SANDISK CORPORATION (SNDK)
Outperform

The Largest Above-Consensus EPS Guide in NAND History: Q3 FY26 $12-14 vs Street $5.50, GM 65-67%, Yokaiichi Extended Through 2034, LTAs Emerging — Maintaining Outperform (High Conviction)

Published: By A.N. Burrows SNDK | Q2 FY2026 Earnings Analysis

Key Takeaways

  • Q2 FY26 print: revenue $3.025B (+31% QoQ / +61% YoY) vs guide $2.55-2.65B; non-GAAP EPS $6.20 vs guide $3.00-3.40 (+94% above mid); non-GAAP GM 51.1% (+2,120bp QoQ!) vs guide 41-43%. The print itself was a 100%+ above-guide blowout.
  • Q3 FY26 guide is the headline: EPS $12-14 (midpoint $13) vs Street $5.50, the largest above-consensus quarterly EPS guide in the entire history of the NAND industry. Revenue guide $4.40-4.80B; GM 65-67% (vs Street ~48-50%). Implied annualized EPS run-rate at the guide midpoint is ~$52.
  • Yokaiichi JV extension with Kioxia through 12/31/2034 (was expiring 2027) secures multi-year supply continuity. $1.165B payment to Kioxia 2026-29 flows through COGS over 9 years, a manageable cost for foundational supply chain security at the moment when capacity is most strategically valuable.
  • Multi-year supply agreement (LTA) conversations advanced "across broad base" of customers and geographies, fixed pricing short-term, variable long-term (mutual upside-capture). The customer-driven LTA shift mirrors HDDs in 2021-23 and structurally reduces NAND cyclicality.
  • Rating: Maintaining Outperform, upgrading conviction from moderate to high. Q1 FY26 thesis confirmed and accelerated. Updated fair value range $675-855 (~15-19x FY27 EPS power of ~$45). Post-print close $576.25; meaningful upside still available as Street estimates climb toward our $40+ FY27 EPS framework.

Results vs. Consensus

MetricQ2 FY26 ActualGuidanceBeat/MissMagnitude
Revenue$3.025B$2.55-2.65BBeat+$425M / +16%
Non-GAAP Gross Margin51.1% (+2,120bp QoQ)41-43%Beat+800-1,000bp above
Non-GAAP Operating Margin37.5% (+2,690bp QoQ)n/an/an/a
Non-GAAP EPS$6.20$3.00-3.40Beat+$3.00 / +94% above mid
Bit growth QoQ+LSD+mid-singleIn linen/a
PricingDouble-digit across all segmentsDouble-digit (guided)Confirmedstrengthened through December
Adjusted FCF$843M (27.9% margin)n/an/an/a
Net cash position+$936M+$91M Q1+$845M QoQn/a

Segment Performance

SegmentRevenueQoQYoYNotable
Edge$1.678B+21%n/aAllocation environment; richer configurations
Consumer$907M+39%+50%Holiday + premium mix; 50% YoY consumer growth
Data Center$440M+64%n/a2nd hyperscaler qualified PCIe Gen5 TLC
Total$3.025B+31%+61%All 3 segments accelerating simultaneously
Quality of Beat: Operationally clean across every dimension. The 2,120bp QoQ gross margin expansion is the largest single-quarter GM move in SanDisk's standalone history and one of the largest in NAND industry history (Samsung's 2017 Q4 expansion was comparable). Bit growth was in line with guide (+LSD QoQ) — the entire upside came from pricing, which is the structurally more valuable dynamic because it sustains. OpEx came in below guide ($413M vs $450-475M) due to non-recurring benefit from R&D management change, but underlying OpEx ratio dropping from 19.3% to 13.7% reflects revenue leverage. The print fully validates the structural NAND tightness framework established at Q1.

