| Aug 5, 2026 | -6.81% | Q4 FY2026 | Read transcript briefingQuarter in one view- Another beat across every guided metric: revenue $8,965M vs $7.75–8.25B guide (+51% QoQ, +372% YoY); non-GAAP GM 84.6% vs 79–81% guide; EPS $39.25 vs $30–33 guide; adjusted FCF $5,035M at a 56% margin (excludes $1,938M of NBM prepayments/deposits sitting in operating cash flow).
- NBM program scaled hard: 5 additional agreements since the April call (3 new customers, 2 expansions of existing NBMs; 3 closed in-quarter, 2 after quarter-end). Total: 8 customers across Datacenter and Edge, weighted-average duration >4 years, minimum total revenue $93.9B at floor pricing, RPO $59.8B at quarter-end ($91.1B including the 2 post-quarter deals), $16.5B in financial guarantees.
- Coverage targets stepped up: NBMs now expected to cover >50% of FY27 bits (already committed, with POs in hand) and ~2/3 of FY28 bits — versus ">1/3 of FY27" last quarter.
- Stargate (QLC eSSD) began revenue shipments in the quarter, as guided. Datacenter revenue $2,977M (+103% QoQ); Datacenter now 38% of the portfolio exiting FY26 vs ~12% of bits a year ago.
- Capital return accelerated: $4.5B of buybacks executed (2.836M shares) and a new $14B authorization on top, $15.5B total remaining. Cash $4,762M, debt-free.
- FQ1 FY27 guide: revenue $10.3–10.8B, GM 83–85% (sequentially flat-to-down at midpoint despite higher pricing), opex $520–540M, EPS $44–46 on 155M shares. FY27 sellable bit growth lowered to mid-teens (from mid-to-high-teens) due to deliberate inventory build for NBMs.
- New TAM framing: NAND market >$300B in CY26 (3x YoY), approaching $500B in CY27; Datacenter share of TAM ~30% CY25 → ~50% CY26.
What management is focused on- NBMs as the dominant business model: "quickly becoming our predominant way of doing business." Selection criteria now explicit — strategic customers, ~5-year duration, growing volumes, attractive financials — and management says it will be "patient" and "highly selective" from here with much of the portfolio already spoken for.
- NBM economics disclosure improved materially: floor-pricing minimum revenue ($93.9B), RPO bridge (recognized revenue explains the gap between total and RPO), guarantee mechanics ($16.5B, mostly third-party-held, released toward end of agreements so guarantee/RPO ratio rises over time), and fixed + variable pricing with floors and ceilings — "attractive margins even at floor pricing."
- NBM margin anchoring: management confirmed NBMs were struck "around 80%" gross margin (Reitzes' question), with some upside capture as prices rise — i.e., signed near but below the current 84.6% print.
- Durability over spot maximization, again: Goeckeler explicitly framed mid-80s GM as a "fair return" and the goal as removing boom/bust, referencing the FY23 downturn. Customer engagement now at CEO/CFO level vs quarterly price negotiation a year ago.
- Inference narrative unchanged but deepened: agentic AI as a "multiplier," KV cache views "continue to mature" and get "more bullish," still no sizing — deferred to next week's Investor Day.
- Supply discipline intact: mid-to-high-teens long-term bit CAGR via nodal transitions (BiCS8 ramping to majority of bits; BiCS10 announced), FY27 capex ~6% of revenue, no acceleration despite allocation extending "beyond calendar year 2027."
- Buybacks as the chosen return mechanism: dividend explicitly considered and passed over — buybacks deemed "more tax efficient... the right thing for us at this point."
Key numbers and quarter mechanics- Segments: Datacenter $2,977M (+103% QoQ); Edge $5,432M (+48% QoQ); Consumer $556M (-32% QoQ). FY26 totals: revenue $20,248M (+175%); DC $5,153M (+437%); Edge $12,160M (+195%); Consumer $2,935M (+29%).
- Sequential revenue bridge: ~1/3 volume, ~2/3 pricing. FQ1 guide embeds both bit growth and "modest price increases."
- Margins: GM 84.6% (+620 bps QoQ); opex $484M (5.4% of revenue, below guide); operating margin 79.2%. R&D ~65% of opex. Note: Luis claimed five straight quarters of GM expansion "22.7% all the way to 86.5%" — the printed figure is 84.6%; the 86.5% is unexplained and should be verified (possible GAAP vs non-GAAP or transcription issue).
- EPS $39.25 on 157M diluted shares. GAAP items: +$807M gain on the Nanya investment, $67M SBC, $175M tax benefit on divested employee equity offset by $170M tax on the Nanya gain.
- Cash flow: $7,126M operating cash flow − $153M net capex = $5,035M adjusted FCF (56% margin), excluding $1,938M of NBM prepayments/deposits. Gross capex $562M (6.3% of revenue — up from 4%, as flagged last quarter).
- Buyback math: $4.5B for 2.836M shares implies ~$1,587 average price; share count 158M → 157M diluted, guided to 155M next quarter.
- FQ1 GM guide of 83–85% (vs 84.6% print) with pricing up modestly: management attributes to mix, "prudent assumptions on component costs," and higher inventory days — and explicitly denied NBMs are a margin drag.
- FY27 planning: capex dollars up YoY (BiCS8 + BiCS10 transitions, later conversions more expensive) but ~6% of revenue; higher inventory days sustained; sellable bits mid-teens.
Product and launch scorecard- Stargate (QLC eSSD): shipped for revenue in the quarter — delivered on the FQ4 timing commitment made last quarter. No revenue sizing, qualification count, or margin detail given; contribution is embedded in the +103% Datacenter growth but not split out. Sizing remains an open item for Investor Day.
- TLC compute eSSD: "scaled across a broad set of hyperscale and AI infrastructure customers" — still the presumed bulk of DC revenue, but no TLC/QLC mix update this quarter (last quarter's ~2/3–1/3 split not reiterated).
- BiCS8: ramped to majority of bit production in FY26 — the cost/performance engine behind the margin structure.
- BiCS10: announced "last week or last couple of weeks"; no specs, pricing, or premium detail ("a little early to talk about pricing"); more at Investor Day.
- HBF: timeline language loosened — no repeat of "NAND die late CY26, controller system early-to-mid CY27." Instead: "deep conversations with customers," new Advisory Board members, customer testimonials at FMS, a global standard announced with SK Hynix, and "we'll tell you more next week" on shipping/release dates. Progress claimed ("come an enormous distance") but the prior firm milestones were not reaffirmed on this call — verify at Investor Day.
- Consumer: -32% QoQ. Management's explanation shifted from "seasonality" (last quarter) to pricing-elasticity dynamics — consumer prices have risen, TAM impacted, and finding the "equilibrium point" takes longer than enterprise. Claims market share is holding. This is now two consecutive declines (-10%, -32%).
- Edge: +48% QoQ to $5.4B despite management stating PC and smartphone units are down mid-teens in CY26 — growth is price/mix/content-driven (smartphone capacity/content up mid-teens; PC content flat). Units expected to flatten in CY27 with content growth resuming exabyte growth.
