SNDK
SanDisk Corporation
NAND Memory · AI Chips & Compute
$1,398.52
Mkt Cap ~$207B
52W: $35.79 – $1,600.00
Next earnings: ~Late Jul 2026 (Q4 FY26 · Est)
↓ Add (pullback)
Entry zone: $1,200–$1,350 · pullback only
The market still prices SNDK as a commodities story that will mean-revert — but Q3 FY26 ($5.95B revenue, 70.9% non-GAAP operating margin, Q4 guided $7.75–8.25B) documents supply getting tighter, not looser, and management's own estimate of 75–100 EB of incremental AI inference demand in 2027 is explicitly "none of that is in the numbers we're talking about."4
Fwd P/E (NTM)
~10–12×3
PEG ~0.04 (g=251% NTM) ⚠ cycle-peak g
Estimates repricing post-Q3
Rev YoY (Q3 FY26)
+251%1
QoQ ↑ · $5.95B
Gross Margin
78.4% Q3 FY26
+27.5pp QoQ · Q4 guide 79–81% ↑1
FCF Margin (TTM)
⚠ pending
Q2 FY26: $980M · 27.9%4
EPS Revision (90d)
↑ major
⚠ post-Q3 upward revision cycle beginning3
Next Earnings
Late Jul 2026
Q4 FY26 · Est
DC Rev %
~24.6%1
of Q3 FY26 total
DC Rev QoQ
+233%1
$1.47B Q3 FY26
Thesis Durability
A−
The market prices SNDK as a commodity cycle trade. The counter: BiCS8 NAND has become structurally embedded in hyperscaler AI inference deployments. Enterprise SSDs are the lowest-cost, high-bandwidth storage tier at inference scale — there is no cheaper substitution available for the workload.
Q3 FY26 validated the demand acceleration, not a peak: DC revenue $1.47B (+233% QoQ), with BiCS8 QLC "Stargate" products confirmed for volume shipments beginning June 2026 at the 128TB SKU (256TB following shortly after). Management stated customer demand will remain "well above supply beyond calendar year 2026" and confirmed that Q3 was more undersupplied than Q2.1 Management's own CY26 DC growth forecast has been revised upward in consecutive quarters: 20s (9 months ago) → 40s → 60s → mid-70s (Q3 print) — roughly a 4× increase over three-month increments, each revision higher than the last.5
The NBM framework is the structural pivot the market is misreading. SanDisk signed 5 NBM agreements (3 in FQ3 + 2 in early FQ4), covering >1/3 of FY2027 BiCS supply. Total RPO reached $42B — over a year of locked-in demand at the FQ4 guide midpoint. Customers committed $11B in financial guarantees (including prepayments and third-party-managed instruments). Critically, NBM pricing is not fixed: volume is committed while pricing remains variable, allowing SanDisk to capture upside if prices rise. Management is targeting >50% of supply under NBMs over coming quarters — this is not an HDD LTA analogy.5
The structural upside optionality: management's estimate of 75–100 EB of incremental AI inference NAND demand in 2027 — doubling again in 2028 — is explicitly not yet in any analyst model.4 At 75–100 EB, that represents 6–8% of the entire flash market as a new incremental demand source.
Bear case: Samsung resolves NAND yield issues in H2 2026 and re-enters as a credible supply competitor. Kioxia and SK Hynix have both signalled capacity expansion plans. Google's TurboQuant (6× KV cache compression) could reduce per-inference NAND demand — though management's read is that compression enables longer context windows and more concurrent requests, making the net NAND impact neutral to slightly positive.