The Q3 FY26 Guide — The Most Consequential Forward Disclosure

MetricQ3 FY26 GuideMidpointStreet (pre-print)Above Street
Revenue$4.40-4.80B$4.60B~$3.00B+$1.60B / +53%
Non-GAAP GM65-67%66%~48-50%+1,500-1,700bp
Non-GAAP EPS$12-14$13~$5.50+$7.50 / +135%
Implied bit growth-mid-single QoQ (seasonal)n/an/aconservative on bits
Implied annualized EPS~$52n/an/an/a

The Q3 FY26 EPS guide is the largest above-consensus quarterly EPS guide in NAND industry history. Two consecutive quarters of triple-digit-above-consensus guides (Q2 was +94% above mid; Q3 +135% above Street) signal that the supply-demand inflection is more severe than any modeler has yet calibrated. The bit decline of -mid-single QoQ (typical Q3 seasonality is -12-14%) reflects management's strategic decision to under-ship and build inventory for the BiCS8 QLC Stargate ramp into Q4 and the recently signed multi-year customer commitments.

Key Topics & Management Commentary

Overall Management Tone: The most direct and explicit articulation we have heard from Goeckeler on the structural-vs-cyclical nature of the current cycle. The framing, "NAND is becoming a more durable, structurally attractive industry with higher average returns", is unprecedented language for a NAND vendor and represents a deliberate break from the historical "cyclical commodity" framing. Management is staking a position on multi-year structural margin elevation.

1. NAND as a Structurally Attractive Industry: The Re-Framing

The most consequential paragraph of the prepared remarks. Goeckeler explicitly framed NAND's evolution from a cyclical commodity industry to a structurally attractive industry with higher average returns. The framing matters because it sets the analyst modeling baseline at higher long-term margins (40-50%) rather than the historical 25-30% mid-cycle assumption.

"NAND is now recognized as indispensable to the world's storage needs, driving a foundational shift in how commercial relationships between suppliers and customers are structured… As a result, we believe NAND is becoming a more durable, structurally attractive industry with higher average returns."
— David Goeckeler, CEO

Assessment: This is the regime-change disclosure. If NAND's through-cycle margin profile re-rates from 25-30% to 40-45%, SanDisk's normalized earnings power transforms from $5-7 to $15-25. The Q2 print + Q3 guide are early data points supporting the framing.

2. Yokaiichi JV Extension Through 2034

SanDisk extended its joint venture with Kioxia through December 2034 (was expiring 2027). The Kitakami JV now has the same expiration date. As part of the extension, SanDisk agreed to pay $1.165B between CY2026-2029 for manufacturing services from Kioxia. The cost flows through COGS over the next 9 years (~$130M/year).

"Earlier today, we announced that we have reached an agreement with Kioxia to extend the Yokaiichi joint venture through 12/31/2034… The JV enables both companies to design and manufacture the highest performing, lowest cost nanotechnology that powers the world's infrastructure."
— Luis Visoso, CFO

Assessment: The Yokaiichi extension secures the most strategic 2027 overhang, the prior expiration was creating uncertainty about post-2027 supply structure. The $1.165B payment over 4 years ($130M/year COGS impact through 2034) is small relative to the strategic value of removing the supply structure uncertainty. This is a meaningful positive that supports the multi-year FCF outlook.

3. Multi-Year Customer Supply Agreements (LTAs)

LTA conversations advanced "across a broad base" of customers and geographies. The structure: shorter-term components are fixed-price; longer-term components are variable (capture upside if prices rise; customer protection if prices decline). Length, price commitment, volume commitment, and prepayment terms all being negotiated.

"We're seeing customers across end markets reach out to us across geographies. So this is not just a few. We're really seeing a broad base, which is it's very interesting for us. And we're making significant progress."
— Luis Visoso, CFO

Assessment: The "broad base" framing is critical. If LTAs were only at the top 3-5 hyperscalers, the structural margin story would be limited. The breadth, across multiple geographies and multiple customer types, supports the industry-wide pricing structure shift. The HDD industry took ~18 months from initial LTA discussions to a fully-LTA market structure; NAND is likely on a similar trajectory.

