Strategy & Finance

SanDisk: How Much NAND Should Be Promised?

Topic
Pricing Strategy, Customer Strategy, Capital Allocation
Industry
Technology Hardware, Storage & Peripherals
Internal ID
S&F-STR-487
Source
Strategy & Finance
Publication date
Pages
6
License
Not specified

Download case PDF (6 pages)Request teacher’s pack

Student strategy case | Information cutoff: September 18, 2026, 11:59 p.m. ET

The decision

During FY2027, as of September 18, 2026, SanDisk’s CEO and board must set rules for signing or expanding New Business Model (NBM) agreements through FY2030. Should multi-year NBMs remain the predominant way SanDisk serves Datacenter and Edge customers? What share of expected output should be committed, and what customer-credit, price, volume, product-mix, and supply-readiness protections should apply before the company takes on more obligations?

This is a portfolio and governance decision, not a request to value SanDisk shares, forecast NAND prices, select a fab, or determine Flash Ventures’ undisclosed production schedule. Public evidence does not provide contract-level economics or give SanDisk unilateral control of its principal manufacturing joint venture.

A new company at a turning point

SanDisk became an independent, NAND-focused company in 2025. Its business spans Datacenter, Edge, and Consumer markets, with products and solutions built around NAND flash memory. NAND demand, prices, customer requirements, production yields, and product generations can shift over time. Management must decide how much supply to sell under multi-year commitments while also serving customers and markets that may value different products or delivery timing.

In FY2026, SanDisk reported revenue of $20.248 billion, compared with $7.355 billion in FY2025. Datacenter revenue was $5.153 billion, Edge $12.160 billion, and Consumer $2.935 billion. Fourth-quarter revenue was $8.965 billion, up 51% sequentially. Management attributed roughly one-third of that sequential growth to volume and two-thirds to price. Those results describe the whole business; they do not isolate NBM shipments, contract margins, or the causal effect of the new model.

Management’s strategic case is that AI and data-center workloads, customer qualification, and multi-year agreements can make demand more visible. An NBM may help a customer secure supply and coordinate product development while helping SanDisk plan production and investment. A more predictable demand base could matter in a capital- and technology-intensive business.

The counterargument is that current tightness and high prices may not persist. Multi-year terms can limit the ability to serve another buyer, capture later market prices, or adjust to new products and workloads. A contracted bit may not be interchangeable across TLC and QLC, product generations, specifications, yields, or delivery quarters. A customer’s purchase obligation is not proof that the customer has deployed the product or that SanDisk will collect all expected value.

What the public contract figures do—and do not—say

SanDisk’s August 5, 2026 Q4 earnings presentation described eight Datacenter and Edge NBM customers. Five agreements had been added since the five previously announced: three with new customers and two expanding existing NBMs. Three of the five were closed before the July 3 fiscal quarter-end and two afterward. The company reported a weighted-average contract duration above four years, with the longest up to five years.

The presentation reported expected coverage of about 50% of FY2027 bits and two-thirds of FY2028 bits. It showed $93.9 billion of minimum contracted NBM revenue at floor pricing, $59.8 billion of remaining performance obligations (RPO) at quarter-end, and $91.1 billion of RPO “including deals signed after.” These are different measures and dates. The slide also described $16.5 billion of financial guarantees in the form of cash deposits and instruments. It said agreements include fixed and variable price components, with floors and ceilings on the variable component, and define supply and demand commitments by year and quarter.

The figures do not form a simple bridge. RPO is transaction price allocated to remaining performance obligations; it is not cash received, revenue recognized, margin, guaranteed recovery, or proof of delivery. The $93.9 billion floor-price measure is a company-reported minimum revenue figure, not a measure of economic profit or cash certainty. The guarantees are not a public buyer-by-buyer schedule of recoverable value.

The August 17 Form 10-K, filed after Investor Day, reported that two NBMs signed after July 3 had an aggregate $31.3 billion transaction price. That subsequent-event disclosure does not make $31.3 billion equivalent to the $59.8 billion quarter-end RPO, the $91.1 billion presentation figure, or the $93.9 billion floor-price measure. Buyers, product and quarter volumes, pricing, remedies, and guarantee amounts by agreement were not disclosed.

The 10-K describes NBMs as requiring SanDisk to deliver and customers to purchase specified product volumes. It warns that capacity, yield, supply-chain, or timing problems could cause shortfalls or disputes; default or breach could leave the company with product that is difficult to resell at comparable prices. Guarantees may offset only part of lost revenue. The public filings do not establish that every agreement is “take-or-pay.” Contract-specific cancellation rights, default remedies, price formulae, and collateral terms are not public.

At July 3, SanDisk reported $1.242 billion of contract liabilities, primarily NBM advances, and $1.500 billion of refund liabilities, primarily refundable NBM security deposits. It also reported maximum potential proceeds of $5.0 billion from third-party collateral. These balances are not interchangeable: contingent collateral proceeds are not cash on hand, and advances and refundable deposits do not by themselves show contract profitability or ultimate recovery.

The supply behind the promises

SanDisk obtains most of its NAND through Flash Ventures with Kioxia. SanDisk reports 49.9% ownership and Kioxia 50.1%, with joint control; Kioxia operates the Japanese fabs. The parents are generally entitled to about 50% of output each. Product and timing interchangeability and detailed schedules are not publicly disclosed. The companies co-develop technology and transfer wafers at cost plus a small markup.

The arrangement brings scale, manufacturing expertise, and shared research and development. It also creates obligations. SanDisk funds its share of capital investment when venture cash flow is insufficient, shares R&D costs, pays half of fixed costs even when it does not purchase output, and uses a rolling forecast with orders in the near-term window binding and non-cancellable. The ventures’ governance and operating arrangements constrain unilateral changes in output, sourcing, timing, and product allocation.