Sell-side read-through- The FQ1 guide was the pressure point: Bernstein (Newman) noted the guide "seems a bit lighter compared to expectations" and the stock was down after hours, asking whether it's volume, locked-in pricing capping upside, or conservatism. Management's answer — both bits and modest pricing contribute — did not directly address the conservatism question that was answered candidly last quarter.
- Margin-durability questions dominated: Melius (NBM margins ~80% — confirmed), Cantor (why GM guided down with pricing up — answered with mix/component-cost language and a durability sermon), Jefferies (same question again — Luis explicitly denied NBMs are a drag and cited "prudent assumptions"). Three analysts probing the same GM-bridge gap signals the market is trying to find the ceiling.
- Hyperscaler identity: Goeckeler volunteered that NBMs exist with "several of the largest hyperscalers in the world" — the first on-record confirmation, answering Goldman's standing question from last quarter — but Goldman (Schneider) was still refused specifics on which agreements.
- Consumer got its first real challenge (BNP's Ackerman): allocation vs demand, and visibility absent LTAs. The exchange was confused — Luis twice asked for the question to be repeated and ultimately answered a different question (NBM customer demand confidence) than the one asked (consumer end-demand visibility). Consumer remains the least-explained segment.
- Supply/demand convergence question (Wells Fargo) drew a philosophical non-answer — convergence happens "with customers and suppliers having deep discussions," not at industry level; no timeline offered.
- BofA (Mohan) pressed the $300B→$500B TAM math: with bit growth not accelerating, the forecast embeds continued price escalation into 2027. Management did not confirm or deny any price-normalization assumption — "we continue to see very robust demand through the end of '27 and into '28" — and committed only to growing "with the market," not outgrowing it (also confirmed to Morgan Stanley: hold segment share).
- Notable absences: no one asked about the $807M Nanya gain's effect on GAAP optics, the guarantee-release mechanics' tail risk, consumer's -32% decline severity, or the prior quarter's unresolved items ($93M legal charge, SBC levels). No challenge to the 3x CY26 TAM claim.
Management credibility- Delivery: fourth consecutive quarter beating every guided metric (revenue $715M above high end; GM ~360 bps above; EPS ~19–31% above). The admitted-conservatism pattern persists — guidance remains a floor, not a predictor.
- NBM specificity stepped up again: total minimum revenue, RPO with and without post-quarter deals, guarantee totals and mechanics, duration (>4 yr weighted avg), bit coverage (>50% FY27, ~2/3 FY28), and margin anchor (~80%) are all new, checkable disclosures. The claim from prior quarters that customers return to upsize "within weeks" is now evidenced structurally — 2 of the 5 new deals are expansions of NBMs signed roughly a quarter earlier.
- Consistency holds on: supply discipline (no acceleration despite allocation beyond CY27), capex philosophy (dollars up modestly, % of revenue down), capital-allocation sequence (buyback executed exactly as telegraphed, then upsized), and Stargate timing (shipped when promised).
- Slippage/inconsistencies to flag: (1) Luis's "86.5%" GM claim vs the 84.6% print; (2) HBF milestones not reaffirmed — language shifted from dated deliverables to "more next week"; (3) FY27 sellable bit growth quietly lowered to mid-teens, attributed to inventory build — reasonable, but it is a downward revision to the standing mid-to-high-teens algorithm; (4) the FQ1 GM guide-down explanation (mix, component costs, prudence) was given only under repeated questioning and never quantified.
- Evasions unchanged: customer names, NBM segment mix, fixed-vs-variable split within NBMs, escalation cadence, and any structural margin floor all still refused. The "we don't ever expect to see that money" line on the $16.5B guarantees is assurance, not evidence.
- Promotional register still elevated ("gold standard," "match-made in heaven," "enviable companies"), and the TAM forecast has now escalated from exabyte-growth revisions to a $500B CY27 industry claim — forecast ambition continues to compound; error bars unaddressed.
What changed versus the prior quarter- NBM count: 5 → 8 customers (10 agreements total: 3 new-customer NBMs + 2 expansions this quarter, 2 of which closed post-quarter-end). RPO: $42B (3 deals) → $59.8B reported / $91.1B pro forma. Guarantees: >$11B → $16.5B. FY27 bit coverage: >1/3 → >50%; FY28 ~2/3 committed — the >50% ambition from last quarter was reached in one quarter.
- NBM pricing structure disclosed for the first time: fixed + variable with floors and ceilings; ~80% GM anchor confirmed; "attractive margins even at floor pricing."
- Stargate moved from "revenue shipments begin FQ4" to shipping for revenue — milestone met.
- Datacenter scale: $1.467B (25% of revenue) → $2.977B (33% of revenue); portfolio share exiting FY26 stated as 38% (vs ~12% of bits a year ago).
- Consumer deteriorated sharply: -10% → -32% QoQ, and the explanation changed from seasonality to price-elasticity/TAM impact — the allocation-vs-demand question from last quarter's watchlist is now acute and was not cleanly answered.
- Bit-growth algorithm revised down for FY27 (mid-to-high-teens → mid-teens sellable) on deliberate inventory build; inventory days guided structurally higher.
- Capex inflected up as flagged: 4% → 6.3% of revenue, ~6% guided for FY27, with BiCS10 now named as a spending driver.
- Capital return scaled: $6B authorization (untouched at last call) → $4.5B executed + $14B added = $15.5B remaining; dividend explicitly rejected for now.
- Market framing escalated: from CY26 exabyte growth (mid-70s%) to dollar TAM (>$300B CY26, ~$500B CY27, DC 50% of TAM by CY26). KV cache sizing again deferred — now to Investor Day.
- New GAAP noise: $807M Nanya mark-to-market gain and offsetting equity-related tax items — GAAP vs non-GAAP gap widening.
Bull case- The guide implies continued compounding: $10.3–10.8B revenue (+15–20% QoQ), EPS $44–46 — annualizing to ~$180+ of EPS power with GM structurally anchored at 83–85% and NBMs struck at ~80% floors.
- Contract coverage is now the story: >50% of FY27 bits and ~2/3 of FY28 bits committed with POs in hand, $91.1B pro-forma RPO against a ~$9B quarterly revenue base, $16.5B of guarantees, and customers expanding deals one quarter after signing — the strongest possible demand signal.
- Floor-pricing language ("attractive margins even at floor") plus variable components with ceilings means the downside case on contracted bits is bounded while uncommitted supply (~half of FY27, ~1/3 of FY28) retains spot upside.
- Allocation extends "beyond calendar year 2027" on management's forecast, with demand signals "to the end of the decade" — the cycle-elongation thesis now has contractual, not just rhetorical, support.
- Cash machine: $5.0B adjusted FCF this quarter (56% margin) plus $1.9B of NBM prepayments collected; $15.5B buyback authorization against ~157M shares is mechanically accretive at scale (Melius framed $5B/quarter ≈ 10% of market cap annually, and management did not push back on "very consistent" execution).
- Stargate adds a second Datacenter engine just as QLC AI-data-lake demand builds; HBF optionality preserved with SK Hynix standardization and customer advisory momentum.
Bear case- The FQ1 guide broke the beat cadence optically: GM guided flat-to-down (83–85% vs 84.6%) with only "modest" pricing, and the stock fell after hours. Management's explanations (mix, component costs, prudence) were qualitative and only extracted under three separate analyst questions — the first time the guide has embedded anything other than acceleration.
