Sandisk (SNDK) stock analysis: full report, October 29, 2026

What the business is

SanDisk makes NAND—flash memory that stores data without power—for data centers, personal devices, vehicles and consumers. It sells drives, embedded memory, cards, USB drives and components to cloud operators, device makers, server makers, retailers and consumers. Fiscal 2026 revenue was $20.2 billion. Its categories were Datacenter, Edge and Consumer. Revenue comes from bits shipped, prices and mix toward enterprise SSDs, fast flash drives designed for data-center servers. No customer exceeded 10% of fiscal 2026 revenue; the ten largest accounted for 44%.

• Flash Ventures, SanDisk’s manufacturing partnerships with Kioxia, generally supply half the output. SanDisk pays half the fixed costs even when it buys less.

• More bits, higher prices and higher-value products can lift revenue; fiscal 2026 capacity sold grew by a mid-teens percentage.

• Eight customer agreements cover about half of fiscal 2027 bits and two-thirds of fiscal 2028 bits. Terms include fixed or variable prices, floors, ceilings and financial protections.

• The ten largest customers accounted for 44% of fiscal 2026 revenue, and no customer exceeded 10%.

The challenges

• The NAND pricing cycle. NAND prices can fall in oversupply while SanDisk’s shared manufacturing costs remain. Fiscal 2026 revenue rose 175.3%, but capacity sold grew by a mid-teens percentage. The risk is that price, not durable volume, drove the rebound.

• Cost per bit and technology ramps. Density gains matter only when new generations achieve good yields—the share of chips that work well enough to sell—and lower total cost. BiCS10’s density claims do not establish profitable volume production.

• Kioxia dependence and fixed obligations. Flash Ventures provides wafers and scale, but SanDisk shares investment and fixed costs even when demand weakens. The ventures run through December 31, 2034, leaving long exposure to NAND economics.

• Enterprise adoption and contract economics. Qualification—customer testing that confirms system compatibility—and shipments do not guarantee recurring volume or attractive margins. Contracts cover bits, not revenue or profit; enterprise SSD adoption must become durable economics.

The numbers

The figures show a sharp recovery from losses: fiscal 2026 operating margin was 61.2% after three consecutive loss-making years. Revenue rose 175.3%, while capacity sold grew by a mid-teens percentage. SanDisk’s real free cash flow reached 55.6% of revenue, making cash generation the key test of durability.

How it earns, and what can hurt it (US$ millions)

Operating margin — The $1.8 billion goodwill write-down caused FY2025’s operating loss; without it, operating margin was 6.2%, compared with −18.7% reported.

Real free cash flow, % of revenue — FY2026 received $2.5 billion in customer advances and deposits; without that cash, free cash flow after stock pay was 43.4% of sales.

What one share gets (US$ per share, except percentages)

The verdict

The share price has steep downside if fiscal 2026’s exceptional profits normalize; its path hinges on enterprise demand and customer commitments sustaining margins as NAND supply responds.

Management guides the quarter to January 2027 to revenue of $12.2–$12.6 billion and non-GAAP diluted EPS of $55.00–$57.00 on its stated share-count basis. For fiscal 2028–2030, it models mid-to-high-teen revenue growth and 80% non-GAAP gross margin. It also models 75% non-GAAP operating margin and 50% adjusted free cash flow margin.

Weighted price is $369.91 in October 2027, a +122% return on a consistent post-split basis. It is $270.32 in October 2029, a +63% return on a consistent post-split basis. The bear case carries 35% probability; the base case carries 50%. The bull case carries 15%.

Fiscal 2026 buybacks were $29.19 per share against $1.50 of stock pay; future cash must support repurchases without weakening resilience. January 2027 results are the first checkpoint: revenue and margins must hold against guidance, while shipped volume must show more than pricing. Through fiscal 2027–2028, contracted bits must become deliveries and recurring enterprise SSD sales, and BiCS8 must ramp without cost or yield setbacks. By 2028–2029, lower cost per bit and durable cash generation support the bull case; oversupply, weaker purchases or slower data-center construction favor the bear case.

