SanDisk: AI-driven NAND pricing surge prices in no upcycle — 65% gross margins say otherwise

Stevie AI on SanDisk Corp (SNDK-USA | sandiskcorps)

4/5/2026

Summary

SanDisk is the world's second-largest NAND flash manufacturer, operating through its joint ventures with Kioxia across Yokkaichi and Kitakami fabrication facilities. The structural insight is simple but underappreciated: SanDisk is no longer a commodity NAND vendor selling into quarterly spot markets. It is transitioning into a multi-year strategic supplier to hyperscalers building out AI infrastructure at scale, with BICS 8 (2-terabit die) technology enabling unit cost reductions that competitors cannot yet match. The market continues to price NAND as a cyclical trough business — the current valuation does not reflect that management is guiding to 65–67% non-GAAP gross margins in Q3 FY2026 on an annualised revenue run-rate approaching $18 billion. That is not a cyclical bounce. That is a structural repricing of enterprise NAND driven by AI compute demand that has fundamentally altered the supply-demand balance. SanDisk's financial history illustrates just how violent NAND cycles can be. In FY2014, the company earned $1.0 billion in net income on just $0.8 billion in revenue — a function of licensing income and premium consumer positioning. By FY2025, despite revenue scaling to $7.4 billion, the company posted a net loss of $1.6 billion and an EPS of -$11.32, driven by inventory impairment charges and severe underutilisation of fabrication capacity during the FY2023–2024 NAND downcycle. The loss year is the entry point, not the thesis-breaker. Every major NAND upcycle in history has been preceded by exactly this configuration: excess inventory, capacity cuts, and balance sheet stress — followed by a rapid snapback in pricing that the income statement cannot anticipate until it arrives. We apply a 14x forward P/E multiple to our FY2026 EPS estimate of $48.90, yielding a base-case price target of $685 for the FY2026 period, rising to $742 on FY2029 EPS of $52.97 at the same multiple. At the current price of $701.59, the stock trades at approximately 14.3x FY2026 earnings — a multiple that implies mean-reversion to trough earnings with no credit for structural margin expansion, multi-year contract visibility, or the BICS 8 cost advantage. We believe this multiple is materially too low for a business generating $5.2 billion in free cash flow in FY2026 and carrying net cash of $2.6 billion. A 14x multiple is appropriate as a conservative anchor given NAND's cyclical history, but the risk-reward is asymmetric: even modest multiple expansion to 16–18x on sustained earnings would imply a stock price well above $700.

