Sandisk
The NAND Pure-Play, Unbundled

Snapshot: Executive Summary

Sandisk Corporation began trading as an independent public company on February 24, 2025 following the spin from Western Digital¹. The transaction separated WDC's NAND-flash business (Sandisk) from its HDD business (WDC standalone), creating two cleaner public-market vehicles for distinct cycle exposures. Sandisk takes ~14% of global NAND market share, fourth behind Samsung (~33%), SK Hynix (~22%), and Kioxia (~17%) but ahead of Micron (~12%), and operates a 50/50 production JV with Kioxia in Yokkaichi, Japan that has been a structural feature of both companies for two decades.

The setup:

• The NAND cycle went parabolic. The 2023 trough saw industry-wide losses; 2024 brought production discipline; through 2025-2026 an AI-driven memory supercycle sent ASPs, revenue and margins to records. Q4 FY26 revenue hit a record $8.97B (+372% YoY, +51% QoQ), gross margin 84.6%, non-GAAP EPS $39.25; full-year FY26 revenue reached $20.25B (+175% YoY). HBM allocations at Samsung / SK Hynix cap non-AI wafer capacity, tightening NAND further.
• The New Business Model locks in visibility. Five multiyear supply agreements secure ~$42B of minimum contractual revenue backed by >$11B of guarantees, converting a spot-priced NAND line into a partially contracted one. Datacenter revenue jumped another +103% QoQ to $2.98B, roughly a third of the mix (up from ~25% in Q3, and 12% a year ago). But the below-consensus Q1 FY27 guide is the first signal the cycle is cresting.
• But this is a cyclical name near a cyclical peak. After ~+2,300% off the spin, the stock has already pulled back ~48% from its $2,354 peak (and ~25% in the past month) as the memory trade cooled, and the down-cycle is the asymmetric risk. The forward P/E (~8x) is cheap only if the peak holds.

At ~6.5x NTM EV/Sales (a multiple inflated by peak gross margins near 85%), SNDK has compressed on the pullback to roughly in line with MU (~6.5x), after trading above it at the July peak. The truer cyclical signal is the ~8x forward P/E on peak EPS. We stay Hold and re-rate the 12-month PT to $1,350 (cut from $1,750) via EV/Sales scenarios (Bull $1,700 / Bear $700), for coherence with the ~$1,212 spot. Probability-weighted blended fair value ~$1,275, modestly above spot. NAND is highly cyclical; after this run, own the backlog and respect the cycle. Position size accordingly.

Tactical: SNDK is trading at ~$1,212, ~10% below our $1,350 12-month target. The $42B locked-in backlog and the record Q4 FY26 print ($8.97B revenue, $39.25 EPS) underpin the base, but the below-consensus Q1 FY27 guide is the first crack in the peak and the multiple already discounts the cycle (forward P/E ~8x). The blended fair value ($1,275) sits modestly above spot. Rating Hold; we would turn more constructive on a deeper pullback that widens the distance to the bull case, or more cautious on the first sign NAND ASPs roll over.

Investment Thesis

Rating: Hold. Probability weights: 25% Bull / 50% Base / 25% Bear → blended fair value ~$1,275, modestly above spot, appropriate for a cyclical name near a peak. We pair this Hold with the MU memory dashboard for two-way memory exposure (MU more diversified across DRAM + HBM + NAND; SNDK pure NAND). The pair gives a complete memory-cycle picture.

Business Overview

Sandisk designs, develops, and sells NAND flash memory products across the full stack: client SSDs (consumer + commercial PC), enterprise SSDs (high-capacity datacenter drives), mobile / embedded (eMMC and UFS for smartphones), and removable storage (memory cards). The company does not operate its own fabs at scale; instead, Sandisk operates a 50/50 production joint venture with Kioxia (the former Toshiba Memory) at the Yokkaichi and Kitakami facilities in Japan. Sandisk takes its allocated share of wafer output and sells finished products separately from Kioxia, the JV is purely upstream manufacturing, not commercial.

Market position

The Kioxia JV: structural feature, not a risk

Sandisk and Kioxia have operated a 50/50 NAND production JV since 1999. The structure: shared capital expenditure, shared manufacturing capacity, separate commercial operations. Sandisk takes 50% of wafer output, sells through its own brand and channels. The JV has survived multiple ownership transitions (Toshiba → Toshiba Memory → Bain consortium → Kioxia Holdings) and is unlikely to dissolve given the deep operational integration. Periodic merger discussions between SNDK and Kioxia have surfaced (most recently 2021) but never consummated; the standalone-public structure makes a future merger easier to mechanically execute if commercial logic ever aligns.

End-market exposure

• Mobile / smartphone: Apple is the largest single customer (multi-billion-dollar relationship); Samsung handsets second-largest. Sensitive to smartphone unit cycles.
• PC client: OEM bundling (Dell, HP, Lenovo) plus retail / channel for upgrades. Tied to PC unit demand.
• Enterprise / datacenter: High-capacity SSDs (61TB and 122TB drives) for AI inference and capacity storage. Growing fastest; Sandisk competes but Samsung leads.
• Industrial / automotive: Smaller but higher-margin. Long lifecycles, slower design wins.
• Removable: Memory cards and USB drives. Declining tail business; high gross margin.