- NBM economics are now confirmed as struck at ~80% GM — below the current 84.6% print. As NBM coverage rises toward 2/3 of bits, the blended margin mathematically gravitates toward the contracted level unless spot pricing keeps rising; the "upside capture" on variable portions is unquantified and has ceilings.
- Consumer is in open decline (-32% QoQ) with an evasive explanation and no LTA-style visibility mechanism; Edge's +48% is price/content-driven while underlying PC/phone units fall mid-teens — two of three segments are shrinking in unit/demand terms, leaving Datacenter (33% of revenue and rising) to carry the model into a concentrated hyperscaler customer set.
- FY27 sellable bit growth was cut to mid-teens and inventory days raised — if NBM ramps or Stargate timing slip, the deliberately built inventory becomes a liability; the same mechanism was flagged last quarter and the inventory commitment has since grown.
- The $500B CY27 TAM requires sustained price escalation with no industry bit acceleration — BofA's normalization question was deflected, and management's own forecasting track record (five straight upward revisions of various metrics) demonstrates the range of error, not precision.
- GAAP quality: an $807M investment gain and stock-price-driven tax items inflate GAAP optics; guarantee recognition, prepayment accounting, and the unexplained "86.5%" GM quote all need verification against the 10-K.
- Bear-case asymmetry: with >50% of FY27 bits at ~80% GM floors, the downside is protected but the upside is increasingly capped — the stock now needs the uncommitted tail and FY28+ pricing to deliver beats, a narrower engine than the one that produced the last four.
Next-quarter watchlist- Investor Day (next week): the promised "new model"/structural margin framework, KV cache dimensionalization, HBF shipping/release dates and any TAM or customer commitments, BiCS10 specs and pricing posture, and NBM economics detail (fixed vs variable split, escalation mechanics). This is now the single biggest catalyst for the durability thesis.
- FQ1 delivery vs $10.3–10.8B / 83–85% GM / $44–46 EPS — and whether the GM guide-down proves to be conservatism (pattern) or the start of NBM-driven blend-down (thesis risk).
- NBM progression: RPO should print near or above $91.1B as the 2 post-quarter deals enter the reported number; watch new signings vs the "patient/selective" language, any segment-mix or hyperscaler disclosure, and evidence of further expansions.
- NBM margin tracking: does blended GM converge toward the ~80% contract anchor as coverage rises, or do variable/spot components hold it in the mid-80s?
- Stargate: first revenue sizing, qualification conversions, QLC mix update (last disclosed ~1/3 of portfolio), and any premium/pricing commentary.
- Consumer: does the decline stabilize, and does management give a real demand-vs-allocation answer? Watch for TAM contraction commentary.
- Edge: whether +48% pricing-driven growth persists as PC/phone units fall mid-teens; any Edge NBM signings as flagged last quarter.
- Inventory: days and dollars vs the "higher for longer" plan; sellable-bit trajectory against the lowered mid-teens FY27 algorithm.
- Buyback pace: execution vs the ~$4.5B/quarter run-rate and the $15.5B authorization; share count path to 155M and beyond.
- Housekeeping: reconcile the "86.5%" GM quote vs 84.6% print; verify NBM prepayment/deposit accounting ($1,938M) and guarantee release mechanics in the 10-K; confirm the Nanya gain treatment and the still-unexplained $93M legal charge from two quarters ago.
|
| Apr 30, 2026 | +8.25% | Q3 FY2026 | Read transcript briefingQuarter in one view- Massive beat across every guided metric: revenue $5.95B vs $4.4–4.8B guide (+97% QoQ, +251% YoY); non-GAAP GM 78.4% vs 65–67% guide; EPS $23.41 vs "$4.12 to $14" guide (transcript prints "$4.12 to $14" — prior guidance was $12–14; the "4" is almost certainly a transcription error, verify against the release); adjusted FCF $2.955B at a 49.7% margin.
- The structural story converted: five NBMs (multiyear supply partnerships) now signed — three in Q3, two more in early Q4 — covering over a third of FY27 bits, with ~$42B minimum contractual revenue (RPO) from the three Q3 deals alone and >$11B in financial guarantees including $400M of prepayments already on the balance sheet. Longest contract: five years.
- Data center inflected: +233% QoQ to $1.467B (~25% of revenue, up from ~15%), almost exclusively TLC eSSD; Stargate (QLC) revenue shipments begin in FQ4 (April quarter).
- Capital allocation flipped: remaining $650M TLB repaid, company now debt-free with $3.735B cash, and a $6B buyback authorized effective immediately — a sharp reversal from last quarter's explicit deferral.
- FQ4 guide: revenue $7.75–8.25B, GM 79–81%, opex $480–500M, tax expense $775–875M (absolute dollars, unusual framing), EPS $30–33 on 158M shares.
- CY26 data center exabyte growth forecast raised again: mid-70s%, from high-60s three months ago, mid-40s before that, mid-20s before that.
What management is focused on- NBMs as the centerpiece of the equity story: committed supply for customers, committed financials for Sandisk, with fixed + variable pricing (short-term mostly fixed, more variable further out), quarterly consumption commitments, and guarantees that pay Sandisk immediately if customers under-consume. Management claims normal course is customers asking to increase volumes "within weeks" of signing.
- Reframing the company around "RPO" and recurring-revenue language — management explicitly invites the software comparison and acknowledges the model is early but intends to push contracted bits "quite high" above 50%.
- Killing cyclicality rather than maximizing spot upside: "we are not interested in trading away that value for certainty; we are interested in getting that value and getting certainty as well." No structural margin target yet — a "new model" will be presented "when we get a little further along."
- AI inference architecture narrative deepened: KV cache, RAG, agentic systems, context retention as durable NAND demand drivers; data center mix "expected to keep rising over the next several quarters and the next several years."
- Supply discipline unchanged: mid-to-high-teens bit CAGR via BiCS8 nodal transitions, some incremental cleanroom CapEx (dollar terms rising modestly, % of revenue falling), no acceleration of supply with Kioxia beyond the aligned BiCS8 plan.
- Extending NBMs to edge customers as PC/phone units decline in CY26 — positioning contracts as the mechanism to see client demand rather than spot markets.
Key numbers and quarter mechanics- Segments: Edge $3.163B (+118% QoQ); Data center $1.467B (+233% QoQ, 25% of revenue); Consumer $820M (-10% QoQ, "in line with historical seasonality").
- Bits: flat YoY, down high-teens QoQ — deliberately building inventory ahead of the FQ4 Stargate ramp and NBM obligations. FYTD bit growth +18%, consistent with the mid-to-high-teens model.
- Margin mechanics: GM 78.4% (+2,730 bps QoQ) on price + mix; opex $448M (7.5% of revenue, below the $450–470M guide); operating margin 70.9%.
- Cash/FCF: $3.038B operating cash flow − $83M net capex = $2.955B adjusted FCF (49.7% margin). Gross capex $240M, 4% of revenue — the prior quarter's garbled capex line is resolved; the trend is real capex compression vs revenue.
- Balance sheet: cash $3.735B, zero debt after repaying the remaining $650M TLB (triggering a $46M write-off of unamortized issuance fees in GAAP). SBC only $20M net of tax (0.3% of revenue — notably low vs $52M last quarter; worth confirming).