SanDisk’s valuation requires volume, lower cost per bit and cash generation to outlast the NAND cycle.

SanDisk Corporation sells the digital storage that lets devices and data centers keep information when power is off. Its business is easiest to understand as a chain: it obtains memory, turns it into components and drives, sells them through business and retail channels, and pays to keep that chain running.

1.1 — What it is and what it sells

SanDisk sells electronic storage for personal devices, vehicles, businesses and large data centers.

Its largest reported customer-use category is Edge, which covers phones, PCs, cars and other devices. In the fiscal first quarter of 2027, ended October 2, 2026, Edge revenue was $5.8 billion. Datacenter revenue was $4.8 billion in the same quarter, while Consumer revenue was $397 million.

Edge is not a product. It is a shelf label for where storage is used; an SSD, by contrast, is a product. An SSD is a solid-state drive with no spinning disk. A PlayStation 5 owner, for example, can install a WD_BLACK SN850P SSD in the console’s expansion slot to add space for games.

SanDisk also sells storage built into devices, removable memory cards, USB drives, business drives and memory components. Its brands include SanDisk, WD, WD_BLACK and SanDisk Professional. WD is now a separate company, but SanDisk continues to sell WD-branded flash products.

SanDisk reports the business as one Flash segment, so investors cannot see exact revenue or profit for each product family. In fiscal 2026, that segment produced $20.2 billion of revenue. Its operating income was $12.4 billion, before interest and tax.

The company’s reported segment view is therefore broad: it shows customer areas, not a complete product ranking.

Results by segment, as the company reports them (US$ millions)

1.2 — How it makes money

SanDisk makes money by turning NAND flash memory into components and finished storage products.

Flash memory is rewritable electronic storage that keeps its data without power. NAND is the type of flash memory at the center of SanDisk’s products. When a phone maker buys a storage module to build into a phone, that is embedded storage: SanDisk sells the component, while the phone maker sells the finished device.

SanDisk obtains its flash wafers through Flash Ventures, its joint ventures with Kioxia, a memory-manufacturing partner. A wafer is a thin silicon disc on which many memory chips are made. Kioxia owns the Japanese factory buildings, while the ventures own or lease much of the equipment. SanDisk generally takes half the output and pays its share of fixed costs, even when it buys less.

That arrangement resembles sharing a bakery: fewer loaves do not eliminate the rent or the cost of the ovens. SanDisk also pays variable manufacturing costs, equipment costs, controllers, assembly, testing, engineers and sales staff. Strong prices can therefore produce powerful profits, but weak prices leave much of the cost in place.

Cash and revenue also arrive at different times. At July 3, 2026, SanDisk had $1.24 billion of contract liabilities, mainly customer advances, and $1.5 billion of refund liabilities, mainly returnable security deposits. It also reported $59.8 billion of remaining performance obligations, but expected only about 19% to become revenue within twelve months.

An advance is cash received before SanDisk delivers what it promised. A deposit may have to be returned. Revenue is generally recognized when products are delivered under the shipping terms, so cash, promises and earned sales are different events.

1.3 — Who pays

SanDisk’s customers range from manufacturers that build devices to retailers that put storage in consumers’ hands.

PC makers can buy drives for computers, while phone and vehicle makers can buy embedded memory. Server makers and large cloud companies buy storage for data centers. These customers often involve engineering teams as well as purchasing departments because the storage must work reliably inside particular systems.

Retail follows a different path. Distributors, retailers and online stores buy cards and drives for resale, and SanDisk also sells through its own online store. Reseller agreements can include incentives and limited price protection, so the first invoice is not always the final economic price.

The customer base is broad but still concentrated. No customer accounted for more than 10% of fiscal 2026 revenue. The ten largest together accounted for 44%.