NAND Cycle Dynamics

NAND is the most cyclical commodity in semiconductors. The 2023 trough, driven by post-COVID demand softness, aggressive 2021-2022 capacity adds, and a smartphone unit recession, produced industry-wide losses. What followed was not just a recovery but a supercycle: through 2025-2026, (a) production discipline from all majors, (b) HBM allocations at Samsung / SK Hynix capping non-AI wafer capacity, and (c) a step-change in AI-datacenter storage demand combined to send NAND ASPs, revenue and margins to records. The open question is no longer whether the cycle recovers but how long the peak lasts before supply catches up.

Demand drivers: where the bytes are going

What the bull case actually requires

• Supply discipline through 2026. No major capacity adds from Samsung or SK Hynix.
• Enterprise SSD share gains for Sandisk. The Bain / Solidigm consolidation gave SK Hynix the leadership position there; Sandisk needs to capture incremental share in the 61TB+ drive segment.
• Smartphone unit stability. A weak smartphone cycle compresses NAND mobile demand by ~20% directly.
• AI inference at edge actually shows up. Bull-case incremental demand assumes on-device LLM inference materially expands NAND storage requirements, this is the speculative tail.

The bear scenario in detail

Samsung has the largest installed NAND capacity globally and the strongest balance sheet to absorb cycle losses. Historically, in NAND down cycles, Samsung's pattern has been to grab share via price (not discipline). If Samsung shifts strategy in CY2026, for example, signaling NAND capex acceleration in their Q4 2025 print, NAND ASPs would compress 20-25% and Sandisk's FY27 revenue + gross margin profile would deteriorate materially.

The WDC Spin & Standalone Structure

The February 21, 2025 spin separated Western Digital into two independent public companies: WDC (HDD only) and Sandisk (NAND only). The transaction was the culmination of a multi-year activist campaign (Elliott Management was the most prominent voice) arguing that the WDC conglomerate structure obscured the value of both businesses and that two pure-play vehicles would each command higher multiples than the blended whole.

Transaction mechanics

Why now

The spin was on the table for nearly five years. The proximate trigger was the combination of (a) Elliott's escalation in 2024, (b) the NAND cycle reaching a recovery point that would let Sandisk debut as a profitable standalone, and (c) WDC management succession that made the structural change politically easier internally. The strategic rationale: as pure-plays, both companies trade on the metrics that matter for their respective end markets, Sandisk on NAND-cycle EV/Sales, WDC on HDD-stability EV/EBITDA. The blended WDC pre-spin traded at a conglomerate discount that the standalone vehicles can recapture.

What operationally changes (and doesn't)

• What changes: Sandisk has its own board, its own capital allocation framework, its own investor relations. Management is no longer competing internally for capital with the HDD business. Reporting cadence and disclosure are tailored to the NAND pure-play comp set.
• What stays the same: The Kioxia JV is unaffected, Sandisk's 50% interest transferred whole. Customer relationships continue uninterrupted (Apple, Samsung handsets, OEM channel). Manufacturing capacity and product roadmaps are unchanged in the near term.
• What's incremental: Modest standalone-company costs (~$30-50M annualized) for board governance, regulatory compliance, separate audit. Largely offset by reduced internal capital-allocation overhead.

The re-rate question, answered

The spin thesis was that a standalone NAND pure-play would recapture the conglomerate discount. The memory supercycle overwhelmed that debate: SNDK re-rated from ~1x EV/Sales at-spin to a ~9x peak (now ~6.5x after the August pullback), and the market value went from ~$10B to a ~$350B peak, now ~$180B. The pure-play structure delivered exactly what the spin promised, clean exposure to the NAND cycle, and that cycle then ran to a historic peak. The forward question is no longer whether the multiple expands but whether peak-cycle sales and margins hold; the standalone story is fully priced.

Financial Health & Trends

Q4 FY26 print highlights (record quarter, reported August 5, 2026)

The Q4 print was a record on every line: revenue rose another +51% sequentially (+372% YoY) to $8.97B, gross margin cleared 84.6%, non-GAAP EPS hit $39.25, and datacenter revenue jumped +103% to roughly a third of the mix (its growth ~1/3 volume, ~2/3 pricing). Full-year FY26 revenue was $20.25B (+175% YoY). Under the New Business Model, five multiyear supply agreements lock in ~$42B of minimum contractual revenue backed by >$11B of guarantees. But the below-consensus Q1 FY27 guide is the first sign the peak is cresting, exactly the cyclical-peak signal a Hold respects.

Capital Allocation

Sandisk emerged from the spin net-debt-neutral (~$1.5B cash, ~$1.5B debt). The supercycle changed the picture: adjusted free cash flow was $2.955B in Q3 FY26 alone and the record Q4 extended the cash build, so the company sits in a net-cash position (~$3B and building) while funding the Kioxia JV capex. Capital allocation priorities: (1) fund the JV capex contribution (rising with the cycle); (2) retire the inherited convert (now deep in the money); (3) begin returning capital as the supercycle throws off cash faster than it can be reinvested.