- NBM economics: $42B RPO covers only the three in-quarter deals (the two Q4 adds are incremental and will appear next quarter); $11B+ guarantees include ~$400M prepayments on the Q3 balance sheet plus third-party-managed instruments. Pricing: mostly fixed near-term, increasingly variable longer-dated; no disclosed price-escalation cadence.
- FQ4 tax guided as an absolute dollar range ($775–875M) rather than a rate — implies continued high profitability; compute the implied rate against the EPS bridge.
- Data center annualized run rate now ~$6B (Mizuho's math, unchallenged).
Product and launch scorecard- TLC compute eSSD: the quarter's engine — 233% sequential data center growth "almost exclusively" TLC, with qualification breadth expanding ("now in a large number of accounts") and strong market pull. Clear product-market fit in inference architectures.
- Stargate (BiCS8 QLC eSSD): revenue shipments guided to begin FQ4 (April) — the first firm timing commitment, tighter than last quarter's "within the next several quarters." Management deliberately built inventory in Q3 to support the ramp but declined to size it. Under qualification at major players "for well over a year."
- TLC/QLC mix: now quantified — roughly two-thirds TLC / one-third QLC across the portfolio, predominantly TLC in data center. The old "QLC ~40% by FY26 exit" target remains unaddressed; mix question now partially answered by disclosure instead.
- HBF: still pre-revenue but timeline confirmed — NAND die late CY26, controller-based system early-to-mid CY27, consistent with prior guidance ("steady as she goes"). Management continues to frame HBF as complementary to, not a substitute for, eSSD; no TAM or economics.
- Consumer: next-gen portable SSD portfolio launched in February with AI-content-creation positioning and "meaningful external visibility"; "Space to Hold More" campaign. Segment declined 10% QoQ on seasonality while management emphasizes prioritizing "the most financially attractive demand" — consistent with allocation shifting to data center.
- KV cache: management now describes deep engagement on sizing variables (concurrent sessions, input tokens, cache hit ratios, storage durations) but explicitly will not size it — last quarter's 75–100 exabyte 2027 estimate was not reiterated or updated this call.
Sell-side read-through- The dominant thread was NBM structure, not the quarter itself: Bernstein (pricing fixed vs variable, pace of Q4 price increases), Melius (ceiling on contracted share — management volunteered >50% ambition), Goldman (hyperscaler participation — refused; ongoing RPO disclosure committed quarterly), BofA (price-escalation rights — refused, "no overall answer on pricing cadence"), Mizuho (segment mix of RPO — refused).
- Bernstein's implied observation that the FQ4 guide embeds decelerating price increases was met with "early in the quarter, extremely dynamic market, pays to be conservative" — the first acknowledgment that guide conservatism is deliberate, which reframes the string of beats.
- Morgan Stanley's market-vs-portfolio question on the 233% data center growth got an honest both/and answer — product readiness (TLC quals broadening) plus market pull.
- CapEx trajectory (Citi): CFO guided investors to think in dollars, not percent — modest dollar increases as later BiCS8 conversions are more expensive, but philosophy unchanged and no Kioxia supply acceleration (Goldman follow-up confirmed).
- Notable absence: no one challenged the mid-70s% exabyte forecast revision (fourth raise in four quarters), the $93M legal charge from last quarter was never raised, and no one pressed NBM pricing vs spot trajectory directly (Bernstein's lock-in concern from last quarter remains analytically open).
- Recurring-revenue framing invited by management and echoed by analysts ("I almost feel like I am on a software call") — the multiple-re-rating pitch is now explicit.
Management credibility- Delivery vs guide: three consecutive quarters of enormous beats (revenue $1.15B above high end; GM ~1,200 bps above; EPS far above guide). Management now explicitly attributes this to deliberate conservatism in a fast-moving market — credible, but it also confirms guidance is not a tight predictor in either direction.
- NBM claims have substance this time: named quantities (five deals, $42B RPO, $11B+ guarantees, $400M prepayments, five-year max duration, >1/3 of FY27 bits), 10-Q verifiable, and quarterly RPO disclosure committed. This is a step change in specificity versus last quarter's single undisclosed LTA.
- Consistency holds on: supply discipline (no Kioxia acceleration despite contract coverage), HBF timeline, mid-to-high-teens bit growth, JV strategy, and the capital-allocation sequence outlined in February (invest → net cash → returns) — the buyback arrived exactly when that sequence said it would.
- Evasions/gaps: customer names, NBM segment mix, price-escalation mechanics, and any structural margin target all refused; the stock "trading like your margin is going back into the 40s" (Melius) was answered with process language, not numbers.
- Promotional register elevated further: "gold standard," "spectacular cash generator," "people told me it would never happen," "this progress has converged in a single moment." Conviction is high; the demand forecast underpinning it has now been revised upward four quarters in a row (mid-20s → mid-40s → high-60s → mid-70s), which cuts both ways on forecast reliability.
- Housekeeping: the transcript's "$4.12 to $14" Q3 EPS guide comparison and unusually low SBC ($20M vs $52M) should be verified against the release; last quarter's $93M legal charge went unexplained and unquestioned.
What changed versus the prior quarter- NBMs scaled from one signed LTA to five, with quantified economics ($42B RPO, $11B+ guarantees, $400M prepayments) and a committed quarterly RPO disclosure — the structural thesis now has auditable proof points.
- Data center went from ~15% to 25% of revenue in one quarter; Stargate timing firmed from "next several quarters" to revenue shipments in FQ4.
- CY26 data center exabyte growth raised again (high-60s → mid-70s); KV cache sizing language walked back from last quarter's hard numbers (75–100 EB in 2027) to qualitative variable discussion.
- Capital allocation inverted: from explicitly deferring returns despite $936M net cash to debt-free, $3.7B cash, and a $6B buyback with no expiration. The Kioxia JV extension and ~$1B Nanya DRAM investment were cited as completed supply-chain prerequisites.
- Mix disclosure improved: TLC/QLC split (~2/3–1/3) quantified; data center share of portfolio (25%) stated directly — partially closing the mix-opacity gap flagged last quarter.
- Consumer shifted from accelerating (+39% QoQ) to declining (-10% QoQ) — attributed to seasonality, but allocation to data center is a plausible contributor; last quarter's watch item on data center starving consumer is now live.
- GM structure: 51.1% → 78.4%, with 79–81% guided — a new plateau, not a peak signal, per management.
- Edge exploded +118% QoQ to $3.163B (from $1.678B) — the largest single segment move and not deeply explained on the call beyond premium-mix and pricing.
Bull case- The guide implies another ~30–39% sequential revenue increase with GM of 79–81% and EPS $30–33; annualizing current-quarter EPS puts earnings power in territory that dwarfs any prior NAND cycle, and management claims NBMs make it durable rather than cyclical.
- $42B of minimum contracted revenue (excluding the two Q4 deals) against a ~$6B quarterly revenue base is extraordinary coverage; >1/3 of FY27 bits already committed with management targeting >50%, and customers reportedly asking to upsize within weeks.
- Guarantees (> $11B, third-party instruments, immediate payment on non-performance) materially de-risk the contracts; variable pricing components preserve spot upside further out.