The JV agreement was extended through 2034. SanDisk’s 10-K describes a $1.2 billion payment to Kioxia over 2026–2029 for manufacturing services and continued supply availability through 2034. The public disclosure does not state a corresponding quantity, unit economics, or NBM-specific capacity allocation. Kioxia and SanDisk announced a Japan investment plan exceeding $31 billion in August, including a Kitakami Fab3 plan with production planned for 2029. These are future plans, not FY2027 usable output or a commitment that SanDisk can redirect capacity unilaterally.

Technology announcements create both opportunity and execution questions. SanDisk and Kioxia announced BiCS10 TLC sampling, BiCS10 QLC, ninth-generation TLC sample shipments, and PCIe 6 enterprise SSD products during 2026. Sampling and roadmaps do not establish production yields at scale, customer qualification, deployment, or a right to substitute these products into existing contracts.

Competing signals in the market

There is evidence for durable demand and evidence for continued cyclicality. SanDisk reported sharply higher revenue and price, expanded Datacenter sales, and meaningful NBM coverage. Kioxia, Micron, SK hynix, and Samsung also described stronger AI/server demand or constrained supply in recent disclosures. For example, SK hynix said second-quarter NAND prices rose significantly and customer demand exceeded its supply capabilities.

Those signals do not establish that current conditions will hold for four or five years. Kioxia’s FY2025 report describes flash-price declines when capacity or customer inventory exceeds demand. It notes a production reduction from October 2022 through March 2024 and cautions that AI growth does not guarantee NAND demand because workloads may also use DRAM or HDDs and customer deployment can be delayed. Competitors’ periods, products, and customer agreements differ; their reports do not identify SanDisk’s NBM counterparties or prove the same demand trajectory.

SanDisk’s August 13 Investor Day release stated a FY2028–FY2030 framework of mid-to-high-teens revenue growth, approximately 80% non-GAAP gross margin, approximately 75% non-GAAP operating margin, and approximately 50% adjusted free-cash-flow margin. These are management targets, not achieved results or proof that NBMs alone will produce them. The full Investor Day presentation and recording were not accessible in the reviewed record. Detailed NBM metrics in this case therefore come from the accessible August 5 Q4 deck; August 13 claims are limited to that day’s official release.

Decision paths

Consider at least three defensible approaches:

  1. Keep NBMs predominant and continue signing selectively. Use multi-year agreements to support customer supply assurance and production planning, while requiring stronger buyer credit and defined protections before expanding commitments.
  2. Cap multi-year commitments and preserve market flexibility. Limit the share or duration of output committed, retaining more capacity for non-NBM customers or market-priced sales. This may preserve allocation and repricing options but could weaken customer assurance or planning visibility.
  3. Adopt conditional, product-specific gates. Approve commitments only when buyer credit, price bands, volume flexibility, product substitution, forecast reliability, and upstream supply-readiness conditions meet defined thresholds. Revisit the portfolio as demand, yields, product generations, and JV plans change.

The case does not supply an optimal contracted share, customer identity, probability of default, price forecast, margin, or usable-capacity estimate. Any sensitivity you construct should be explicitly illustrative. Do not treat total expected contract value, RPO, guarantees, or market forecasts as interchangeable measures.

Questions for discussion

  1. What decision must the board make now, and which parts of an NBM commitment are reversible or difficult to change?
  2. What share of expected FY2027–FY2028 output, if any, should be available for multi-year commitments? What evidence would change your limit over time?
  3. What protections should be required for customer credit, floors and ceilings, minimum and maximum volumes, delivery timing, product mix, qualification, and supply readiness? Which terms must be reviewed together?
  4. How should SanDisk test whether a buyer can and will perform without assuming undisclosed identity, deployment, or guarantee recovery?
  5. How do Flash Ventures’ fixed-cost, near-term order, investment, and joint-control arrangements affect customer commitments? What can SanDisk control, influence, or only monitor?
  6. Which indicators should trigger a pause, renegotiation, or expansion of NBM commitments? Identify the evidence owner, review frequency, and action associated with each trigger.

Selected exhibits

Measure Public disclosure by cutoff Interpretation
FY2026 revenue $20.248B; Datacenter $5.153B, Edge $12.160B, Consumer $2.935B Filed actuals; not NBM-only revenue or margin.
NBM customers / duration Eight customers; weighted-average duration above four years, longest up to five August 5 management presentation.
Expected bit coverage About 50% of FY2027 bits; about two-thirds FY2028 bits Management expectation; not shipment, customer deployment, or payment.
Minimum contracted revenue $93.9B at floor pricing August 5 management presentation; distinct from RPO and cash.
RPO $59.8B at July 3; $91.1B including later deals Distinct as-of dates and presentation scope; not margin or cash.
Later agreements Two post-July 3 NBMs; $31.3B aggregate transaction price August 17 10-K subsequent-event note; not directly additive to other measures.
Guarantees $16.5B, described as cash deposits and instruments Company presentation; buyer-level terms and actual recoveries unavailable.
NBM-related balances $1.242B contract liabilities; $1.500B refund liabilities; $5.0B maximum potential collateral proceeds Different accounting and contingent measures; not equivalent to guarantees or cash.
Flash Ventures SanDisk 49.9%, Kioxia 50.1%; joint control; generally equal output Filed JV structure; not unilateral capacity control or universal product fungibility.

Source notes

Cutoff discipline: This case uses information available through September 18, 2026, ET. It contains no later execution, results, or recommendation. Fiscal periods are labeled separately from calendar dates.

Back to strategy case studies