4. KV Cache Opportunity (NVIDIA Discussions)

Goeckeler disclosed initial estimates of the KV Cache opportunity: 75-100 additional exabytes of NAND demand in CY27 from AI inference workloads, with potential doubling in CY28. None of this is in current demand numbers.

"Our initial looks at it when we look at let's say, 2027 demand, we think, you know, that's you know, roughly maybe 75 to 100 additional exabytes. And then a year after, that you can you can double that."
— David Goeckeler, CEO

Assessment: The KV Cache opportunity is the most quantitative new demand vector disclosed on the call. 75-100 exabytes represents roughly 5-7% additional industry demand on top of already-elevated forecasts. Combined with the existing AI inference storage demand (already in numbers), the structural NAND tightness has another multi-year supply-side challenge.

5. Data Center as Largest NAND Market in 2026

For the first time in NAND industry history, data center is expected to be the largest end market (surpassing mobile) in 2026. The forecast revision sequence: mid-20% growth (3 quarters ago) → mid-40% (1 quarter ago) → high-60% (current), three consecutive quarterly upward revisions to CY26 data center exabyte demand.

"For the first time, data centers are expected to become the largest market for NAND in 2026. Driven by some of the world's largest and well-capitalized technology companies."
— David Goeckeler, CEO

Assessment: The data center becoming largest NAND market in CY26 is the structural shift that justifies the multi-year re-rate. Data center has higher pricing discipline, lower cyclicality, longer planning cycles, and higher gross margins than mobile or consumer NAND. The mix shift to data center alone supports through-cycle gross margin in the 40-45% range.

6. CapEx Discipline Maintained

Despite the most aggressive pricing environment in NAND history, management explicitly maintained the mid-to-high teens bit growth framework with no CapEx expansion. Capital allocation remains: invest in BiCS8 transition, then return cash to shareholders (no specific buyback announcement yet).

"We continue to be prudent and are not changing our capital spending plans, which support mid to high teens bit growth through the BiCS8 transition. Our investment posture remains focused on serving attractive sustained demand and healthy profitability levels."
— Luis Visoso, CFO

Assessment: Continued CapEx discipline is the key bull-case support. The structural margin re-rate requires no producer to expand capacity at peak pricing. SanDisk's second consecutive quarter of explicit CapEx restraint commitment sets the industry standard.

7. Enterprise SSD Acceleration

Data center revenue grew 64% sequentially to $440M, driven by completion of 2nd hyperscaler qualification on PCIe Gen5 TLC drives. Stargate (BiCS8 QLC) shipping for revenue "within next several quarters", implying Q3 FY26 to Q4 FY26 ramp. Working with 5 hyperscalers across the data center portfolio.

Assessment: Data center +64% QoQ on a quarter where mix was constrained by Edge/Consumer demand validates the share-gain narrative. The Stargate ramp into Q4 FY26 + sustained TLC qualification expansion supports data center revenue trajectory to $1B+ quarterly run-rate by exit FY26.

8. Pricing Mechanism Update

Pricing strengthened "during the quarter", December was meaningfully stronger than October/November. The acceleration was broad across all end markets (not concentrated in data center). Pricing visibility for Q3 FY26 is at the "double-digit increases" level.

"In December, we experienced a clear and significant improvement in conditions across end markets, which led to higher pricing."
— Luis Visoso, CFO

Assessment: December acceleration is the typical pattern when supply tightness becomes acute. The broad-based pricing increases (vs. concentrated in data center) reflect industry-wide allocation rather than customer-specific premium pricing. This is the strongest structural signal.

9. Net Cash Position $936M; $750M Debt Paydown

SanDisk closed Q2 with $1.539B cash and $603M debt = +$936M net cash. Paid down $750M of TLB during the quarter. The capital structure is now lean and positioned for capital return acceleration in H2 FY26 and FY27.

Assessment: Net cash of $936M plus Q2 FCF of $843M means SanDisk generated roughly $1B of distributable cash in a single quarter. Buyback authorization is a near-term catalyst, likely announced at Q3 FY26 or Q4 FY26 print.