- Data center compounding: +29% → +64% → +233%, TLC eSSD qualified broadly, Stargate QLC ramp starting FQ4 as a second engine, CY26 exabyte growth mid-70s and still excluding unmodeled KV cache upside.
- Balance sheet is pristine: debt-free, $3.7B cash, ~$3B quarterly FCF, $6B buyback active — per-share compounding now mechanically supports the story.
- Capex intensity falling (4% of revenue) while bit growth persists via node transitions — the FCF margin structure (49.7% this quarter) could persist well beyond a typical memory peak if NBMs hold.
Bear case- Pricing/mix remains the entire near-term driver: bits were down high-teens sequentially and flat YoY; the 78–81% GM plateau assumes the pricing environment holds. NBM financials are only "consistent with" current guidance — i.e., struck at today's elevated levels, which locks in peak-adjacent economics, not upside, on the fixed portions.
- Guidance conservatism is now admitted policy; the same opacity that produced +$1.15B beats can produce misses when pricing turns, and management declined to give any structural margin floor for the downturn case.
- NBM disclosures are aggregate-only: no customer names, no segment split, no escalation cadence, no price-vs-spot detail. If the fixed-price near-term layers are below where spot goes, Bernstein's upside-cap concern is unresolved; if hyperscalers aren't the counterparties (Goldman's unanswered question), the AI-demand linkage to the contracts is assumed, not proven.
- Demand forecasting credibility: four consecutive upward revisions (mid-20s → mid-70s in nine months) means the error bars are enormous; management simultaneously claims unprecedented visibility via NBMs and says the market moves "literally every day," making forecasting "difficult."
- Edge/consumer unit contraction is happening now (management concedes PC/phone units down, recovery only in CY27); consumer already declined 10%, and inventory was deliberately built ahead of Stargate/NBMs — if the ramp slips or NBM timing shifts, that inventory becomes a margin liability.
- Concentration risk rising: data center is 25% of revenue and heading higher on management's own words, into a customer set whose CapEx plans drive the entire exabyte forecast.
- Tax guided at $775–875M in absolute dollars and interest income guided oddly ($10–30M "income" against $3.7B cash — verify), and last quarter's $93M legal charge remains unexplained.
Next-quarter watchlist- FQ4 delivery vs $7.75–8.25B / 79–81% GM / $30–33 EPS — and whether the pricing-deceleration signal Bernstein flagged in the guide proves conservative or real.
- NBM progression: updated RPO including the two Q4 deals (must exceed $42B — quantify the increment), number of new agreements, any disclosure of hyperscaler participation, segment mix, and whether contracted FY27 bit share moves meaningfully above one-third toward the stated >50% ambition.
- Stargate ramp: actual revenue contribution, qualification completions converting to volume, whether the Q3 inventory build translates into shipments, and any sizing of the QLC opportunity (management declined to forecast it — check if that changes).
- RPO quality: how much is fixed vs variable pricing, any color on escalation mechanics, and evidence customers are upsizing as claimed.
- Buyback execution: pace and price of repurchases against the $6B authorization; cash build vs deployment.
- Mix disclosures: TLC/QLC progression as Stargate ships; eSSD bit share (was high-teens — update expected); the promised price-vs-mix ASP breakout from last quarter's commitment.
- Data center share trajectory toward and beyond 25%; edge's +118% quarter — was it price, mix, or one-time, and does it hold?
- Consumer: whether seasonal recovery materializes or allocation to data center structurally shrinks it; edge NBM signings with PC/phone customers as flagged.
- HBF milestones: NAND die late CY26, controller system early-to-mid CY27 — any customer commitment or TAM disclosure.
- Housekeeping: verify the "$4.12 to $14" prior-guide print, the $20M SBC figure, the $93M legal charge resolution, and the implied tax rate inside the $775–875M guide.
|
| Jan 29, 2026 | +6.85% | Q2 FY2026 | Read transcript briefingQuarter in one view- Revenue $3,025M, +31% QoQ, +61% YoY, well above the $2,550–2,650M guide; beat driven by pricing that "strengthened during the quarter" (bits up low single digits QoQ, +22% YoY).
- Non-GAAP gross margin 51.1% (51.9% ex $24M start-up costs) vs 41–43% guide — a ~900 bps beat driven by pricing; unit cost reductions merely "as expected."
- Non-GAAP EPS $6.20 vs $3.00–3.40 guide; opex $413M (13.7% of revenue) came in below guide due to a ~$35M one-time NPI/qualification-unit accounting benefit (see mechanics).
- Adjusted FCF $843M (27.9% margin); paid down another $750M of debt; net cash now $936M (cash $1,539M, debt $603M).
- Q3 FY26 guide is a step change: revenue $4.4–4.8B, GM 65–67%, EPS $12–14 on 157M shares, opex $450–470M; market "more undersupplied" than Q2 with bits down only mid-single digits (below historical seasonality) on data center strength.
- Structural narrative escalated: first LTA signed (with prepayment component, terms undisclosed, "several in the queue"); data center CY26 exabyte growth forecast raised again to "high 60s%" (from mid-40s last quarter, mid-20s two quarters ago).
What management is focused on- Rewiring industry business practices: moving from "quarterly auction" pricing to multiyear agreements with firmer supply/price commitments and prepayments. One agreement signed and closed; several more in progress; management frames this as the key enabler for any future capacity adds.
- Supply discipline unchanged: mid-to-high-teens bit growth CAGR via BiCS8 node productivity; any material CapEx increase requires multi-year committed demand at attractive pricing. Explicitly rejected adding supply in response to the current squeeze.
- Allocation to "strategic customers": prioritizing customers with multiyear frameworks and shared planning; acknowledged being unable to fulfill demand this quarter.
- Data center as the new center of gravity: data center expected to become the largest NAND market in CY26 (now stated as fact, first time in prepared remarks); management emphasizes inference-driven content growth.
- Kioxia JV extension: Yokaichi JV extended through 12/31/2034 (aligned with Kitakami); Sandisk paying $1,165M for manufacturing services over CY2026–2029, flowing through COGS over nine years — a real, multi-year gross margin headwind embedded in the model.
- Capital allocation: debt paydown and "prudent cash reserves" prioritized; shareholder returns explicitly deferred ("at the right time, we'll expand and give you an update"). No buyback commitment despite $936M net cash.
Key numbers and quarter mechanics- Segment revenue: Edge $1,678M (+21% QoQ); Consumer $907M (+39% QoQ, >50% YoY per CEO); Data center $440M (+64% QoQ, ~15% of revenue — analyst cited; consistent with math).
- Balance sheet: cash $1,539M; debt $603M (down from ~$2B at separation); net cash $936M.
- FCF: $1,019M operating cash flow − $176M net cash capex = $843M adjusted FCF. Transcript states "gross capital spending totaled $5.255 billion... 8.4% of revenue" — internally inconsistent (8.4% of $3,025M ≈ $255M); likely transcription error, verify against press release.
- Opex mechanics: $413M includes a ~$35M one-time gain from reclassifying qualification-unit costs from period expense to inventory (now charging customers for qual units). Management says the recurring run-rate benefit continues; treat Q3's $450–470M guide as the clean baseline. ~75% of opex is R&D.
- Q3 margin math: GM 65–67% at midpoint implies another ~1,500 bps sequential expansion, again pricing-led; bits guided down mid-single digits vs the ~12–14% historical seasonal decline flagged last quarter — mix shift to data center is cushioning seasonality as predicted.