The convertible bond is now deep in the money

$1.5B of convertible senior notes due 2028 were inherited from the pre-spin WDC capital structure, with a conversion price around $70 per SNDK share. At ~$1,212 the converts are ~17x in the money and will convert, the incremental shares are already reflected in the diluted count. With supercycle cash flow, retiring or cash-settling the notes is straightforward; the historical "overhang" framing no longer applies.

JV capex

Sandisk's annual share of the Kioxia JV capex runs ~$1B-$1.2B depending on cycle position. This is the largest single capital commitment and is structurally inflexible, once a fab capacity expansion is contractually committed, Sandisk must fund its 50% share regardless of cycle conditions. In NAND downturns this can pressure FCF; in upturns the operating leverage is meaningful.

Capital return policy

With supercycle cash flow, capital return is now a question of when and how much, not whether. Our base case assumes a sizable buyback authorization as the company generates cash faster than the JV can absorb it. A fixed dividend remains less likely near-term: NAND cyclicality argues against committing to a payout set at peak earnings, since the down-cycle would make it hard to sustain. Buybacks (flexible, opportunistic) fit a cyclical far better than a dividend.

Valuation Overview

Why EV/Sales over P/E for SNDK, and why the forward P/E matters more here

NAND is cyclical enough that P/E is uninformative at cycle extremes (negative earnings in trough years; explosive earnings in peak years). EV/Sales smooths that, but at a cyclical peak it does the opposite of what it does at a trough: peak gross margins (85%) inflate the multiple, so ~6.5x EV/Sales looks full even though the forward P/E is only ~8x. For a peak memory name the forward P/E is the more honest signal, and ~8x forward earnings is cheap only if the peak holds. That asymmetry is the core of the Hold.

The MU comparison, inverted

SNDK briefly traded above MU's EV/Sales multiple at the July peak; the August pullback closed that gap, and the two now screen roughly in line. The premium existed at all because of margin mix: at the peak, SNDK's pure-NAND margins run hotter than MU's blended DRAM+HBM+NAND book, so per-dollar-of-sales SNDK screened richer. That was never a durable premium, it was a peak-cycle artifact. As margins normalize off 85%, the EV/Sales multiple compresses even if the share price holds, which is why we lean on scenario weights rather than a single multiple.

EV/Sales Scenario Model

FY27E revenue × EV/Sales multiple = enterprise value, plus net cash, divided by diluted shares = forward fair value. For this cyclical peak we take the 12-month PT equal to the forward fair value (no separate discount); the low multiple already prices the cycle.

Sensitivity grid: FY27E revenue × EV/Sales multiple (PT in $)

Quick PT calculator

Risk / Reward calculator

Upcoming Catalysts

Risk Factors

• The cycle rolls over. The dominant risk, and after a +2,300% run the asymmetric one. When supply catches up (Samsung/SK Hynix adding capacity is their historical pattern), NAND ASPs fall 25-40%, gross margin normalizes off 85%, and peak EPS collapses. The multiple compresses even if volumes hold, driving the $700 bear case.
• Post-parabola volatility. The stock has already pulled back ~48% from its $2,354 peak (and ~25% in the past month) and swings violently on cycle sentiment. Even with an intact thesis, the mark-to-market risk is large.
• Backlog cushions volume, not price. The $42B New Business Model backlog secures minimum volumes but does not fully insulate ASPs; a price down-cycle still compresses margins and the multiple.
• Smartphone unit weakness. Mobile / embedded is a meaningful share of revenue. A weak iPhone cycle or China smartphone weakness compresses NAND demand directly.
• Kioxia JV restructuring. Any change to the 50/50 production JV (capex allocation, output split, ownership transfer) could materially shift Sandisk's cost structure or supply position. Low probability but high severity.
• Concentration and comparability. Datacenter is now ~a third of revenue and growing fast, concentrating exposure to a handful of large buyers; and the supercycle's YoY comps will be brutal on the way down.
• JV partner (Kioxia) financial health. Kioxia is privately held (Bain-led consortium ownership). Any financial stress at Kioxia could affect JV capex commitments and operating cadence.

Bull vs Bear Debate

Technical Analysis

Trader's view

• 52-week range $40.10-$2,354.39; now ~$1,212 after a ~48% pullback off the $2,354 peak.
• Key support: ~$1,200 (the spring base), now being tested. A decisive break opens a deeper de-rating as the cycle-peak trade unwinds.
• Key resistance: the $2,354 ATH. Reclaiming it requires a fresh leg of the supercycle, not just mean-reversion.
• R/R from ~$1,212 to a $1,700 target with a ~$1,000 stop ≈ 2.3:1. Improved by the pullback, but still a Hold near a cresting peak.
• Post-parabola tape: momentum indicators whipsaw and gaps are common. Size positions for the volatility, not the trend.