioxia, operating wafer fabs in Yokkaichi and Kitakami, Japan (K1, Y7, K2 facilities per Sandisk’s 10-Q). This JV structure means Sandisk shares world-class fab capacity and R&D costs without bearing the full capital burden alone — but also means its supply is tied to a partner it doesn’t control.
The “New Business Model” (NBM) is the most important strategic shift to understand. Historically, NAND was a brutal commodity business: prices swung wildly with supply/demand, and producers suffered multi-year loss periods. Sandisk is now signing multi-year supply contracts with committed volumes, backed by enforceable financial guarantees (prepayments and third-party financial instruments that compensate Sandisk if customers walk away). This attempts to convert a boom-bust commodity business into something closer to a contracted industrial supplier — the central question of the entire investment story.
Revenue Streams
| Segment | What it includes | Q3 FY26 (Apr 2026) | Trend |
|---|---|---|---|
| Datacenter | Enterprise SSDs (TLC performance + QLC capacity tiers) for AI infrastructure, hyperscalers, and the Stargate project | $1.47B (~25%) | +233% QoQ, +645% YoY |
| Edge | NAND for smartphones, PCs, automotive, IoT — sold to device makers | ~$3.1B (~52%) | Growing on AI phone storage upgrades |
| Consumer | Retail SanDisk-brand products: memory cards, USB drives, portable SSDs | ~$1.4B est. (~23%) | −10% QoQ — price-sensitive demand fading |
| TOTAL | — | $5.95B | +97% QoQ, +251% YoY |
Critical nuance for beginners: most of this growth is price, not volume. TrendForce reported the top-five NAND suppliers’ combined Q1 2026 revenue jumped 83.7% QoQ to $38.9 billion “primarily fueled by elevated prices due to the scarcity, not by higher output volumes” (IndexBox citing TrendForce, June 2026). When the product you sell triples in price while your costs barely move, margins explode — Sandisk’s gross margin went from industry-typical ~30% levels to 78.4% in Q3. That’s also exactly why the durability question matters so much: prices that triple can also fall.
Industry Trends — The NAND Supercycle
1. The AI Storage Shortage. AI workloads — inference, retrieval-augmented generation (RAG), KV cache, agentic systems — require massive amounts of fast, low-latency storage sitting next to GPUs. Data center became the largest NAND end market in 2026 for the first time. TrendForce upgraded its Q1 2026 NAND contract price forecast from +33–38% QoQ to as high as +55–60% as AI demand decisively outpaced supply (Accio/TrendForce summaries, early 2026).
2. Supply Discipline — Deliberate and Coordinated. Samsung, SK Hynix, Kioxia, and Micron all scaled back NAND production in H2 2025 (SK Hynix cut ~10%), after the brutal 2023–24 downturn taught producers that flooding the market destroys everyone’s profits (TrendForce citing Chosun Biz, Nov 13, 2025). Crucially, TrendForce reports major suppliers will not add substantial new capacity in 2026 — new fabs won’t reach meaningful volume until late 2026–2027 (NAND Research, March 2026).
3. The HBM Squeeze — A Structural, Not Cyclical, Driver. Memory makers are converting wafer capacity to High Bandwidth Memory for AI accelerators at roughly a 3-to-1 ratio (every HBM wafer displaces ~3 conventional wafers’ worth of output, per Micron via NAND Research, May 2026). This structurally starves conventional DRAM and NAND supply for as long as the AI buildout continues — a key reason analysts argue this cycle differs from previous ones.
4. Edge AI Storage Upgrades. As 2026 smartphones integrate local LLMs, baseline storage is shifting from 256GB toward 1TB per device — a quiet but massive volume tailwind for mobile NAND (FinancialContent Deep Dive, March 31, 2026).
Competitive Landscape
NAND is a concentrated oligopoly: the top manufacturers control ~92% of global capacity. Sandisk holds roughly 13% market share — ranked 4th–5th by volume, tied with Micron in recent quarters — but is regarded as among the most agile players in the high-margin enterprise segment (FinancialContent, March 2026; TrendForce Q1 2026 data).
Financial Performance
| Metric | Q1 FY26 (Oct 2025) |
Q2 FY26 (Jan 2026) |
Q3 FY26 (Apr 2026) |
Q4 FY26 Guide (due ~Jul 2026) |
|---|---|---|---|---|
| Revenue | ~$2.31B | $3.03B (+31% QoQ) | $5.95B (+97% QoQ) | $7.75–8.25B |
| GAAP Net Income | $112M | $803M | $3,615M | — |
| GAAP Diluted EPS | ~$0.76 | $5.15 | $23.03 | — |
| Non-GAAP Diluted EPS | ~$1.30 | $6.20 | $23.41 | $30.00–33.00 |
| Gross Margin | ~29% | ~40s% | 78.4% | Elevated |
| Datacenter Revenue | ~$0.3B | ~$0.44B (+64% QoQ) | $1.47B (+233% QoQ) | QLC Stargate revenue begins |
Valuation — The Hardest Part of This Story
Valuing SNDK is genuinely difficult, and honest analysts disagree sharply. The stock closed at $1,643 on June 10, 2026, with an all-time closing high of $1,831 on June 3 — versus a 52-week low of $39.44 (MacroTrends, June 2026). A ~4,000% one-year return means the market has already priced in an enormous amount of good news. The core problem: which “E” do you put under the P/E?