- Tax: Malaysia loss carryforwards being consumed quickly; ongoing tax rate guided to ~14–15%.
- GAAP reconciling items: $52M SBC (1.7% of revenue); $93M for "certain legal matters" — new and notable, no detail given.
- eSSD = high-teens % of total bits (disclosed only in Q&A).
- Enterprise SSD growth sequence: +29% QoQ (Q1), +64% QoQ (Q2), "substantial step-up next quarter" guided verbally.
Product and launch scorecard- Stargate (BiCS8 QLC storage-class eSSD): still in qualification at two major hyperscalers; revenue shipments now expected "within the next several quarters" — language consistent with but not firmer than the prior mid-CY26 ramp framing; no new hyperscaler adds confirmed this call.
- Compute TLC eSSD (PCIe Gen5): second hyperscaler qualification completed; more quals expected "over coming quarters," BiCS8 TLC versions "soon thereafter." This product is driving the 64% sequential data center growth — the clearest product-level win.
- QLC/TLC mix: portfolio still "tilted towards TLC"; prior-quarter target of QLC reaching ~40% by FY26 exit was not reiterated — flag for follow-up.
- KV cache / NVIDIA context: management quantified potential 2027 NAND demand at ~75–100 incremental exabytes, doubling the year after; explicitly NOT yet in demand forecasts. Early-stage sizing, but the first hard numbers attached to the AI-inference storage thesis.
- HBF: still pre-revenue; NAND die being designed, controller in build, "deep conversations with customers on use cases." No timeline update (prior: memory CY26, controller CY27), no TAM. Management acknowledged "lots of folks working on that" — implicit concession of competitive convergence.
- Consumer: Extreme Fit USB-C launch; Crayola and FIFA World Cup 2026 licensed products; "Don't Delete Your Games" campaign; Sandisk Optimus rebrand of WD Black/Blue NVMe at CES. >50% YoY consumer growth supports the premium-mix claim.
Sell-side read-through- LTA economics remained the dominant thread (Bernstein, Wells Fargo, SIG): Wells Fargo extracted the quarter's key disclosure — one signed LTA with a prepayment component, terms undisclosed. Bernstein's lock-in-vs-upside tradeoff question was answered only directionally ("value accretive... super thoughtful" on length, price, volume, prepayment).
- Demand credibility probed (Cantor, Morgan Stanley): management's answer to "quantify AI demand" is the data center exabyte forecast — which has now been revised mid-20s → mid-40s → high-60s in two quarters, excluding any new CapEx announcements and excluding KV cache. Analysts did not challenge the volatility of these numbers directly.
- Capacity discipline confirmed (Goldman): fabs at full utilization for "a couple quarters," no underutilization charges in actuals or guide; Kitakami K2 fab opened and is the expansion vehicle; competitor greenfield announcements dismissed as "normal course," multi-year lead times.
- Mix/ASP opacity flagged (SIG): analyst asked for mix breakout so raw ASP isn't misread as data center mix shifts; management deferred to next quarter's report — a disclosure gap persisting at least one more quarter.
- SIG's sharpest challenge: why no customer urgency to commit given multi-year fab lead times? Management's answer was behavioral (decades-old quarterly practices change slowly) and optimistic, but conceded "we're not quite there yet" on announcing contracts beyond the one signed.
- Model hygiene items surfaced (Jefferies, BofA): the ~$35M one-time opex benefit was disclosed only under questioning; tax rate guided to 14–15% only when asked. Both matter for Q3+ modeling.
Management credibility- Delivery vs guide: massive beats on all guided metrics (revenue +$475M above high end; GM ~900 bps above; EPS ~2x). However, guidance quality is now a live question — guides this loose in a volatile pricing environment cut both ways on the way down.
- Consistency: the undersupply and allocation story is intact and now supported by a signed LTA with prepayment; the start-up cost glide path held ($30M guided → $24M actual, ~zero going forward); underutilization confirmed gone from actuals and guide.
- Improved transparency where pressed: disclosed eSSD bit share (high teens), the $35M opex one-timer, tax rate path, and the Kioxia payment schedule ($1,165M, COGS over nine years).
- Continuing gaps/evasions: no LTA terms (price, duration, volume, counterparty) on the one signed deal; no data center % of revenue target for year-end ("stay tuned"); no QLC mix update; KV cache sizing offered as rough ranges with explicit "not in our numbers" caveats; no HBF economics.
- Watch the promotional register: "perfect moment," "completely demand-driven phenomenon," high-60s% exabyte growth forecasts that have tripled in two quarters — conviction is high, but the forecast error bars are demonstrably wide, and management now attributes pricing beats partly to conditions "improving significantly in December," i.e., late-quarter visibility was poor.
- Debt figure inconsistency: CFO said "$600 million this quarter" and "$2 billion to $650 million" in different answers; reported debt is $603M — minor, but sloppy.
What changed versus the prior quarter- LTA narrative converted from "very, very little committed beyond one quarter" to one signed multiyear agreement with a prepayment, plus "several in the queue" and conversations extending to 2029–2030. Still immaterial in disclosed volume, but the first hard proof point.
- Data center CY26 exabyte growth forecast raised again: mid-40s → high-60s% (it was mid-20s two quarters ago). KV cache demand (75–100 exabytes in 2027, doubling in 2028) identified as entirely incremental and unmodeled.
- Data center went from ~12% to ~15% of revenue; Q3 guided for another "substantial step-up." Segment growth commitments from last quarter (growth every quarter through FY26) are on track.
- Margin structure reset: GM 51.1% actual, 65–67% guided — management's prior "35% through-cycle" framing was explicitly abandoned ("clearly, that's not where we want to be"), with no new structural target given.
- Balance sheet: net cash $91M → $936M; debt $1,351M → $603M. Kioxia JV extended to 2034 with a $1,165M payment obligation (CY26–29, COGS over nine years) — a new structural cost line.
- Q3 seasonal bit decline guided to only mid-single digits vs the ~12–14% historical pattern flagged last quarter — data center mix is muting seasonality as management predicted.
- Capital allocation posture hardened against returns: despite net cash approaching $1B, buybacks/dividends explicitly deferred in favor of cash reserves and remaining debt paydown — a shift from last quarter's "returning cash to shareholders" framing.
- New GAAP item: $93M for legal matters (no prior analog disclosed in the summary).
- Consumer accelerated (+39% QoQ, >50% YoY) rather than fading; PC 2026 unit forecast of 285M cited as above what "anybody would have picked at the beginning of the year," though management concedes unit base-effect declines ahead.
Bull case- The Q3 guide implies revenue +45–58% QoQ and EPS roughly doubling again at midpoint ($12–14 vs $6.20); GM 65–67% approaches memory-industry peak-cycle territory with bits declining — pure pricing/mix leverage.
- Structural repricing evidence is accumulating: one signed multiyear LTA with prepayment, broad-based customer engagement across geographies, and management explicitly stating margins are "expected to reset at a structurally higher level."
- Demand visibility keeps improving: data center exabyte growth now high-60s% for CY26 before any new hyperscaler CapEx raises and before KV cache demand (75–100 incremental exabytes in 2027, doubling after) enters the model.