Q3 FY26 validated the demand acceleration, not a peak: DC revenue $1.47B (+233% QoQ), with BiCS8 QLC "Stargate" products confirmed for volume shipments beginning June 2026 at the 128TB SKU (256TB following shortly after). Management stated customer demand will remain "well above supply beyond calendar year 2026" and confirmed that Q3 was more undersupplied than Q2.1 Management's own CY26 DC growth forecast has been revised upward in consecutive quarters: 20s (9 months ago) → 40s → 60s → mid-70s (Q3 print) — roughly a 4× increase over three-month increments, each revision higher than the last.5
The NBM framework is the structural pivot the market is misreading. SanDisk signed 5 NBM agreements (3 in FQ3 + 2 in early FQ4), covering >1/3 of FY2027 BiCS supply. Total RPO reached $42B — over a year of locked-in demand at the FQ4 guide midpoint. Customers committed $11B in financial guarantees (including prepayments and third-party-managed instruments). Critically, NBM pricing is not fixed: volume is committed while pricing remains variable, allowing SanDisk to capture upside if prices rise. Management is targeting >50% of supply under NBMs over coming quarters — this is not an HDD LTA analogy.5
The structural upside optionality: management's estimate of 75–100 EB of incremental AI inference NAND demand in 2027 — doubling again in 2028 — is explicitly not yet in any analyst model.4 At 75–100 EB, that represents 6–8% of the entire flash market as a new incremental demand source.
Bear case: Samsung resolves NAND yield issues in H2 2026 and re-enters as a credible supply competitor. Kioxia and SK Hynix have both signalled capacity expansion plans. Google's TurboQuant (6× KV cache compression) could reduce per-inference NAND demand — though management's read is that compression enables longer context windows and more concurrent requests, making the net NAND impact neutral to slightly positive.
Thesis validated by Q3 execution; AI inference demand is structural. A− not A because NAND pricing is spot-driven — durability depends on supply remaining tight, which is a market variable, not a contractual guarantee as with MU's HBM.
Business Quality
A
Q3 FY26 was exceptional on every dimension: $5.95B revenue (+97% QoQ, +251% YoY), GAAP EPS $23.03, non-GAAP operating margin 70.9%.1 Q4 FY26 guide of $7.75–8.25B with non-GAAP gross margin 79–81% implies continued margin expansion — the sequential improvement from Q3 to Q4 is the steepest in the company's public history.
Balance sheet: zero debt and $3.74B cash after repaying the outstanding $603M debt in Q3 (funded entirely by operating cash flow). This is the cleanest NAND balance sheet on record — no debt-refinancing risk entering the cycle's second derivative.5
Capital allocation: $6B share repurchase authorised — the largest capital return since the Feb 2025 spinoff from Western Digital. At current prices this represents ~2.9% of the market cap. Management conviction in the forward earnings trajectory is the clearest possible signal.
BiCS8 ramp: expected to reach majority of bits shipped by end of FY26. The mix shift to higher-capacity nodes improves both COGS and margin structure — not merely a revenue story but an operating leverage play.
Balance sheet: zero debt and $3.74B cash after repaying the outstanding $603M debt in Q3 (funded entirely by operating cash flow). This is the cleanest NAND balance sheet on record — no debt-refinancing risk entering the cycle's second derivative.5
Capital allocation: $6B share repurchase authorised — the largest capital return since the Feb 2025 spinoff from Western Digital. At current prices this represents ~2.9% of the market cap. Management conviction in the forward earnings trajectory is the clearest possible signal.
BiCS8 ramp: expected to reach majority of bits shipped by end of FY26. The mix shift to higher-capacity nodes improves both COGS and margin structure — not merely a revenue story but an operating leverage play.
Insider flag (last 90 days):
- One director sale: 3,500 shares @ $627.53 on 25 Feb 2026 ($2.2M total).2 Single transaction, single insider — not a cluster. Well below the $50M flag threshold.
- Limited history: Feb 2025 spinoff gives only 15 months of insider track record. Insufficient base for full directional read.
Q3 FY26 execution is the cleanest high-conviction quarter in the watchlist. $6B buyback puts real capital behind management conviction. Only asterisk: supercycle margins are not steady-state — the A grade is for execution quality, not margin durability.
