SNDK

SNDK — SanDisk

TL;DR

SanDisk (spun off from Western Digital in 2024) is the pure-play publicly traded NAND company in the same supply tightness cycle as Micron. At ISSCC 2026, the SanDisk/Kioxia BiCS10 NAND (332 layers, 37.6 Gb/mm²) claimed the highest reported NAND bit density — dethroning SK Hynix’s 321L V9. Irrational Analysis rates SanDisk and Kioxia “tied for first place” on NAND technology. The thesis mirrors MU but is NAND-only: enterprise SSD demand exceeds supply, LTAs are locking in multi-year pricing floors, and new-node transitions create supply constraints at precisely the moment datacenter demand is inflecting.¹²³

Business

NAND flash manufacturer, spun off from Western Digital in February 2024. Operates through a joint venture with Kioxia (50% ownership) at Yokkaichi and Kitakami fabs in Japan. Product lines: NAND flash chips, enterprise SSDs (data center), client SSDs (PC/laptop), and consumer flash (SD cards, USB). Enterprise SSD is the high-margin, fast-growing segment. NAND technology roadmap follows the BiCS (Bit Cost Scalable) naming: Gen8 (BiCS8) → BiCS9 → BiCS10 in development.¹⁴

Thesis

• BiCS10 leading-edge density — first mover at 332 layers. At ISSCC 2026, SanDisk/Kioxia demonstrated BiCS10 with 332 layers in 3 decks, achieving 37.6 Gb/mm² — the highest reported NAND bit density at time of disclosure, above SK Hynix’s 321L V9. Key technical differentiators: (1) 6-plane configuration (1×6 approach, vs. SK Hynix’s 2×3) adds 50% IO bandwidth; (2) CBA (Cell Bonded Array) architecture enables CMOS customization, including dual top-metal layers for stronger power distribution; (3) die-gating cuts unselected-die idle current by two orders of magnitude. Lower cost-per-bit at BiCS10 = gross margin tailwind as mix shifts.¹
• Enterprise SSD demand exceeding supply. Datacenter/enterprise SSDs are SanDisk’s highest-margin segment and are operating at record levels with demand exceeding supply. The AI server buildout (storage + boot SSDs in GPU racks, AI training checkpoint storage, inference KV-cache SSDs) is driving structural demand growth that didn’t exist in prior NAND cycles.²³
• LTA pricing floors — multi-year visibility. FundaAI: new LTA negotiations now run 3-5 years (vs. one year historically), require prepayments, and are anchoring “bottom price” floors at current elevated Q2 levels (up 80-100% from prior). SanDisk explicitly used as a benchmark for the new vendor pricing power regime. If floors hold, SNDK EPS has structural support through 2027-2028.²
• Kioxia JV economics improving for SNDK. The Kioxia-SanDisk JV terms were renegotiated: Kioxia receives $1.165BfromSanDiskover2026-2029($290M/year) and recognizes ~$131M/year of margin-accretive revenue from Feb 2026-Dec 2034. This reframes what had historically been throughput economics into explicit consideration — reduces the risk of future JV disputes and puts a longer-term floor on the JV’s contribution to SanDisk output.⁴
• Cycle read-through from MU. FundaAI’s SNDK EPS estimates were revised up alongside Micron’s post-earnings beat (March 2026). The two are in the same NAND cycle; MU’s revenue and margin trajectory is the leading indicator for SNDK’s own NAND pricing environment.³

Risks

• NAND cycle normalization. Unlike DRAM/HBM (where AI directly consumes capacity), NAND is more exposed to consumer/PC demand softness. If enterprise SSD demand plateaus and consumer demand remains weak, NAND can cycle down faster than DRAM.
• Kioxia JV execution. SanDisk has minority control over the JV’s production decisions. Any yield problem at BiCS10, fab downtime, or Kioxia capital allocation shift could affect SanDisk bit output without SanDisk being able to course-correct unilaterally.
• LTA breach risk. In prior cycles, customers defaulted on LTAs when spot prices fell. The new prepayment and penalty structures reduce this risk but don’t eliminate it entirely; if a major hyperscaler disputes or restructures an LTA, pricing power erodes quickly.
• No HBM exposure. SNDK is pure NAND. It has no DRAM or HBM products — the highest-margin, fastest-growing memory category. The upside of the memory supercycle is more concentrated in MU (which has DRAM + HBM + NAND) than in SNDK.
• Accounting transition. Ridire Research flags accounting complexity in the Kioxia JV revenue treatment as a risk to earnings quality visibility.

