SNDK — Sandisk Corporation

▸ The full interactive deep dive — interactive price chart (updated weekly), scores, fair value, and the expert-claim panel: research.synthosresearch.com/SNDK

Theme: Compute (memory/NAND) | Sector: Technology / Semiconductors | Price: $2,032.22 | Market Cap: ~$301B Note date: 2026-07-02 | Fiscal year end: last Friday of June

1. What it is

Sandisk Corporation designs, manufactures, and supplies NAND flash-based storage solutions — SSDs, embedded memory, removable memory cards, USB drives, and NAND wafers/components — sold into PC, mobile, consumer, and enterprise/datacenter channels. Sandisk was spun off from Western Digital in February 2025 as a standalone pure-play NAND company; it is not a DRAM or HBM producer. Headquartered in Milpitas, CA, ~12,000 employees, CEO David Goeckeler.

2. The exponential thesis

The Synthos quant screen shows +168% forward revenue growth with slight positive acceleration, landing SNDK in the Compute/memory theme alongside DRAM/HBM names. The driver is real but narrower than the headline suggests: NAND flash is undergoing an acute, industry-wide supply-demand imbalance in 2025-2026 — datacenter SSD demand (driven by AI training/inference storage buildouts) plus a wave of supplier capacity discipline (Samsung, SK Hynix, Kioxia, Sandisk itself all cut bit-supply growth after years of losses) has flipped NAND from oversupply to shortage. FMP quarterly data confirms this concretely: quarterly revenue went from $2.31B (Q1 FY26, Sep-2025) → $3.03B (Q2 FY26) → $5.95B (Q3 FY26, Apr-2026), with gross margin exploding from ~30% (FY2025 full year) to an extraordinary ~78% in the most recent quarter (gross profit $4.66B on $5.95B revenue) as NAND contract/spot pricing spiked.

This is a genuine cyclical demand/pricing inflection, not a new product category or S-curve in the classic sense (no new addressable market is being created — the same bits are simply being repriced upward and volumes are recovering). It is directly analogous to prior NAND up-cycles (e.g., 2017, 2021) rather than a structural, durable growth vector like HBM's attach-rate-to-AI-accelerator story. This is the central caveat to the thesis and is addressed head-on in Section 7.

3. Financial health

Standalone annual history is short (spun off Feb 2025; FMP shows FY2021 as a placeholder/carve-out stub with mostly zeroed ratios) — four real annual data points (FY2022–FY2025, fiscal years ending late June):

Revenue fell 38% from FY2022 to the FY2023 trough, then only partially recovered through FY2025 — three consecutive annual net losses despite the FY2025 gross-margin recovery (net loss driven largely by a non-operating/other-expense charge, likely spin-off related). The turn is happening in FY2026 (not yet in annual filings): quarterly gross margin snapped from ~30% (FY25 average) to 78% by Q3 FY26, and Q3 FY26 alone produced $3.615B of net income (EPS $24.43) — more profit in one quarter than the prior three full fiscal years combined lost. This is a real, dramatic inflection, but it is a classic NAND boom quarter, not a gradual margin-expansion story.

Balance sheet (FY2025, most recent annual): cash + short-term investments $1.48B, total debt $2.04B, net debt $561M, total equity $9.22B — a manageable net-debt position, current ratio 3.6x, quick ratio 2.1x. Balance sheet strengthened materially vs FY2024 (cash was only $328M then). Cash flow: FY2025 operating cash flow was only +$84M and free cash flow was negative (-$120M) despite the reported profit recovery in gross margin — working capital consumption (inventory build, AR growth) ate the operating cash flow. FY2023 and FY2024 both had deeply negative operating cash flow (-$713M, -$309M) and free cash flow (-$932M, -$475M). No dividend, no buybacks in the FMP data — all "other financing activities" instead (likely spin-off-related debt/equity restructuring, not shareholder returns).

4. Competitive position / moat

NAND is an oligopoly (Samsung, SK Hynix/Solidigm, Kioxia, Micron, Sandisk — roughly 5 players controlling >90% of bit supply), which gives some pricing power in tight markets, but NAND itself is largely a commoditized, interchangeable product — competition is on cost-per-bit and process node execution, not durable product differentiation. Sandisk's moat is scale, decades of process IP (originally Toshiba/SanDisk joint ventures in Japan), and channel relationships built from being the historical #2 NAND supplier behind Samsung. It has no meaningful position in DRAM or HBM, so it does not participate in the AI-accelerator-attached memory story the way Micron or SK Hynix do — its AI leverage is indirect, via enterprise/datacenter SSD attach rates, not compute-adjacent memory content per GPU.

5. Valuation

Using FMP's own annual ratios (FY2025, based on then-prevailing ~$6.8B market cap in the ratios file — note this predates the ~$301B market cap in the current quote/profile, implying the stock has re-rated enormously since FY2025 close): P/S 0.93x, P/B 0.74x, EV/Sales ~1.0x — these were "cheap" multiples typical of a company in a loss-making trough. At the current $301B market cap and ~$2,032 share price, using run-rate annualized Q3 FY26 figures (~$24B revenue annualized, ~$14.5B annualized net income), the multiples look completely different — roughly 12-13x sales and ~20x annualized earnings on the best quarter of the cycle. The quant screen's PEG of 0.09 is a function of dividing a low forward P/E by an extreme (and almost certainly cyclically-inflated) growth rate; PEG is not a reliable valuation signal in a business this cyclical, because the "E" and "growth" inputs are both artifacts of a trough-to-peak swing, not steady-state earnings power.