Entry Price Discipline
C+
At $1,398, SNDK trades at:
At current $1,398 vs revised entry zone: bull case +29% to $1,800, bear case −79% to $300. Asymmetry is 2.7:1 downside/upside — unfavourable. At $1,200–$1,350 entry: bull +33–50%, bear −75–78%, improving to ~1.7:1. That asymmetry is more tolerable for a watch-list conviction name.
- Fwd P/E ~10–12× — reasonable on post-Q3 revised estimates; prior historical avg ~9.7× (limited 15-month spinoff history)3
- Fwd P/S 8.4× vs trough 0.6× last August — already re-rated ~14×
- 52W return +2,720% from ~$49 one year ago; stock previously touched $1,600 intraday
- 3.6–16.5% above the revised entry zone of $1,200–$1,350
At current $1,398 vs revised entry zone: bull case +29% to $1,800, bear case −79% to $300. Asymmetry is 2.7:1 downside/upside — unfavourable. At $1,200–$1,350 entry: bull +33–50%, bear −75–78%, improving to ~1.7:1. That asymmetry is more tolerable for a watch-list conviction name.
Trading 3.6–16.5% above the $1,200–$1,350 entry zone. Post-Q3 analyst revisions are the near-term re-rating catalyst; bear scenario at $300 is still on the table if the supply-demand balance shifts before end of 2026.
Verdict — Not Held, New Position
Add on pullback into $1,200–$1,350. Q3 FY26 definitively validated the AI inference NAND thesis — the question is no longer whether the thesis is correct but whether $1,398 is the right entry. It is not, but only marginally. Post-Q3, the consensus PT revision cycle (from stale $1,409 to likely $1,600–1,900) improves forward risk-reward. The $6B buyback provides structural support near $1,300. Add on any macro-driven pullback, cycle-fear overshoot, or post-Q4 report volatility that brings price into zone.
Cyclicality reminder: NAND gross margins of ~77% (Q3 FY26) and 79–81% (Q4 guide) are supercycle peaks. Trough margins from 2022–23 cycle were ~25–30%. The bear DCF at $300 is not a tail scenario — it is the base outcome if Samsung and Kioxia execute their stated capacity expansion plans. Any Add recommendation on SNDK carries this downside explicitly. Do not size SNDK to compete with MU — combined memory exposure should cap at ~12% of the active book.
Cyclicality reminder: NAND gross margins of ~77% (Q3 FY26) and 79–81% (Q4 guide) are supercycle peaks. Trough margins from 2022–23 cycle were ~25–30%. The bear DCF at $300 is not a tail scenario — it is the base outcome if Samsung and Kioxia execute their stated capacity expansion plans. Any Add recommendation on SNDK carries this downside explicitly. Do not size SNDK to compete with MU — combined memory exposure should cap at ~12% of the active book.
Entry zone: $1,200–$1,350 on pullback. Size to 3–5% NLV (not to compete with held MU at 8.00% NLV). Do not initiate above $1,400 before Q4 FY26 print (~Late Jul 2026). 20% single-stock cap is not the binding constraint — combined memory cyclicality is.
Opportunity-cost check
▾
vs CSPX (40% NLV)
CSPX at ~21× fwd P/E offers ~10% expected CAGR with full diversification and zero single-cycle blow-up risk. SNDK at 10–12× fwd P/E with 251% YoY revenue growth and AI inference structural demand optionality offers higher expected return — but only if the cycle extends beyond the 2-year consensus window. Edge over CSPX is the NAND undersupply pricing power. Only valid at entry-zone prices, not at $1,398 ATH-adjacent.
vs current holdings
MU (held, 8.00% NLV) covers HBM/DRAM with price-locked contracts through 2026 — structurally more defensible margin than SNDK's spot NAND. SNDK adds the AI inference storage layer (SSD/NAND) distinct from MU's training/compute-memory exposure. Partially correlated cycle, different customer qualification timelines. Initiating SNDK at 3–5% NLV alongside MU at 8.00% puts combined memory at ~11–13% of active book — at the upper boundary of tolerable correlation risk. Do not add both simultaneously; sequence the entry.
Thesis-breakers
▾
Three falsifiers. Any one trips → do not initiate. If already held → exit position.