Receivables are unpaid bills for sales already made. At year-end, two customers represented approximately 19% and 12% of receivables, respectively. A third represented approximately 10%. That exposure can look more concentrated than annual revenue if a large invoice was issued shortly before year-end.

SanDisk assigns revenue to the customer’s ship-to location. The regional figures therefore show where goods were sent, not necessarily where the final user lived.

The table lists revenue for the United States, China, Hong Kong, Rest of Asia, Europe, Middle East and Africa, Other, and Total.

Revenue by region, as the company reports it (US$ millions)

1.4 — The push into data-center and enterprise storage

SanDisk is trying to sell more complete, high-capacity drives to companies that run servers and data centers.

An enterprise SSD is a solid-state drive built for those machines rather than for one person’s laptop. SanDisk’s SN670 is one example. Its QLC memory stores four bits in each memory cell, which increases capacity, although buyers must weigh that against write performance and endurance.

Large customers do not simply plug in a new drive and order thousands. They conduct qualification: testing compatibility, speed, power use, reliability and failure behavior in their own systems. A server maker may run another qualification before offering the drive to its customers.

SanDisk said its high-performance TLC drives had completed qualification at a second hyperscaler, meaning a very large cloud company. It also said its QLC Stargate platform had begun shipping for revenue. Those milestones open the door to volume sales, but SanDisk has not disclosed customer names, product-specific volumes or pricing.

The opportunity is greater value per product because SanDisk combines memory, controllers and software into a system. The risks are equally clear: development and qualification cost money, a few large buyers have bargaining power, and approval does not guarantee immediate volume.

1.5 — Longer customer agreements

SanDisk’s longer agreements trade some market uncertainty for more predictable supply commitments.

Management said that, at its August 13, 2026 investor day, agreements with eight customers covered approximately half of fiscal 2027 bits and two-thirds of fiscal 2028 bits. A bit is a unit of stored information, so this measures committed storage capacity rather than guaranteed revenue or profit.

The agreements help SanDisk plan factory capacity and equipment. Management described an average contract length of more than four years and a longest contract of five years. Some contracts have fixed prices; others have variable prices bounded by floors and ceilings.

A price floor limits how far the selling price can fall. A ceiling limits SanDisk’s benefit if market prices rise sharply. The contracts can also include financial protections if customers fail to meet purchase obligations.

That protection reduces uncertainty, but it does not eliminate risk. Customers may need less storage, prices may move beyond expectations, costs may rise, or either side may fail to deliver. The important question is therefore not just how long a contract lasts, but how much the customer must buy, at what price and with what remedies.

1.6 — Memory cards and USB flash drives

Memory cards and USB flash drives give people a simple way to add or move storage.

A memory card fits into a compatible camera, phone or gaming device. A USB flash drive plugs into a compatible port. Buyers choose based on capacity, speed and physical fit: capacity determines how much data the product holds, while speed affects how quickly files move or are recorded.

These products also show the importance of distribution and compatibility. Retailers and online stores make them easy to find, but consumers can compare alternatives readily, so price competition remains strong.

In October 2026, SanDisk announced that it would be the exclusive storage partner for Meta VR Glasses. Meta, the technology company, had tested microSD Express cards and USB-C flash drives planned for the glasses’ spring 2027 launch. That announcement represents a future product commitment, not sales already made.

The cards and drives are the most familiar part of SanDisk’s business, but the larger story is industrial: SanDisk buys or jointly produces memory, adds engineering and manufacturing work, sells storage into many kinds of systems, and carries fixed costs while it waits for customers to buy.

SanDisk Corporation sells storage built from flash memory to cloud operators, device makers and consumers. Its market is moving toward data centers, but its fortunes still depend on a cyclical manufacturing industry.