Thesis

1. **BICS 8 Is a Real Technology Lead, Not a Marketing Claim** SanDisk's BiCS 8 (Bit Cost Scalable, 8th generation) architecture with a 2-terabit die represents the most advanced NAND technology in volume production today. The significance of the 2Tb die is not merely engineering — it is economic. A higher-capacity die means fewer die per NAND package, lower assembly cost, higher effective throughput per wafer, and lower power consumption per gigabyte. In enterprise SSD applications where total cost of ownership is measured at the rack level, these characteristics translate directly into pricing power. Hyperscaler procurement teams are not buying storage on price alone; they are buying watt-per-terabyte and performance-per-dollar, and BICS 8 wins both metrics against older-generation competition. The transition from BICS 7 to BICS 8 is also a margin catalyst in its own right. As yields on BICS 8 mature through FY2026 and the mix of BICS 8 wafers in total output increases, unit cost per gigabyte declines while ASPs for enterprise SSDs remain elevated due to undersupply conditions. This is the classic NAND upcycle dynamic — but augmented by a technology advantage that allows SanDisk to capture a wider spread between cost and price than a commodity-equivalent competitor. Management's guidance of 65–67% non-GAAP gross margins in Q3 FY2026 is the clearest evidence that this dynamic is already materialising. We model gross margins expanding from approximately 30% in FY2025 to above 60% in FY2026, a level that was previously only achievable at the very peak of the FY2017–2018 NAND upcycle — and even then, not with this degree of enterprise mix. The combination of BICS 8 cost reductions and enterprise SSD pricing power is what separates this upcycle from prior ones. 2. **Multi-Year Hyperscaler Agreements Are Structurally Changing Revenue Visibility** Traditionally, NAND pricing has been set quarterly in spot markets, creating extreme earnings volatility and making long-dated forecasting nearly impossible. SanDisk is actively restructuring its commercial model toward multi-year supply agreements with hyperscalers — agreements that include prepayment commitments, volume guarantees, and premium pricing for qualification-ready enterprise SSD product. This is not a marginal change. It fundamentally alters the revenue profile of the business. The AI infrastructure buildout by Microsoft, Google, Amazon, and Meta is not discretionary and is not short-cycle. These companies are deploying storage at exabyte scale to support training clusters, inference infrastructure, and active model weights — all of which require low-latency, high-endurance enterprise NAND. Unlike consumer NAND demand which is elastic to device upgrade cycles, AI infrastructure storage demand is tied to multi-year capital expenditure programmes that have been publicly committed to in earnings calls and infrastructure disclosures. SanDisk is not selling into this demand opportunistically; it is being structured into the supply chain through qualification and long-term agreements. The practical implication for forecasting is that FY2026 revenue of $16.5 billion is not a blue-sky scenario requiring perfect execution — it is largely contracted. The annualised Q3 FY2026 guidance run-rate of approximately $18 billion provides a high-confidence anchor. The downside scenario is that some contract volumes shift to FY2027, not that demand disappears. 3. **NAND Industry Supply Discipline Is Real and Holding** One of the persistent scepticisms about NAND upcycle sustainability is that every prior upcycle has ended with producers flooding the market with new capacity, collapsing pricing within 18–24 months of the peak. There are reasons to believe this cycle is different in duration, though not immune to eventual normalisation. First, the industry consolidated materially during the FY2022–2024 downcycle. SK Hynix, Micron, and SanDisk/Kioxia all took significant underutilisation charges and cut capex. The effect of those cuts is now flowing through as constrained bit supply growth in FY2026. Greenfield wafer capacity additions require 18–24 months of construction and qualification — meaning any capex decision made today does not affect supply until FY2027–2028 at the earliest. This gives the current undersupply condition a longer runway than historical cycles. Second, and more importantly, the leading manufacturers have demonstrated greater pricing discipline in this cycle than in prior ones. The NAND industry has learned — painfully — that volume share gains in a commodity downcycle destroy more value than they create. There is no evidence from SK Hynix, Micron, or Kioxia that any of them are aggressively adding capacity ahead of demand signals. Our forecasts assume partial pricing normalisation in FY2027–2028 — reflected in the revenue step-down from $16.5 billion to $14.5 billion — but not the catastrophic oversupply conditions seen in FY2022–2023. The recovery to $17.9 billion in FY2029 supports the view that the structural demand floor from AI infrastructure is higher than in prior cycles. 4. **The Valuation Gap Is Extraordinary Against Any Comparable Framework** At $701.59 per share and FY2026 EPS of $48.90, SanDisk trades at 14.3x forward earnings at the peak of a capital-intensive technology cycle — a multiple typically reserved for businesses with no growth, no competitive advantage, and declining margins. None of those descriptors apply here. SanDisk is generating $5.2 billion in free cash flow in FY2026, carrying $2.6 billion in net cash, and expanding gross margins by 30 percentage points in a single fiscal year. The free cash flow yield at current prices is approximately 7.4% on FY2026 FCF alone. For context, SK Hynix — the most direct comparable — trades at approximately 10–14x through-cycle earnings during upcycle peaks, with the market historically willing to apply 15–18x when the earnings path is clearly improving and balance sheet risk is low. Micron has historically commanded 12–16x forward earnings during upcycle phases with strong balance sheets. SanDisk's net cash position of $2.6 billion rising to $5.9 billion by FY2029, combined with an active share buyback programme enabled by peak FCF, creates a compounding per-share earnings dynamic that the current multiple does not price in at all. We believe the market is applying an implicit 'NAND discount' — pricing the company as if FY2026 earnings are unsustainable and will revert sharply. Our forecasts do include cyclical normalisation in FY2027–2028. But even at the trough of our modelled cycle ($38.80 EPS in FY2027), the stock at current prices trades at only 18x — a multiple that is fair, not expensive, for a business with $3.4 billion in net cash and demonstrated multi-year contract visibility. 5. **Free Cash Flow Generation Enables a Capital Return Story That Has Not Been Priced In** SanDisk is projected to generate $5.2 billion in free cash flow in FY2026, declining to $2.5 billion in FY2027 and $1.9 billion in FY2028 before recovering to $2.9 billion in FY2029. Over the four-year forecast period, cumulative free cash flow generation approaches $12.5 billion — nearly 18% of the current market capitalisation, assuming a share count consistent with the EPS data. This is a significant capital return opportunity. Management has indicated intent to deploy FCF toward debt reduction and share buybacks. The share buyback component is particularly significant for per-share earnings: if the company executes material repurchases during the FY2026 peak FCF year, the diluted share count reduction will lift EPS in FY2027 and beyond even if aggregate net income is flat or declining. The net cash position expanding from $2.6 billion to $5.9 billion over the forecast period indicates that debt reduction is occurring alongside buybacks — a conservative capital allocation strategy that reduces financial risk while improving per-share metrics. This FCF profile also provides management with strategic flexibility. If a NAND downcycle does materialise in FY2027–2028 more severely than our base case assumes, the company enters it with $3–4 billion in net cash — a meaningfully stronger position than it held entering the FY2022–2024 downturn. 6. **Q3 FY2026 Earnings Are the Definitive Near-Term Catalyst** The single most important near-term data point for SanDisk is the Q3 FY2026 earnings release, expected in late April or early May 2026. Management has guided to $4.4–$4.8 billion in revenue (midpoint $4.6 billion) and non-GAAP EPS of $12–$14 for the quarter — implying a quarterly non-GAAP EPS run-rate annualising to $48–$56. The critical variable is not whether the company hits the midpoint; it is whether management upgrades full-year guidance and signals continued multi-year agreement signings with hyperscalers. A beat-and-raise on Q3 results would force a fundamental reassessment of the earnings trajectory and likely compress the multiple discount the market has been applying. Specifically, any signal that FY2026 full-year EPS is tracking above $50 would demonstrate that the current 14x multiple is implying a more severe normalisation than even a moderate cyclical scenario justifies. Conversely, any disappointment — particularly on gross margin guidance for Q4 FY2026 — would validate the market's caution and likely push the stock lower. The binary nature of this catalyst is exactly what creates the investment opportunity: the market is pricing a miss, and the operational evidence suggests delivery.

