Snapshot
- Ticker: SNDK (Nasdaq, spun from Western Digital February 21, 2025)
- Bucket: Storage & Memory
- Q4 2025 fund position: $250.2 M (1.05 M sh) — 4.53 % of 13F, #9 holding
- Fiscal year: ends late June (inherited from WDC)
- HQ: Milpitas, CA
Business Overview & Spin Context
Sandisk is the NAND flash and SSD pure-play spun out of Western Digital on February 21, 2025 (1:3 distribution to WDC shareholders). The spin completed a separation that had been logical for years: WDC’s HDD business and Flash business have very different capex requirements, customer mixes, and cyclicality.
Sandisk holds a long-standing 50/50 fab joint venture with Kioxia (the former Toshiba Memory) operating multiple 3D NAND fabs at Yokkaichi and Kitakami in Japan, plus its own controller/firmware engineering and a strong consumer brand (SanDisk-branded SSDs, memory cards, USB).
Following the spin, Sandisk is a clean exposure to:
- Client SSDs — PC and laptop OEM
- Consumer storage — cards, USB, portable SSDs (the SanDisk retail brand)
- Mobile NAND — UFS for smartphones
- Enterprise SSDs — data-center capacity drives, high-performance NVMe (the AI-relevant segment)
Financial Trajectory (combined Flash segment / pro-forma WDC era)
| Metric (USD M) | FY22 (Jun-22, WDC Flash) | FY23 | FY24 | FY25 (Jun-25, first as Sandisk) |
|---|---|---|---|---|
| Revenue | 9,750 | 6,320 (−35 %, NAND down-cycle) | 6,660 (+5 %) | 7,400–7,600 (+12–14 %) |
| Gross margin | — | trough negative | recovering | ~30 % non-GAAP by F4Q’25 |
| Operating margin | — | negative | mixed | swung positive FY25 |
| Net income | — | loss | mixed | modest profit FY25 |
| Adj. EBITDA | — | — | — | 1,200–1,500 FY25 |
Quarterly Trajectory
Recent quarters: ~1.9 B → $2.0 B+ trajectory through CY 2025 as NAND ASP recovered and bit shipments grew.
Cash Flow & Balance Sheet
- Capex: Shared with Kioxia at the JV fabs (Yokkaichi/Kitakami); Sandisk standalone capex ~$0.6–1.0 B
- FCF: Positive in upcycle FY25
- Total debt (allocated at spin): ~$2.0–2.5 B
- Net debt / EBITDA: 1.5–2× — manageable
Segment / Product Mix
| Segment | % of revenue | Notes |
|---|---|---|
| Client SSD | ~30–35 % | PC OEM (Dell, HP, Lenovo) |
| Consumer (retail) | ~25–30 % | SanDisk-branded — high-margin retail |
| Mobile NAND (eMMC/UFS) | ~15–20 % | Smartphones |
| Enterprise SSD | ~15–20 % | The AI-relevant segment, smallest leg |
Enterprise SSD is historically the weakest leg for the Sandisk/Kioxia consortium vs. Samsung and Solidigm. This is both the bull and bear case: bull = room to grow share; bear = competitive disadvantage.
NAND Cycle Context
Three structural shifts make NAND interesting in the AI cycle:
- Inference-time KV-cache offload to fast NAND — long-context inference creates pressure to swap KV state out of HBM and onto fast NAND tiers (NVMe-attached or CXL-attached). Per-token economics favor cheaper-but-fast storage where latency permits.
- Training dataset persistence at multi-PB scale — frontier labs hoard multi-petabyte training datasets that need fast NAND for repeated epoch reads.
- Supply discipline — NAND is a commodity oligopoly (Samsung, Kioxia/Sandisk, SK Hynix, Micron, YMTC). After the 2023–2024 down-cycle, surviving suppliers cut capex sharply. Prices recovered in 2025.
Operational KPIs
- Bit shipments: Mid-teens % YoY growth FY25
- ASP/bit: Up double-digit % YoY in FY25 on supply discipline
- Capex: Disciplined; BiCS8 (218-layer) ramping; BiCS9 in development
- Kioxia JV: 50/50 wafer split; node transitions are the key cost-down lever
Why It Fits the Thesis
Two AI-relevant pressures favor NAND structurally (above), and the fund’s choice of SNDK over Micron (which it was actually short via puts in Q3 2025) reflects a view that NAND is the better-positioned memory class vs. DRAM/HBM, where consensus is already crowded.
Specifically:
- The Q3 2025 MU put position (closed by Q4) signals the fund is negative on DRAM/HBM near-term (consensus too bullish, supply ramping fast).
- The SNDK long signals positive on NAND (less crowded, structural inference-tier demand, supply discipline).
The bet is therefore not on memory broadly but on the NAND vs. DRAM relative call.
Position History in the Fund
| Quarter | Position |
|---|---|
| Q3 2025 | Held WDC + initial SNDK exposure (post-spin) |
| Q4 2025 | Rotated entirely into SNDK at 1.05 M sh; exited residual WDC |
Risks
- NAND pricing volatility — commodity exposure remains.
- Kioxia JV dependency for fab capacity and capex decisions.
- HBM cannibalization if AI inference architectures shift more aggressively to DRAM-based caching.
- Enterprise SSD competitive disadvantage vs. Samsung and Solidigm — slower share gain in the highest-margin AI workload.
- Limited stand-alone history — first stand-alone 10-K only covers a stub period; less data for analysts to anchor on.
Sources
- WDC FY24 10-K (pre-spin)
- Sandisk Form 10 / spin filings
- Sandisk first 10-Q post-spin
- Sandisk investor day post-spin (February 2025)