2.1 — The market it sells into

SanDisk sells flash storage in several forms to three broad customer groups. NAND is the memory chip that keeps information when power is off, while flash is the broader name for this persistent memory. An SSD combines NAND with a controller and firmware to become a complete storage device; a memory card performs the same basic job in a smaller consumer product.

The field is large, though no single figure captures every finished storage product. The five largest publicly listed NAND suppliers generated $68.87 billion of NAND revenue in the April–June 2026 quarter. That is quarterly supplier revenue, not the annual value of every storage device sold worldwide.

SanDisk divides its fiscal 2026 revenue into Datacenter, Edge and Consumer. Datacenter supplied about 25%, Edge about 60% and Consumer about 14%. Edge includes flash for products such as phones and PCs; Consumer includes retail cards and drives.

Datacenter’s importance is rising quickly. It supplied about 13% of fiscal 2025 revenue, then about 25% in fiscal 2026, and about 44% in the fiscal first quarter of 2027, ended October 2, 2026. That change reflects both more capacity sold and higher revenue per unit, so the next question is whether data-center demand will last.

2.2 — Where the market is heading

Flash demand is shifting from personal devices toward data centers, especially those built for artificial intelligence. SanDisk estimates that data centers consumed roughly half of industry flash capacity in 2026, up from 30% in 2025. It expects the share to exceed 55% in 2027.

AI needs storage because every interaction creates information that must be stored, retrieved and served quickly. Working memory is like a small, fast kitchen counter; flash is like a large refrigerator. During inference—the running of a trained model—an AI system can retrieve documents, earlier conversation material or generated content from SSDs, then move what it immediately needs into faster memory.

That demand can recur beyond the construction of a new AI facility. But consumer demand has weakened as higher storage prices make phones and PCs more expensive, which can delay replacement or push buyers toward less capacity.

The investment question is therefore whether inference creates durable storage use or whether current demand mainly reflects a construction boom. SanDisk forecasts data-center bit demand growing about 40% in calendar 2027 after more than doubling in 2026, but future orders remain exposed to inventory, software efficiency and slower construction.

2.3 — The rivals

SanDisk faces different rivals depending on whether a customer is buying memory chips, enterprise SSDs, hard-disk capacity or a retail product. Samsung is the leading NAND supplier by revenue, while SK hynix competes through its memory business and Solidigm, its enterprise-SSD operation. Micron also makes memory and sells to data-center and device customers.

Western Digital and Seagate principally offer hard-disk storage. Hard disks can compete with flash for data-center budgets when capacity matters more than speed, so SanDisk does not face only other NAND manufacturers.

Kioxia is both a manufacturing partner and a rival. The companies share flash technology and production, then can sell competing products to the same customer. YMTC adds pressure through high-volume manufacturing, though its business remains more concentrated in consumer products and ranks lower by revenue. On retail shelves, Kingston, ADATA and Transcend compete for the shopper even when another company made the underlying NAND.

The table compares the listed and private rivals by business revenue, margin, market value and valuation.

The rivals

2.4 — What protects it

SanDisk’s protection comes from manufacturing scale, technical knowledge and customer relationships, but much of its manufacturing advantage is shared with Kioxia. A new rival would need expensive equipment, a reliable process, experienced engineers and enough customers to absorb high-volume output. One reported new-plant proposal put the investment at $45 billion.

SanDisk and Kioxia jointly develop flash technology and operate manufacturing ventures in Japan. Kioxia owns the facilities and provides manufacturing services, while SanDisk holds a 49.9% position and jointly controls the ventures. Each partner generally has the right to buy half of the output.

SanDisk also adds controllers, firmware and complete-system design. That matters because a cloud customer needs a drive that performs reliably inside its servers, not merely a dense memory chip.

The partnership brings risks as well as protection. SanDisk cannot direct the factories alone, must pay half of certain fixed costs even if it buys less output, and faces restrictions on using third parties while the ventures operate. Its brand and tested products help retain customers, but many products remain largely interchangeable with competitors’ products.