Recent catalysts

• 2026-04-14 — Edgewater Research (NEUTRAL, $952.50;STMixed/LTPositive):FY26EPS$48.52 (vs. consensus $20.09);FY27consensusEPS$96.92 (Edgewater estimates below consensus). Model: Rev $7.4BFY25\to$8.2B FY26e → ~$17.1B FY27e; GM 10.2% → ~30% → ~77%. SNDK noted as most aggressive in C1Q/C2Q pricing across all NAND solutions. May have signed at least one additional multi-year supply agreement in C1Q26.⁵
• 2026-04-14 — Cleveland Research (NEUTRAL, $391.72): Forecasts moving higher on further NAND ASP upgrades and strengthening eSSD demand. NAND ASPs higher in CQ2 & CQ3 with server demand expected to uptick in CQ3 on KV Cache & VR ramp up. Hears SNDK LTAs driving scarcity premiums vs. peers for renewing eSSD contracts.⁶
• 2026-04-15 — ISSCC 2026: SanDisk/Kioxia BiCS10 NAND disclosed — 332 layers, 37.6 Gb/mm², highest bit density reported; 6-plane 1×6 configuration, CBA architecture, die-gating.¹
• 2026-04-01 — FundaAI: LTA structure deep dive — SanDisk cited as pricing floor benchmark; 3-5 year terms with prepayments and downside locks at Q2 price levels.²
• 2026-03-19 — FundaAI: SNDK EPS revised up alongside MU post-earnings; NAND supply tightness to persist beyond 2026.³
• 2026-03-05 — Irrational Analysis: “Sandisk and Kioxia are tied for first place” on NAND technology; SNDK rated as best NAND equity in author’s framework.⁷
• 2026-03-02 — Ridire Research: Kioxia Q3 FY2025 deep dive — JV economics rewritten ($1.165B consideration 2026-2029); enterprise SSD 60% of Kioxia SSD revenue at record highs.⁴

Second-order reads

• 2026-04-14 — Edgewater Research, Memory/Storage Update — SNDK viewed as most aggressive in pricing; near-term pricing risk if hyperscalers push back on Q/Q increases; ST/LT mixed signals for SNDK vs. MU (which remains LT positive). → SNDK’s price aggressiveness could be a double-edged sword if customers delay orders.⁵
• 2026-03-19 — FundaAI, Review MU — MU NAND supply shortage persists; both MU and SNDK benefit from same datacenter SSD demand wave → MU beat is the leading read-through indicator for SNDK EPS trajectory.
• 2026-04-15 — SemiAnalysis, ISSCC 2026 — BiCS10 leading SK Hynix V9 on density → competitive implication for SK Hynix NAND market share; if BiCS10 yields hold at scale, SNDK/Kioxia gain cost-per-bit advantage over SK Hynix in enterprise SSD.

Valuation & positioning

Edgewater (NEUTRAL, $952.50):FY26EPS$48.52 vs. consensus $20.09;FY27EPSvs.consensus$96.92 (Edgewater below consensus on FY27). Model revenue: $7.4BFY25\to$8.2B FY26e → $17.1BFY27e;GM:10.2$391.72): FY26 Rev $17.14B(+132$32.4B (+47% Y/Y); FY27 EPS $108.79. Different fiscal year conventions between firms — Cleveland’s FY26 appears to run through mid-2027.⁶ Irrational Analysis considers SNDK and Kioxia the top two NAND equities but does not provide a specific multiple target in available corpus.⁷

Sources

Related

MU — NAND/DRAM cycle companion; MU has DRAM + HBM exposure that SNDK lacks; MU beat is the leading indicator for SNDK pricing

Footnotes

1. SemiAnalysis — ISSCC 2026: Nvidia and Broadcom CPO — 2026-04-15 ↩ ↩² ↩³ ↩⁴

2. FundaAI — Deep Memory: Sustaining the Supercycle — 2026-04-01 ↩ ↩² ↩³ ↩⁴

3. FundaAI — Review MU: Results and Guidance Crushes Expectations; SNDK EPS Update — 2026-03-19 ↩ ↩² ↩³ ↩⁴

4. Ridire Research — Not Your Last NAND Cycle — 2026-03-02 ↩ ↩² ↩³

5. SNDK) Update — 2026-04-14 ↩ ↩²

6. NAND Update — 2026-04-14 ↩ ↩²

7. Irrational Analysis — March Morgan Stanley TMT Madness — 2026-03-05 ↩ ↩²