- 01Non-GAAP gross margin falls below 65% in any quarter — first unambiguous signal of NAND spot-price rollover; margins compress before revenue does, making this the leading-indicator falsifier · By Q2 FY27 print (~Jan 2027)
- 02Q4 FY26 revenue misses $7.5B (below guide midpoint $8B by >6%) — breaks the consecutive-beat streak and introduces cycle-deceleration narrative; current price has no buffer for a miss · At Q4 FY26 report date, ~Late Jul 2026
- 03Samsung or Kioxia announces NAND capacity expansion >15% YoY for calendar 2027 — supply-glut precursor; arrives 2–3 quarters before it shows in SNDK margins, but the derating begins at the announcement · By end of Dec 2026
Key risks
▾
- 01NAND cycle turn: Spot-priced NAND has historically compressed 30–40 gross margin points when supply surpasses demand. At $8B FCF peak, a 35 ppt GM compression on the Q4 FY26 revenue run-rate implies FCF collapsing to ~$2B. Bear DCF: ~$300/share, −79% from $1,398.
- 02251% comp wall: Q1 FY27 compares against Q1 FY26 ($2.31B est revenue). Even strong absolute Q1 FY27 results will show dramatic YoY deceleration vs the +251% base — market sensitivity to even minor sequential misses is extreme from ATH-adjacent pricing.
- 03Samsung NAND supply normalisation: Samsung was qualification-excluded from some NAND supply chains through H1 2026. If Samsung resolves BiCS yield issues and re-enters as full-volume NAND supplier in H2 2026, it adds meaningful supply into a market management has been calling undersupplied — a direct thesis risk.
- 04MU correlation: Initiating SNDK alongside held MU creates a correlated memory pair. Both collapse simultaneously on any DRAM+NAND cycle signal (common macro trigger: hyperscaler AI capex pause, inventory build). Combined exposure caps the active book's effective diversification.
- 05Enterprise SSD displacement: If hyperscalers build proprietary CXL-attached or processing-in-memory tiers that displace external enterprise SSDs for inference caching, SNDK's DC wallet shrinks structurally. Not the base case — but any NVIDIA announcement on disaggregated memory architecture warrants re-assessment.
DCF scenarios
▾
Discount rate: 11% — profitable, FCF-generative, lighter capex profile than MU (Flash Ventures JV with Kioxia shares manufacturing investment); memory cyclicality prevents a lower rate. Shares outstanding: ~148M (implied $207B mkt cap ÷ $1,398). FCF base FY26e: ~$8B estimated (Q1 ~$200M + Q2 $980M + Q3 ~$2.2B + Q4 ~$3.5B on $8B guide at peak margins). ⚠ FCF base is estimated pending Q3/Q4 FY26 actual FCF disclosure.
FCF base (FY26e)
~$8B
5Y FCF CAGR
15%
Terminal growth
4%
Fair value / share
~$1,800
Bull: NAND cycle extends through FY28; AI inference demand materialises per management's 75–100 EB 2027 estimate; BiCS8 majority production delivers mix-driven margin improvement. FCF compounds 15% annually from $8B base. Fair value ~$1,800 implies +29% from $1,398. Terminal multiple reflects NAND as semi-infrastructure, not pure commodity.
FCF base (FY26e)
~$8B
5Y FCF CAGR
−30%
Terminal growth
2%
Fair value / share
~$300
Bear: Samsung and Kioxia execute capacity expansion in H2 2026; NAND spot prices roll 30%+ QoQ; gross margins compress from ~80% to 30–35% (prior-cycle trough). FCF collapses from $8B to ~$2B by FY28. Consistent with 2022–23 NAND cycle. Fair value ~$300 implies −79% downside. This is not a tail scenario — it is the historical NAND playbook when supply surpasses demand.
Position: At $1,398, stock sits between bull ($1,800) and bear ($300). 2.7:1 downside/upside asymmetry. At the entry zone ($1,200–$1,350) the asymmetry improves to ~1.7:1 — acceptable for a watch-list addition at limited size. The $6B buyback sets a practical floor nearer $1,300.