2.5 — The memory supply, demand and pricing cycle

NAND is cyclical because factories expand slowly while customers can change orders and inventories quickly. A bit is a unit of digital information, and bit supply means the storage capacity manufacturers make available. The NAND cycle is the repeated swing between shortage and oversupply; ASP, or average selling price, is revenue divided by capacity shipped.

When buyers need more capacity than factories can provide, suppliers gain pricing power. When customers have stocked up or producers add too much output, prices fall. ASP can also rise because sales shift toward expensive enterprise SSDs, even if the price of an identical chip does not change.

SanDisk’s fiscal 2026 revenue rose 175% while its total capacity sold rose by a mid-teens percentage. In its latest quarter, about 60% of the sequential revenue increase came from volume and 40% from higher pricing. Both helped, but neither alone measures the price change for every product.

Wafer starts count wafers entering production, while utilization measures how much available capacity is operating. SanDisk estimated that the industry had reduced wafer capacity by about 30% from its peak yet still achieved mid-to-high-teens bit growth through denser technology. Inventory, technology transitions and actual shipments therefore matter as much as factory counts.

2.6 — Competition and customer adoption

Winning a data-center customer requires approval for a complete drive, not merely a promising flash chip. The customer tests performance, power use, reliability and compatibility with its servers. Qualification and initial shipping are encouraging, but they do not prove a large, recurring order.

SanDisk said its enterprise SSDs had been qualified at major cloud customers and equipment makers and were shipping. It also said its high-capacity Stargate platform had begun shipping for revenue. Those statements show progress, but public disclosures do not establish each customer’s volume or repeat purchases.

SanDisk’s technology must still compete against Samsung, SK hynix and Micron in each customer’s system. A higher layer count or a claim of leadership does not automatically mean the lowest cost per bit or the best performance for every buyer.

Long-term agreements can make adoption more durable. Eight agreements cover more than half of expected fiscal 2027 bits. They cover two-thirds of expected fiscal 2028 bits and extend up to five years. Floors, ceilings and guarantees help secure supply and protect SanDisk if a customer buys less than promised, but they do not eliminate manufacturing costs or weak demand.

2.7 — NAND bit shipments

Bit shipments show how much storage capacity SanDisk sells, separate from the number of drives or customers. One exabyte equals one billion gigabytes, so a drive with twice the capacity contributes twice as many shipped bits.

In fiscal 2026, SanDisk’s total capacity sold grew by a mid-teens percentage. Datacenter capacity rose almost 120%, while Edge capacity grew by a high-single-digit percentage. Consumer capacity fell by a mid-teens percentage, showing how sharply the company was reallocating supply toward data centers.

Bit mix and revenue mix describe different things. Data centers represented 38% of SanDisk’s portfolio when fiscal 2026 ended, while Datacenter supplied about 25% of revenue across the full year. The first is a point-in-time capacity measure; the second is a period-wide dollar measure.

To judge SanDisk’s market position, compare its bit growth with independent industry growth over the same period. Then ask whether the difference reflects customer wins, scarce-supply allocation or denser technology. Capacity growth is useful evidence, but it does not by itself prove lasting end-user demand.

Sandisk Corporation makes memory that stores data after power is turned off, but its financial story is also a story about factories, partnerships and timing. It spent heavily to secure technology and supply, then used a sudden cash windfall to repay debt and buy shares; whether that worked depends on which decision we measure.

3.1 — How it got here

Sandisk grew from a flash-memory pioneer into a storage company, then became independent again.

Eli Harari, Sanjay Mehrotra and Jack Yuan founded SunDisk in 1988. Flash memory gave portable devices storage that survived a power loss, and the company’s 1995 Nasdaq listing provided public capital as digital cameras and similar products created demand for removable storage.

Manufacturing memory chips required much more investment than designing cards. Sandisk’s partnership with Toshiba, begun in the late 1990s, shared factory and technology costs; Toshiba’s memory successor is Kioxia.