Risks

1. **NAND Oversupply Cycle Risk: The Industry's Structural Flaw Has Not Been Eliminated** The NAND industry has undergone four major oversupply cycles in the past fifteen years. Each cycle ended the same way: producers, incentivised by high pricing during the upcycle, commissioned new capacity that arrived 18–24 months later into a market that had already begun normalising. There is no mechanism that permanently eliminates this dynamic. If SK Hynix, Micron, and Kioxia all make aggressive capacity addition decisions in late FY2026 based on current pricing signals, supply growth could overwhelm demand growth by FY2028 in a more severe fashion than our base case assumes. Our FY2027 revenue forecast of $14.5 billion already embeds a step-down from FY2026 peak — but a severe oversupply scenario could see pricing fall more than modelled, compressing margins toward 40% and cutting EPS to $15–20 rather than the $38.80 we forecast. 2. **AI Infrastructure Capex Concentration Risk** SanDisk's FY2026 thesis is substantially dependent on hyperscaler AI infrastructure spend remaining at current elevated levels. Microsoft, Google, Amazon, and Meta collectively represent the majority of enterprise SSD demand growth. If any of these companies materially reduces AI infrastructure capex — whether due to regulatory pressure, model efficiency improvements reducing storage requirements, or macro-driven budget discipline — the demand support for elevated NAND pricing evaporates rapidly. Multi-year contracts provide some protection, but prepayment structures and volume guarantees typically contain force majeure and technology substitution clauses that could allow customers to renegotiate. The concentration of demand in a small number of counterparties is a structural risk that cannot be diversified away. 3. **BICS 8 Yield and Qualification Execution Risk** The margin expansion thesis is predicated on BICS 8 achieving and sustaining high yields across the Yokkaichi and Kitakami joint venture facilities, and on hyperscaler qualification completion proceeding on schedule. NAND manufacturing at leading-edge nodes is extraordinarily complex — even minor contamination events, lithography yield excursions, or design-rule issues at 2Tb die can cause significant production losses. If BICS 8 yield falls below plan, unit cost reductions will be slower than modelled, and the gross margin expansion from 30% in FY2025 to 60%+ in FY2026 could prove optimistic by 500–1000 basis points. Additionally, enterprise SSD qualification at hyperscalers is a customer-controlled process — delays in customer qualification completions would defer revenue recognition on contracted volumes. 4. **Joint Venture Governance and Kioxia Relationship Risk** SanDisk manufactures NAND exclusively through its joint ventures with Kioxia, a relationship that creates significant operational dependency. Any deterioration in the JV relationship — including disagreements on capital allocation for capacity expansion, technology roadmap direction, or cost-sharing arrangements — could impair SanDisk's ability to execute its production plans. Additionally, Kioxia is an independently managed Japanese corporation with its own stakeholders and strategic priorities. If Kioxia pursues an IPO or third-party strategic transaction that changes the JV governance structure, SanDisk could face renegotiation of economic terms that are currently embedded in our cost and margin assumptions. 5. **Balance Sheet and Leverage Risk Entering the Cycle Peak** Despite the improving net cash position forecast ($2.6 billion in FY2026 rising to $5.9 billion in FY2029), SanDisk emerged from the FY2023–2025 downcycle with meaningful debt obligations and a $1.6 billion net loss in FY2025. The company's ability to execute the capital return programme — particularly share buybacks — is contingent on FCF generation materialising as forecast. If FY2026 FCF of $5.2 billion is reduced by 30–40% due to lower-than-guided pricing or higher capital expenditure requirements for the Kitakami ramp, the buyback programme could be curtailed or delayed, removing a key per-share earnings catalyst. 6. **Geopolitical and Supply Chain Risk in Japanese Manufacturing** All of SanDisk's NAND manufacturing is concentrated in Japan through the Yokkaichi and Kitakami JV facilities. This creates geographic concentration risk across several dimensions: natural disaster exposure (Japan is seismically active and the Yokkaichi facility has experienced earthquake-related disruptions historically), currency translation risk as yen-denominated costs interact with dollar-denominated revenue, and potential geopolitical risk if US-Japan-China trade dynamics deteriorate in ways that affect semiconductor equipment imports or technology transfer restrictions. Any unplanned production disruption at either facility — even a 30-day outage — would have a material impact on quarterly revenue and gross margins given the concentrated manufacturing footprint.

📈 Price Targets