Catalyst timeline
▾
-
Jun–Jul 2026Analyst PT revisions post-Q3. Street repricing from stale consensus $1,409 toward $1,600–2,000 range as models incorporate Q3 actuals and Q4 guide. Creates near-term re-rating catalyst; also the last opportunity for the stock to pull back into entry zone before Q4 print.
-
Late Jul 2026Q4 FY26 earnings (~4 Jul quarter-end · Est). Watch: (1) Revenue vs $7.75–8.25B guide — any miss triggers cycle-fear sell-off; (2) Non-GAAP GM vs 79–81% guide — first read on whether margins can hold above 80%; (3) FY27 revenue guide tone — this is the single most important catalyst in the next 12 months.
-
Aug–Oct 2026Samsung, Kioxia, SK Hynix H2 capex announcements. Any NAND capacity expansion >15% YoY for 2027 signals supply glut arriving 2–3 quarters ahead. This is the leading indicator for thesis-breaker 03 — arrives before it shows in SNDK margins.
-
Nov 2026Q1 FY27 earnings (est). First brutal YoY comp ($2.31B Q1 FY26 base). Sequential growth continuation = cycle-extension narrative intact. Any sequential miss → derating begins. This quarter sets the tone for the FY27 narrative.
Named analyst commentary
▾
"There has been roughly a 4× increase in management's CY26 data center growth forecast over nine months, across three-month increments, with each revision higher than the last. Against a NAND bit supply base growing in the high teens, the demand-supply gap is widening every quarter rather than closing."
— Beth Kindig / IO Fund · SanDisk FQ3 FY26: Data Center Inflects 233% QoQ While New Business Models (NBMs) Weigh on the Stock · May 20265
No public-access coverage from Serenity or Dylan Patel on file as of 14 May 2026.
Sources
1 SanDisk Q3 FY26 earnings press release and slides — investor.sandisk.com, 30 Apr 2026
2 SEC Form 4 — SanDisk director, filed ~Feb 2026. 3,500 shares sold @ $627.53, 25 Feb 2026
3 MarketBeat / Stockanalysis.com — NTM consensus estimates, analyst ratings, PT range, 13 May 2026
4 SanDisk Q2 FY26 earnings call transcript — management commentary on KV cache NAND demand and supply-demand dynamics, Feb 2026 (est)
5 IO Fund (Beth Kindig / Royston Roche) — "SanDisk FQ3 FY26: Data Center Inflects 233% QoQ While New Business Models (NBMs) Weigh on the Stock", May 2026 (premium)
⚠ Fwd P/E range (~10–12×) reflects rapidly repricing post-Q3 NTM estimates. Q3 FY26 gross margin confirmed at 78.4% from IO Fund / SanDisk earnings release; FCF Q3 actual: ~$2.96B (49.7% margin) per IO Fund financials section.
1 SanDisk Q3 FY26 earnings press release and slides — investor.sandisk.com, 30 Apr 2026
2 SEC Form 4 — SanDisk director, filed ~Feb 2026. 3,500 shares sold @ $627.53, 25 Feb 2026
3 MarketBeat / Stockanalysis.com — NTM consensus estimates, analyst ratings, PT range, 13 May 2026
4 SanDisk Q2 FY26 earnings call transcript — management commentary on KV cache NAND demand and supply-demand dynamics, Feb 2026 (est)
5 IO Fund (Beth Kindig / Royston Roche) — "SanDisk FQ3 FY26: Data Center Inflects 233% QoQ While New Business Models (NBMs) Weigh on the Stock", May 2026 (premium)
⚠ Fwd P/E range (~10–12×) reflects rapidly repricing post-Q3 NTM estimates. Q3 FY26 gross margin confirmed at 78.4% from IO Fund / SanDisk earnings release; FCF Q3 actual: ~$2.96B (49.7% margin) per IO Fund financials section.