Sandisk then pursued business storage, including its 2014 purchase of Fusion-io. Western Digital, a larger storage company, bought Sandisk in May 2016 to combine flash technology with its hard-drive business.

That arrangement ended when Western Digital separated the flash business on February 21, 2025. Sandisk began independent Nasdaq trading on February 24, and Western Digital shareholders received Sandisk shares. Because older accounts were prepared inside Western Digital, their costs, debt and per-share figures do not always describe the independent company directly. That history explains why the next question is not simply what Sandisk bought, but what those decisions produced.

3.2 — The big decisions

Sandisk’s major decisions bought capabilities and supply, but the returns have been uneven.

In 2014, Sandisk bought Fusion-io for about $1.1 billion net of cash acquired. Fusion-io brought hardware and software that placed flash storage close to a server’s processor. In 2015, Sandisk cancelled one acquired research project and wrote off its $61 million accounting value. That shows one project failed, but it does not measure the whole acquisition’s return.

The Toshiba relationship also exposed a governance risk. After Western Digital acquired Sandisk, the companies fought over technical databases, engineering wafers and investment in a new factory. They settled in December 2017 and extended their collaboration, preserving the supply relationship while showing that shared ownership can create conflict.

In 2024, Sandisk sold 80% of its interest in SDSS, a Chinese flash-product manufacturer, to JCET. The deal produced $659 million of pretax proceeds and a $34 million accounting gain, but Sandisk kept 20% and a $550 million minimum annual purchase commitment. It raised cash without ending the commercial relationship.

Goodwill is the accounting value of expected advantages such as customers, talent or technology above identifiable assets. Sandisk impaired goodwill by $671 million in fiscal 2023 and $1.83 billion in fiscal 2025. These were not new cash payments, but they suggest that some inherited value no longer supported its recorded price.

In March 2026, Sandisk invested about $970 million in Nanya, a Taiwanese memory maker, alongside a multiyear supply agreement. Nanya supplies DRAM—fast working memory that helps an SSD controller organize data—and Sandisk recorded an $807 million unrealized gain on the shares in fiscal 2026. The investment may secure an important component, but its market value can also make earnings swing independently of storage sales.

3.3 — Where the cash went

Sandisk moved from several years of cash consumption to a fiscal 2026 cash surge, then directed much of that money toward debt repayment and buybacks.

Free cash flow was positive in fiscal 2022 but negative in each of the next three years. Those pre-separation figures include Western Digital’s cash management, so they cannot be treated as a pure record of independent Sandisk choices.

Sandisk also invested in its own equipment, while its Kioxia ventures bought and operated much of the manufacturing capacity. That spending does not appear simply as Sandisk-owned factory investment, which makes the partnership central to understanding where the capital went.

At separation, Sandisk issued debt and paid cash to Western Digital. Later, stronger operating cash enabled it to retire the term loan and launch a large share-buyback program. Buying shares leaves remaining shareholders with a larger ownership fraction, but it also removes cash that might be needed when memory prices fall.

Customer advances complicate the picture. They resemble paying a builder before construction: Sandisk receives cash today but owes future deliveries. The fiscal 2026 inflow therefore improved liquidity without representing only completed sales.

This table tracks free cash flow, payouts, acquisitions, asset sales, investments, new shares and debt across the five fiscal periods.

Follow the money: where the free cash flow went (US$ millions)

The cash was real, but its durability depends on future deliveries, factory commitments and the next memory downturn.

3.4 — The Kioxia manufacturing and technology partnership

Kioxia gives Sandisk a large shared manufacturing base, but it also makes some costs difficult to avoid.

Kioxia is the Japanese company descended from Toshiba’s memory business. Together, Kioxia and Sandisk develop and manufacture NAND, the flash memory that stores files in products such as solid-state drives. Their factories operate through ventures called Flash Ventures.