Analyst Q&A Highlights

Pricing Sustainability and LTA Structure

The dominant topic on the call. The questioning pressed on whether multi-year agreements risk locking in pricing at exactly the wrong moment, with prices rising as fast as they are. Management's answer was that the terms are still being defined, and that the inbound interest is broad rather than concentrated.

Q: "How are you thinking about long-term agreements? Obviously, there's pros and cons in long-term agreements, because long-term agreements lock in the prices. When prices are going up so fast, you actually don't want so many long-term agreements, I guess. But I guess I'd just like to understand how you are thinking about that, how we should think about that in terms of your portion of agreements that are going longer-term, and how that may impact going forward."
— Mark Newman, Bernstein Research

A: "Yeah. I mean, David covered most of it. What I would say, Mark, is we're seeing customers across end markets reach out to us across geographies. So this is not just a few. We're really seeing a broad base, which is it's very interesting for us. And we're making significant progress. We're making significant progress with several of our customers who really want us to prioritize or assure supply to David's point, that they see that as a critical enabler for their business, and that's what they're looking for. Now to your point, we're being very thoughtful. On how do we define a few metrics. One is the length of the agreement, the price at which we will transact, the quantities, how much of our business we want to put in there, and any prepayment component of that. So we're being super thoughtful and this should be a value accretive and not the opposite."
— Luis Visoso, CFO

Assessment: Note what is disclosed and what is not. The five terms under negotiation are named explicitly (length, price, quantity, how much of the book to commit, and any prepayment), but no pricing structure is committed to and nothing is signed. The more useful signal is the breadth: inbound across end markets and across geographies says supply assurance has become a purchasing priority through the whole customer base rather than a hyperscaler-only concern. The visibility benefit, though, is a FY27 story. None of this quarter's revenue is contracted.

KV Cache Demand Quantification

A question on the AI inference and KV cache opportunity produced the most quantitative forward demand commentary on the call: 75 to 100 additional exabytes of 2027 demand, explicitly outside current forecasts.

Q: "At the Consumer Electronics Show, Jensen talked about this key value cache and, you know, gave some numbers in terms of, I think, terabytes per GPU. Seems like a pretty big market. Are you getting indications around that? Do you think there's should take that as kind of straight math? Does everybody have different implementations? And just the ramifications for what happens to data center NAND?"
— Joe Moore, Morgan Stanley

A: "Yeah, Joe. We're working through that right now. You know, we're working through it with NVIDIA and kind of how they're thinking about it. And, of course, then we'll work through it with our customers about how they're gonna configure it in deployments. So it's still a bit early. I'll say a couple things about it. First of all, none of that demand is in the numbers we're talking about… Secondly, our initial looks at it when we look at let's say, 2027 demand, we think, you know, that's you know, roughly maybe 75 to 100 additional exabytes. And then a year after, that you can you can double that. So it is a significant amount of demand."
— David Goeckeler, CEO

Assessment: The load-bearing phrase is that none of this demand is in the numbers. 75 to 100 exabytes is incremental to a 2026 data center forecast that has already been revised up twice, and the "double that" framing for 2028 implies another 150 to 200 on top. Treat it as optionality rather than base case until it enters the forecast, which management indicated happens in the back half of the year.

Supply Plans and Demand Visibility

Asked directly whether there was flexibility to add supply, management declined, and reframed the question as one about visibility rather than capacity. The data center exabyte forecast was revised up again this quarter, to high-60% growth for 2026 against mid-40% a quarter ago.