- Data center momentum is compounding: +29% → +64% → guided "substantial step-up," second hyperscaler TLC qual done, Stargate revenue starting "within the next several quarters" as an additional tailwind.
- Balance sheet: $936M net cash, $843M FCF in one quarter (27.9% margin), JV certainty through 2034 — capacity for large returns once debt is retired (~$600M remaining).
- Fabs full, no wafer adds, competitors' greenfields years away — undersupply "beyond CY26" intact.
Bear case- Guidance is being set with poor late-quarter visibility (December pricing inflected after the guide was presumably built); the same dynamic in reverse would produce equally large misses in a downturn. Q3's 65–67% GM guide bakes in continued steep price escalation.
- The LTA moat is still thin: exactly one signed deal, terms undisclosed; if LTAs are signed at prices below spot trajectory (Bernstein's concern), they cap upside; if customers won't commit (SIG's challenge), the structural-reset thesis rests on quarterly spot pricing after all.
- Demand forecasts are whipsawing (mid-20s → mid-40s → high-60s in six months); management itself says it lacks demand visibility, which is precisely why it wants LTAs. Planning a multi-year supply posture on these numbers is high-risk.
- Pricing is the entire story near-term: bits down mid-single digits in Q3, no wafer capacity additions, and management openly "trading out the lowest margin business" — allocation-driven mix gains are finite.
- The Kioxia extension adds $1,165M of committed payments flowing through COGS over nine years — a structural margin drag that partially offsets the "reset higher" narrative, and it was disclosed without analyst-modeling detail.
- Legal charges ($93M) appeared with zero explanation; tax rate stepping up to ~14–15%; opex one-timer ($35M) inflates the Q2 beat optically.
- Edge/consumer base effects: management concedes PC/smartphone unit declines ahead in 2026; 70%+ of revenue still sits outside data center.
Next-quarter watchlist- Q3 delivery vs $4.4–4.8B / 65–67% GM / $12–14 EPS — and the promised mix-breakout disclosure (committed on the call for next quarter's report) to separate price from customer-mix effects on ASP.
- LTA conversion: number signed beyond the first one, and any disclosure of structure — duration, price vs spot, prepayment size, percentage of bit supply committed, and whether data center hyperscalers are the counterparties.
- Stargate: whether the two hyperscaler quals complete and revenue shipments begin; the third hyperscaler/storage OEM adds flagged for CY26 in prior quarters were not mentioned this call — confirm they remain on schedule.
- Data center: another "substantial step-up" as verbally guided; eSSD bit share progression from high-teens; second-hyperscaler TLC ramp; BiCS8 TLC qualification timing.
- Kioxia payments: how the $1,165M phases into COGS (per-year run rate) and its net GM impact; reconciliation of the $5.255B vs ~$255M gross capex transcription discrepancy; the $93M legal charge explained.
- KV cache: whether the 75–100 exabyte 2027 estimate gets incorporated into formal demand forecasts in the back half of the year, and any NVIDIA design specifics.
- Capital allocation: remaining ~$600M debt retired? Any concrete buyback framework now that net cash is ~$1B and FCF is running at ~$850M/quarter?
- Consumer/edge: evidence of the flagged PC/smartphone unit base effects; whether NAND allocation to data center starves these segments and what that does to consumer's premium-mix story.
- QLC mix: prior 20%→40% by FY26-exit target was not reiterated — ask directly.
- Opex: confirm the $450–470M guide as the clean run rate and that the qualification-unit charging benefit persists (~$35M ongoing).
|
| Nov 6, 2025 | +15.31% | Q1 FY2026 | Read transcript briefingQuarter in one view- Revenue $2,308M, +21% QoQ, +23% YoY, well above guide of $2.1–2.2B; beat driven by mid-teens sequential bit growth (pricing up mid-single digits, firming through the quarter).
- Non-GAAP gross margin 29.9% (+350 bps QoQ), above 28.5–29.5% guide; underlying GM ex $61M start-up and $11M underutilization charges was 33.1%.
- Non-GAAP EPS $1.22 vs $0.29 prior quarter and $0.70–0.90 guide; opex $446M was above guide ($415–430M) on higher variable comp from the beat.
- Adjusted FCF $448M (19.4% margin); reached net cash of $91M ~6 months ahead of the February Analyst Day target; repaid $500M of TLB; inventory days cut 135 → 115 (analyst cited 107 — figures differ in transcript, verify).
- Q2 FY26 guide: revenue $2,550–2,650M (double-digit pricing, mid- to low-single-digit bit growth), GM 41–43% (includes $30M start-up costs), opex $450–475M, EPS $3.00–3.40 on 155M shares.
- Structural call: demand exceeds supply through calendar 2026 and "beyond"; products on allocation across all end markets; data center becomes the largest NAND market (by exabytes) in CY26.
What management is focused on- Allocation discipline: prioritizing "strategic customers" with growth and mutual value; explicitly framing Sandisk as moving from mobile-centric toward data center.
- Long-term agreements: customers proactively seeking multi-quarter volume/price deals (first half CY26) and, nascently, visibility/commitments into CY27. Management stresses today there are "very, very little" volume/price commitments beyond one quarter — the LTA narrative is early-stage conversation, not signed contracts.
- BiCS8 transition: 15% of bits shipped in Q1, targeted to be the majority of bit production exiting FY26; described as the key enabler for data center share gains.
- Disciplined capacity stance: FY26 CapEx unchanged; investing to mid- to high-teens long-term bit demand CAGR; explicitly not adding wafer capacity — bit growth comes from node productivity, not wafers.
- Gross margin trajectory: management notes GM is still "4 quarters below model" despite +720 bps over two quarters.
- Capital allocation: net cash goal achieved; priorities now investment plus returning cash to shareholders (no specifics on buyback/dividend size).
Key numbers and quarter mechanics- Segment revenue: Edge $1,387M (+26% QoQ); Consumer $652M (+11% QoQ); Data center $269M (+26% QoQ, still only ~12% of revenue). Note nomenclature change: "cloud" → data center; "client" → edge.
- Cash $1,442M; gross debt $1,351M; net cash $91M.
- FCF build: $488M operating cash flow + $10M net from Flash Ventures activity − $50M back-end/office capex = $448M. Gross capex $387M (16.8% of revenue); $337M Flash Ventures gross capex, of which $240M externally funded (subsidies, equipment leasing).
- Reconciling items: $47M SBC (2% of revenue), $9M separation charges, $17M one-time SSD/separation costs.
- Q2 mechanics: ~900 bps sequential GM expansion (ex start-up/underutilization) is "majority pricing," with modest BiCS8 cost-per-bit benefit; start-up costs step from ~$61M → $30M → ~0 thereafter; underutilization gone (fabs at 100%).
- Supply/demand math offered: CY25 supply +8%; CY26 supply ~17%; constrained demand mid-teens; unconstrained demand ~mid-20s% (raised from ~20% "a couple of weeks ago"). Data center exabyte growth estimate for CY26 raised from mid-20s to mid-40s in three months.
- Q3 caution (management-flagged, not guided): historical seasonal bit decline of ~12–14% QoQ in the March quarter; mix shift toward data center may soften but not eliminate it.