The FMP forward estimates data itself is unreliable past FY2027 — the FY2029 estimate shows revenue jumping to $58.8B (net income $37.0B, i.e., net margin >60%, and net income exceeding EBITDA), and FY2030 shows revenue falling to $11.6B with net income of $1.96B — internally inconsistent, almost certainly analyst-count/data-integrity artifacts (FY2029/2030 estimates are built on only 2-6 analysts vs 10-14 for FY2026/2027) and should be disregarded entirely.

Valuation stance: rich, not cheap, despite the low headline PEG. The stock has already priced in the current NAND up-cycle peak (100x-plus run off the 52-week low of $40.10 to $2,032, per the quote file's yearLow/yearHigh range) — this is one of the more extreme cyclical re-ratings in the dataset. Buying here is a bet that peak-cycle NAND pricing is durable, which historically it has not been.

6. What the panel says

KB evidence found: yes — genuine, relevant matches, not a ticker collision. ~24 claims across the KB mention SanDisk/SNDK by name (grepped both spellings), overwhelmingly from analysts covering the memory/semiconductor space (jordi_visser / jordi_visser_ai / jordi_visser_m — appears to be the same commentator tracked under multiple source variants — and compound_and_friends).

• Bullish contingent (majority, conviction 75-88): Frames memory (Micron, SanDisk, SK Hynix) as a "5-year-plus super-cycle" and an "AI pick-and-shovel" scarcity trade — NAND/DRAM prices spiking, supply sold out, comparisons to Micron's 7x and Intel's 6x runs. One claim (2026-02-01) specifically notes SanDisk's EPS beat "~100% with huge guidance," consistent with the margin inflection seen in the FMP quarterly data.
• Bearish/skeptical contingent (conviction 55-72): Explicitly flags memory as "the most cyclical stocks" where low multiples reflect earnings that "swing wildly, not value" — directly echoing this note's valuation caution. Others flag the semiconductor index as "68% above its 200-day moving average, the most stretched in 25 years," call recent SanDisk price action "pockets of nonsense (+25% in a day)," and question whether the "memory-shortage melt-up" can go "straight up" while software signals weaken. One claim (2026-05-31) flags the key structural risk directly: "betting memory stays non-cyclical with no Chinese competition or algorithmic change is the key risk" once Micron/SanDisk/SK Hynix all crossed $1T combined market cap.
• Net panel read: Bullish-leaning but with unusually vocal, specific cyclicality skepticism embedded in the same source family — the panel itself is debating the exact tension this note raises, not uniformly bullish.

7. Key risks

1. NAND is a brutally cyclical commodity business, and this data shows it plainly. FY2022 profitable → FY2023/24/25 three straight annual net losses → FY2026 explosive rebound. Gross margin alone ranged from 7% (FY2023) to an estimated ~78% (latest quarter) in under three years. A snapback of that magnitude is a strong prior for a subsequent snap-down when supply discipline breaks or demand normalizes — this has happened in every prior NAND cycle (2017-18, 2021-22).
2. The "+168% forward growth" figure is very likely trough-to-peak cyclical snapback, not durable structural growth. FY2025 (the base year) was itself a loss-making, depressed year following two even worse years. Compounding off a depressed base mechanically inflates the growth percentage; nothing in the business model (no new product category, no attach-rate story comparable to HBM) suggests this rate is sustainable into FY2027+ once the pricing cycle rolls over. The FMP forward estimates data for FY2029-2030 are internally inconsistent (net income exceeding EBITDA, revenue reversing) and should not be relied upon for terminal-value work.
3. No DRAM/HBM exposure limits the durable, structural AI-buildout angle relative to peers (Micron, SK Hynix) that participate directly in HBM content-per-GPU growth — Sandisk's AI leverage is entirely via cyclical NAND/SSD pricing, the most commoditized and volatile part of the memory stack.

8. Verdict

Watch.

One-line thesis: Real, sharp NAND-cycle earnings inflection currently underway, but the stock's ~50x move off 52-week lows already prices in peak-cycle conditions in a historically boom-bust commodity business with a two-year annual loss streak directly behind it.

Valuation stance: rich for the growth — the low PEG is an artifact of cyclical trough-to-peak earnings math, not a genuine growth-adjusted discount.

Biggest risk: NAND pricing/supply-discipline reversal — the same cyclicality that produced three consecutive annual net losses (FY2023-FY2025) before this upswing is the base-rate outcome for this industry, and nothing in the current setup (no structural new-demand category, oligopoly supply discipline that has broken before) rules out a repeat.

Data-quality flags: - Standalone annual history is thin (real data starts FY2022; FY2021 is a mostly-zeroed carve-out stub) — a direct consequence of the Feb-2025 Western Digital spin-off, exactly as anticipated. Four real annual data points is not enough to establish a long-run structural trend independent of the cycle. - FMP estimates_annual data for FY2029 and FY2030 is internally inconsistent/unreliable (net income implausibly exceeds EBITDA/revenue relationship in FY2029; revenue reverses downward in FY2030 versus FY2029) and is built on very thin analyst counts (2-6 analysts) — excluded from valuation work above. - The ratios_annual/key_metrics_annual files reference a market cap (~$5.2-6.8B) far below the current quote/profile market cap (~$301B), reflecting how much the stock has re-rated since FY2025 close — cross-period multiple comparisons must account for this, which this note does in Section 5. - KB entity check: both "SNDK" and "Sandisk"/"SanDisk" spellings returned genuine, on-topic matches (memory/NAND semiconductor commentary) — no evidence of ticker collision or wrong-entity false positives.