Q: "And any quick comments on how you're thinking about supply-demand longer term and any flexibility to add supply?"
— Mark Newman, Bernstein Research

A: "No. I mean, Mark, we've got our supply plans. We've been again, we've been very clear on what our CapEx plans are, what our bit growth plans are. That's what they are. It's about meeting our customers at that supply level and understanding how we allocate that. And then, as we said, it's about you know, all of us picking up our head and looking a little further out on the horizon. As to what demand is really gonna be in this market, and what sustained demand is going to be. And, you know, we just really need to get out of this idea that this is a transactional market where we only get a strong signal a quarter at a time."
— David Goeckeler, CEO

Assessment: A flat "no" on adding supply, in the strongest pricing environment this industry has seen, is the most important thing management said this quarter. Supply stays capped at mid-to-high-teens bit growth while the data center forecast has moved mid-20% to mid-40% to high-60% across three consecutive revisions. That widening gap is the thesis, and it is a choice rather than a constraint, which means it can be reversed. Watch for any softening of this answer.

Enterprise SSD Trajectory

Within the same response, management quantified the enterprise SSD ramp and guided to further acceleration through the back half of the fiscal year.

A: "I talked last quarter, we're gonna see growth of that throughout the fiscal year. We saw, I think, 29% sequential growth in the first fiscal quarter. Now we just saw 64% sequential growth in the second fiscal quarter, and I think you'll see that accelerate from here in the second half of the fiscal year."
— David Goeckeler, CEO

Assessment: 29% then 64% sequential, with acceleration guided from here, is the clearest evidence that the data center mix shift is a product cycle rather than a pricing artifact. Note the timing carefully on the QLC half of it: Stargate is described this quarter as expected to begin shipping for revenue "within the next several quarters," not next quarter, so the ramp through the back half of FY26 remains almost entirely TLC.

Market Reaction

  • Pre-print setup: Closed $539.30 on January 29, itself a 52-week closing high (range $29.62-$539.30). Up 127.2% year-to-date from $237.38 at 2025 year-end, against +1.8% for the S&P 500, and up 124.5% in the trailing 30 days.
  • January 30 session: Gapped open at $651.23, +20.8%, ran to $676.69 (+25.5%), then gave almost all of it back, trading as low as $533.00 (-1.2%) before closing at $576.25, +6.9% (+$36.95). Volume 40.9M shares vs. a 12.5M 30-day average (3.3x).
  • Sector read-across: Sharply negative. Memory peers were sold hard the same session: WDC -10.1%, STX -8.7%, MU -4.8%, against the S&P 500 at -0.4%.

The headline is +6.9%, but the session is the story. The market opened this print up 20.8% and spent the day selling it, closing barely a third of the way up from the low. A guide of $12-14 against a Street at $5.50 was met with a gap, a fade, and heavy volume: 3.3x average is distribution, not accumulation. Read alongside a memory complex down 5-10% on the same session, the most plausible reading is that the print landed into a sector being de-risked, and that holders used the gap as an exit rather than a confirmation.

We do not think that changes the fundamental call, but it does change what the setup is. This is no longer a market that will pay up on announcement. At $576.25 SanDisk trades at roughly 13x our updated FY27 EPS estimate, set out in Model Update Needed below, still inside the 15-25x band premium semiconductor franchises command, with an exceptional FCF yield. The structural re-rate is incomplete and roughly 20-50% upside remains to our fair value range. But the path there now runs through delivered quarters rather than through guides, because the tape just demonstrated it will fade a guide of this magnitude inside a single session.

Street Perspective

Debate: Is the Q3 Guide Sustainable Beyond Q3?

Bull view: The Q3 EPS guide of $12-14 reflects sustained NAND tightness through CY26 and emerging multi-year LTAs that lock in pricing. Q4 FY26 (BiCS8 mix expansion + Stargate revenue + sustained pricing) likely delivers $14-17 EPS; FY27 full-year EPS lands $40-50.

Bear view: The Q3 guide is the cyclical peak. Korean producers (Samsung, SK Hynix) will respond with capacity additions over 12-18 months. Pricing corrects 30-50% by FY28 mid-year. FY27 EPS lands $20-25.

Our take: The structural tightness extends 4-6 quarters beyond Q3 (through end CY26 and into H1 CY27). Korean CapEx response takes 18-24 months minimum to manifest. FY27 EPS in the $35-45 range is the realistic outcome. Even at bear-case $25 FY27 EPS, the stock at $576.25 trades 23x, which is no longer cheap on the bear's numbers. The bull-bear gap here is carried almost entirely by the estimate, not by the multiple.