Product and launch scorecard- Stargate (QLC storage-class eSSD, 128TB): two hyperscaler qualifications underway; a third hyperscaler plus a major storage OEM planned for CY26; ramp expected mid-CY26 (qualifications take several quarters). Prior reference to a "256TB ramping 1H CY26" was corrected by management — 256TB (UltraQLC) hits market mid- to late next year, ramps the following year.
- Compute eSSD (TLC): "continues to do very well," customer base broadening; no numbers given.
- QLC mix: expected to go from 20% to 40% of (data center?) business by end of FY26 — management phrasing ambiguous on denominator; verify.
- BiCS8: 15% of bits; majority exiting FY26; cost tailwind begins in Q2.
- HBF (high-bandwidth flash): timeline reaffirmed — memory later CY26, controller CY27; active customer engagements for inference use cases in edge and cloud; management declined to size TAM or detail competitive advantage vs HBM incumbents. No revenue contribution; pre-product.
- Consumer: Nintendo Switch 2 co-branded microSD Express >900K units sold in FQ1; new ROG Xbox Ally microSD; "Memory Man" campaign for holiday. Consumer framed as margin-accretive through cycle.
Sell-side read-through- LTA structure was the dominant thread (Cantor, Bernstein, Fox Advisors): analysts pushed on HDD-style build-to-order/LTA parallels. Management's answers consistently dialed this back to "early conversations" — Bernstein extracted the key fact that virtually nothing is committed beyond one quarter today.
- Supply discipline questioned (Goldman): what would trigger wafer adds? Management said that discussion hasn't started; needs multi-year demand proof. Bullish for pricing durability, but a constraint on bit share gains.
- Margin guide credibility probed (BofA, Citi): analysts reverse-engineered the 41–43% GM guide; management confirmed it's mostly pricing, with Citi catching that revenue guide implies double-digit pricing plus some bit growth.
- Share vs mix: Benchmark asked directly if Sandisk is gaining data center share — management said yes, growing faster than market. BofA confirmed data center will lead back-half growth.
- HBF questions (Bernstein, Arete) were largely deflected — no TAM, no competitive specifics — a disclosure gap.
- Management declined to re-engage on cost-down percentages (called Bernstein's high-teens framing "too aggressive"), continuing a multi-quarter policy of not guiding cost declines.
Management credibility- Strong quarter of delivery: beat on all guided metrics (revenue, GM, EPS), net cash reached ~6 months early, start-up/underutilization cost glide path ($60M → $30M → ~0) tracking as previously communicated.
- Undersupply call made "at least a year" in advance is now playing out; management can credibly claim foresight on the cycle turn.
- Consistency checks: corrected the record on 256TB timing (analyst's prior-quarter recollection was off, per management); Q2 guide built with internally consistent price/bit math; opex miss explained transparently as variable comp on the beat.
- Evasions/gaps: no QLC/TLC revenue split; no AI vs traditional cloud split beyond "majority AI"; no HBF economics; no LTA economics ("stay tuned"); declined cost-down guidance; inventory days figure inconsistent (115 prepared vs 107 cited in Q&A).
- Language is promotional in places ("new era," "technology arriving at exactly the right time") but the substantive claims (quals, ramps, allocation) were consistently hedged on timing.
What changed versus the prior quarter- No prior-quarter summary available; transcript-only deltas:
- End-market reporting renamed (cloud → data center; client → edge) — comparability footnote for models.
- EPS inflected sharply ($0.29 → $1.22) and the balance sheet flipped to net cash.
- Undersupply view extended from "through end of CY26" to "beyond CY26."
- Data center CY26 exabyte growth forecast raised from mid-20s to mid-40s in one quarter; unconstrained demand view raised from ~20% to mid-20s within weeks.
- Customer behavior shifted: multi-quarter and CY27 supply conversations are new this quarter.
- Start-up costs stepping down ($61M → $30M → ~0); underutilization eliminated (100% fab utilization).
- Data center described as becoming the largest NAND market in CY26 — a first.
Bull case- Pricing-led EPS ramp: Q2 guide implies EPS roughly 2.5x QoQ at midpoint with double-digit price increases and fabs full; GM path 29.9% → 41–43% in one quarter.
- Structural supply discipline: no wafer adds planned, industry supply growth (~17% CY26) below unconstrained demand (mid-20s), allocation persisting beyond CY26 — supports pricing durability and LTA optionality.
- Data center inflection with low base: only ~12% of revenue today, growing faster than market, five hyperscaler engagements, Stargate quals converting to ramps mid-CY26, QLC mix doubling to 40%.
- Balance sheet: net cash achieved early, $448M FCF in one quarter, capacity for returns.
- AI inference optionality (HBF) as unpriced upside; consumer/gaming (Switch 2) providing profitable volume.
Bear case- The LTA story is aspirational: "very, very little" committed beyond a quarter; if LTAs get signed now, they could cap participation in further price upside (Bernstein's point, unanswered on economics).
- Bit growth is capped by no wafer adds — mid-single-digit bit growth in Q2 and "grow with the market" stance means the story is almost entirely pricing; any demand air pocket hits hard with fabs at 100% and inventories already lean.
- Q3 seasonality (management's own 12–14% historical bit decline) lands right after a pricing-driven spike; data center ramp (mid-CY26) arrives just as consumer seasonally fades.
- Data center remains small (~$269M); quals "take several quarters"; 256TB pushed to mid/late CY26 market entry — share-gain proof points are 2–4 quarters out.
- Demand estimates are moving weekly (mid-20s → mid-40s data center revisions) — cuts both ways; forecasts this volatile have wide error bars.
- Opex rising ($450–475M guide) and share count up to 155M on stock appreciation; HBF consumes R&D with no disclosed TAM or timeline certainty beyond memory-in-CY26/controller-CY27.
Next-quarter watchlist- Q2 delivery vs $2.55–2.65B / 41–43% GM / $3.00–3.40 EPS — specifically the price/bit split and whether pricing momentum persisted through the quarter (recall Q1 exit pricing was better than entry).
- LTA conversion: do the multi-quarter/CY27 conversations produce signed volume/price commitments, on what terms (price collars, prepayments, CapEx sharing), and with which hyperscalers?
- Stargate qualification progress: completion of the two underway hyperscaler quals, start of the third hyperscaler + storage OEM, and any revision to the mid-CY26 ramp timing.
- Data center sequential growth (management committed to growth every quarter through FY26) and segment mix shift toward BiCS8/QLC (40% target by FY26 exit).
- BiCS8 progress toward majority of bits exiting FY26, and confirmation start-up costs go to ~$0 as promised.
- Q3 framing: does management quantify or soften the seasonal bit decline given data center mix?
- HBF: any TAM sizing, design-win disclosure, or timeline slip; watch the SK hynix partnership for ecosystem traction.
- Capital allocation specifics now that net cash is achieved — magnitude and form of shareholder returns; any change to "unchanged" FY26 CapEx or wafer-capacity posture.
- Reconcile the inventory-days discrepancy (115 vs 107) and clarify the QLC 20%→40% denominator.
|
| Aug 14, 2025 | -4.58% | — | Transcript briefing unavailable |
| May 7, 2025 | +4.83% | — | Transcript briefing unavailable |
| Mar 7, 2025 | +3.92% | — | Transcript briefing unavailable |
Corbanu