| Metric / View | Value | Comment |
|---|---|---|
| Market Cap | ~$126B | From ~$5–6B at spinoff — now larger than many S&P 100 companies |
| P/E on FY2026 consensus EPS (~$65) | ~25x | Looks reasonable — but FY26 includes pre-boom quarters |
| Implied annualized run-rate (Q4 guide ~$31.50 EPS × 4) | ~13x | Looks cheap — IF peak-cycle earnings were sustainable (historically they are not) |
| Forward P/E per 24/7 Wall St (Jun 4, 2026) | ~57x | “Rich for a memory cyclical, even one positioned for AI” — methodology differences show how contested the number is |
| Street consensus price target | ~$1,550 | BELOW the market price — the market is more bullish than the average analyst (TIKR, May 2026) |
| FY2027 consensus revenue | ~$45B | Implies a further +125% from FY2026’s ~$20B (Quartr summary, May 28, 2026) |
The classic cyclical valuation trap, explained simply: commodity-cyclical stocks often look cheapest on P/E at the exact top of the cycle (earnings are peaking) and most expensive at the bottom (earnings are depressed). A 13x multiple on peak earnings can be far more dangerous than 50x on trough earnings. The bull counterargument is that the NBM contracts and AI structural demand mean “this time the cycle is dampened” — which is precisely the proposition the current price is paying for.
Growth Drivers
Near-term (FY2026–27):
Q4 FY2026 guidance of $7.75–8.25B implies yet another ~35% sequential step-up. QLC-based “Stargate” capacity solutions begin shipping revenue in Q4 — connecting Sandisk directly to the largest AI infrastructure project in the world. NAND contract prices continue rising (TrendForce: +33–60% QoQ ranges through early 2026) while no meaningful new industry supply arrives before late 2026–2027.
Medium-term — the NBM flywheel:
Five NBM agreements signed to date (three in Q3, two more in early Q4), spanning up to five years. The three Q3 contracts alone carry $42 billion in minimum contractual revenue (RPO); guarantees across all five exceed $11 billion, including $400M of prepayments already on the balance sheet. Over one-third of FY2027 bit supply is committed, with more agreements in active discussion. Contracts include both fixed and variable pricing elements — preserving upside participation if market prices keep rising. (SEC 8-K Apr 30, 2026; Zacks via TradingView, Jun 8, 2026; Motley Fool, May 27, 2026)
Long-term — High Bandwidth Flash (HBF):
HBF targets DRAM-like bandwidth at a fraction of HBM’s cost, aimed squarely at AI inference — where the industry’s compute spending is migrating. Product launches are expected late 2026 into 2027. If HBF is adopted as a standard tier in the AI memory hierarchy (alongside or partially replacing HBM for inference), it could roughly double Sandisk’s addressable market. Edge AI (1TB-standard smartphones) provides a second, quieter long-term volume driver. (Quartr May 2026; FinancialContent Mar 2026)
Key Risks
| Risk | Severity | Explanation |
|---|---|---|
| NAND Cycle Reversal | High | The defining risk. Most of the revenue explosion is price-driven, and memory prices have crashed 50–70% in past downcycles. If hyperscaler demand pauses or new supply (late 2026–2027 fabs, Kioxia’s plan to double capacity by 2029) arrives into softening demand, margins compress violently. TIKR: a demand slowdown “could compress both volumes and margins faster than the NBM agreements can offset.” Sandisk itself was cutting fab utilization as recently as H1 FY2026. |
| Peak-Multiple-on-Peak-Earnings | High | After a ~4,000% run, the stock trades above street consensus (~$1,550). Published bear cases sit ~40% below current prices. Even if fundamentals stay strong, any guidance miss or NAND spot-price wobble could trigger outsized drawdowns — high-beta cyclicals fall faster than they rise. |
| Kioxia JV Dependence | Medium | All wafer supply flows through Flash Ventures with Kioxia (K1/Y7/K2 fabs in Japan). Sandisk doesn’t unilaterally control capacity, technology cadence, or expansion decisions — and its JV partner is also a competitor. Earthquake/operational risk at concentrated Japanese sites adds a physical dimension (a 2022 contamination incident at these fabs disrupted global NAND supply). |
| Customer Concentration (Hyperscalers) | Medium | Datacenter growth is concentrated among a handful of AI infrastructure builders and the Stargate ecosystem. NBM guarantees ($11B+) mitigate but don’t eliminate this — guarantees compensate for broken commitments, they don’t replace a lost growth trajectory. (24/7 Wall St, Jun 2026) |