Debate: Yokaiichi JV Extension Cost vs. Benefit

Bull view: The $1.165B payment over 4 years is small relative to the strategic value of supply continuity through 2034. Removing the 2027 expiration overhang strengthens FCF predictability and supports multi-year LTA conversations with customers who needed supply visibility.

Bear view: Paying Kioxia $1.165B at peak pricing represents Kioxia capturing some of the structural margin benefit. Could have been negotiated more aggressively given tightness.

Our take: Strategic positive. The JV continuity is more valuable than the $130M/year COGS impact. SanDisk avoids the disruption + uncertainty of post-2027 supply restructuring at the exact moment supply security is most valuable.

Debate: Capital Return Timing

Bull view: SanDisk should announce a $2-3B buyback in Q3 FY26 print, returning excess cash before the next debt paydown cycle. At the $576.25 post-print close, buybacks are still defensible given $40-50 FY27 EPS power, though the case is weaker than it was a quarter ago.

Bear view: Capital should be retained for either CapEx flexibility or potential M&A consolidation. Buybacks at peak-cycle pricing risk being mistimed.

Our take: Initial buyback announcement is appropriate at Q3 FY26 print. $1-2B initial authorization with measured deployment over 12 months balances shareholder return with optionality.

Model Update Needed

ItemPrior Model (Q1 FY26)Suggested ChangeReason
FY26 Revenue$11.5B$15.5BQ2 + Q3 guide running far ahead
FY26 EPS (non-GAAP)$10.50$28Q3 guide $12-14 alone + Q4 acceleration
FY27 Revenue$14B$22BSustained tightness + Stargate scale
FY27 EPS$18$45Margin step-up + Stargate ramp + LTAs
FY27 FCFn/a$9BFCF margin to ~40%

Valuation impact: Fair value range raised to $675-855 (~15-19x FY27 EPS). From the $576.25 post-print close that is +17% to +48% upside. Catalysts: Q3 FY26 print Apr/May 2026; buyback announcement; multi-year LTA contract signings.

Thesis Scorecard

Thesis PointStatusNotes
Bull #1: NAND structural tightness multi-yearStrengthenedQ3 guide validates 4-6 quarter visibility
Bull #2: CapEx discipline preserves pricingMaintainedMid-to-high teens bit growth cap reaffirmed
Bull #3: BiCS8 mix accelerates marginConfirmedGM 51.1% Q2; 65-67% Q3 guide
Bull #4: Data center largest NAND market 2026NEW — ConfirmedFirst time in NAND history
Bull #5: Multi-year LTAs locking in pricingEmergingBroad-base customer conversations advancing
Bull #6: Yokaiichi extension through 2034NEW — ConfirmedSupply continuity secured
Bull #7: KV Cache adds 75-100 EB CY27 demandNEW — ConfirmedNVIDIA discussions ongoing
Bear #1: NAND cyclical correction riskPushed outFY28 risk only; not FY26-27
Bear #2: Korean CapEx responseOpenWatch Samsung / SK Hynix 2H 2026

Overall: Thesis materially strengthened across all dimensions. Six bull points confirmed/strengthened (three new this quarter); two bears pushed out or open. The combination of Q3 EPS guide blowout + Yokaiichi extension + KV Cache disclosure + structural reframing makes this the most positive quarterly print SanDisk has delivered in its standalone history.

Action: Maintaining Outperform, upgrading conviction from moderate to High. Fair value $675-855. Post-print close $576.25. Continued upside on FY27 EPS revisions higher.

Independence Disclosure As of the publication date, the author holds no position in SNDK and has no plans to initiate any position in SNDK within the next 72 hours. Aardvark Labs Capital Research maintains a firm-wide policy of not trading any security we cover. No compensation has been received from SanDisk Corporation or any affiliated party for this research.