| NBM Model Unproven Through a Downturn | Medium | The contracts have never been tested in a falling-price environment. Variable pricing elements cut both ways; customers under financial stress renegotiate. The thesis that NBM “structurally dampens the cycle” is a forward-looking claim, not yet an observed fact. |
| Consumer Segment Softness | Medium | Consumer revenue fell ~10% QoQ in Q3 as retail buyers balked at higher prices — an early demonstration that NAND price hikes destroy price-sensitive demand. If edge/mobile buyers also push back, the volume side weakens. |
| Technology Execution (BiCS10 / HBF) | Low-Med | Aggressive node transitions (BiCS8→BiCS10) carry yield/defect risk; HBF is a new architecture that must win ecosystem adoption against entrenched HBM roadmaps. Slippage would hand momentum to Samsung/SK Hynix. (FinancialContent, Mar 2026) |
| Tariffs / Geopolitics | Low | Japan-concentrated manufacturing with global sales creates tariff exposure; flagged by analysts but currently manageable. YMTC becoming unconstrained (export-control easing) is a tail risk to long-term industry pricing. |
Bull / Base / Bear Cases
Research Summary
Sandisk — The Analyst’s View in Plain Language
Sandisk is the most extreme expression of the AI infrastructure trade in the entire semiconductor universe. Sixteen months ago it was an unloved spinoff trading near $40, cutting fab utilization because nobody wanted flash memory. Today it is a ~$126 billion company posting 78% gross margins, $23 of quarterly EPS, a $42 billion contracted backlog, zero debt, and a $6 billion buyback — with guidance implying the numbers get even bigger next quarter. Both of those companies are the same company. That is the essential lesson of memory economics.
What’s genuinely new and real: the AI inference storage demand is not imaginary — data center became NAND’s largest market, management keeps revising demand upward, and the entire supplier oligopoly is exercising unprecedented discipline. The NBM contracts, with $11B+ in enforceable guarantees, are a serious structural attempt to break the boom-bust pattern, and no competitor moved on this faster than Sandisk. The zero-debt balance sheet means Sandisk — unlike leveraged AI infrastructure plays — can survive any downturn.
What’s old and unchanged: NAND remains a commodity whose price tripled mostly because supply was deliberately restricted into a demand spike. Every previous memory supercycle ended the same way — with supply arriving into decelerating demand — and new fabs are already scheduled for late 2026–2027. The consumer segment’s 10% decline is the canary: high prices destroy demand at the margin. And critically, the stock now trades above the average analyst’s target, meaning the market is paying today for the bull case to be right.
The honest framing for a beginner: SNDK is no longer an early-cycle opportunity; it is a late-recognition, high-momentum position where the central question has shifted from “is the business inflecting?” (answered: yes) to “how much of the inflection is already in the price, and is this cycle truly different?” (unanswered). The ~$1,000-to-$3,400 spread in credible published scenarios is the widest of any large-cap this report series has covered — wider than ASML’s, wider than Oracle’s. Position sizing and a pre-defined view on NAND pricing in 2027 matter more here than for almost any other stock. Watch: Q4 results (~July 2026), NBM contract additions, TrendForce monthly NAND pricing, competitor capacity announcements, and the HBF launch timeline.
Primary Sources Referenced
Sandisk Q3 FY2026 Press Release / SEC Form 8-K (April 30, 2026) · Sandisk Q2 FY2026 Press Release / 8-K (January 29, 2026) · Sandisk 10-Q Filings (periods ended Jan 2 & Apr 3, 2026) · Sandisk Q3 FY2026 Earnings Call Transcript (Globe and Mail summary, April 30, 2026) · Zacks “SNDK’s New Business Model Boosts Profitability” (TradingView, June 8, 2026) · Motley Fool “Sandisk Has Climbed From $36 to $1,590 in a Year” (May 27, 2026) · TIKR “Sandisk Has Returned Over 4,000% Since Its Spinoff” (May 2026) · 24/7 Wall St SNDK Price Prediction (June 4, 2026) · MacroTrends SNDK Price History & Market Cap (June 9–10, 2026) · TrendForce “Top Five NAND Suppliers Post 23.8% QoQ Growth in 4Q25” (March 3, 2026) · IndexBox / TrendForce Q1 2026 NAND Revenue Data (June 2026) · NAND Research “Memory & Flash Crisis Updates” (March & May 2026) · TrendForce / Chosun Biz “NAND Giants Cut Output in 2H25” (November 13, 2025) · FinancialContent “Deep Dive: SanDisk and the 2026 NAND Flash Shortage” (March 31, 2026) · Quartr Sandisk Investor Summary (May 28, 2026) · eeNews Europe / TrendForce Memory Price Forecasts (January 7, 2026)