Sandisk is generally entitled to buy 50% of the ventures’ output, while paying half their fixed costs whether it uses all that capacity or not. Orders covering up to three months are binding, so a sudden fall in demand cannot immediately remove the obligation.

In fiscal 2026, Sandisk’s net payments for wafers and loans to Flash Ventures were $3.6 billion. It must also help fund roughly half of venture capital investment when operating cash is insufficient, guarantee some equipment leases and meet a $138 million joint-research commitment for fiscal 2027.

The downside appears in under-used capacity: Sandisk recorded $249 million in fiscal 2024 and $11 million in fiscal 2026. At July 3, 2026, it estimated maximum reasonably measurable exposure of $2.897 billion under the venture arrangements. That is not an immediate bill, but it shows the scale of the commitments.

The partners extended all three ventures through December 31, 2034. The arrangement supplies scale and technology, yet its success still depends on keeping fixed costs and new investment aligned with demand.

3.5 — The 2025 separation and standalone costs

The separation made Sandisk easier to examine, but its earlier accounts are not a clean standalone baseline.

A carve-out removes a business from a parent’s accounts and assigns it portions of shared costs such as finance, legal and information technology. A pro forma statement then models what the business might have looked like independently. Sandisk’s model showed a $826 million net loss and a $5.65 diluted loss per share for fiscal 2024, but it warned that the result was not necessarily what an independent company would have earned.

The separation changed ownership and financing. On February 21, 2025, Western Digital distributed 80.1% of Sandisk’s shares, giving each Western Digital shareholder one-third of a Sandisk share for every Western Digital share owned. Sandisk also borrowed $2 billion and paid approximately $1.5 billion to Western Digital, beginning as an independent company with its own balance sheet and debt.

Sandisk repaid the term loan on March 4, 2026. The repayment improved flexibility but created an accounting charge for unamortized financing costs.

Actual transition-service expense fell from $5 million in fiscal 2025 to $2 million in fiscal 2026. That decline does not mean corporate costs disappeared; some functions simply moved into Sandisk’s own expenses. The fair conclusion is that separation clarified the company’s future, while making comparisons with its past more demanding. Sandisk’s plan worked best when it secured scarce capability and converted cash into flexibility; it worked less clearly when value depended on an acquisition, a full memory cycle or a high future demand forecast.

Sandisk is a flash-memory company whose fortunes depend on factories, executives, directors and shareholders that do not all pull in the same direction. To decide whether they are on an outside shareholder’s side, follow the chain from management to oversight, ownership, trading and pay.

4.1 — Who runs it

David Goeckeler gives Sandisk continuity, while Luis Visoso brings finance experience to a business built on costly manufacturing commitments.

Goeckeler led Western Digital, Sandisk’s former parent, after joining it from Cisco’s networking and security business. He therefore knows the flash-memory operation both as part of a larger storage company and as the leader of an independent business.

Sandisk designs and sells storage products, but Kioxia, its manufacturing partner, makes the memory wafers inside them. A wafer is a round sheet of silicon on which many chips are produced. NAND flash memory keeps data after power is turned off. Sandisk obtains all its flash memory through joint ventures with Kioxia, owns 49.9% of those ventures and generally receives half their output. The partnership accounted for 29% of global NAND bit output in the fiscal year ended March 31, 2025. Its extension through December 31, 2034, including $1.2 billion in payments during 2026–2029, makes factory access a central management responsibility.

Visoso became finance chief when Sandisk separated. His earlier roles at Unity Software, Palo Alto Networks and Amazon Web Services give him a finance background beyond memory chips. At Sandisk, that work includes funding, supply allocation and pricing.

The leadership is experienced, but succession remains a question. The board says it reviews leadership development and an emergency CEO plan; outside shareholders cannot see how ready particular successors are.

The officers table lists the named officers, their roles and when they began serving.

Who runs it

That makes the board—not just the executives—the next test of who is accountable.

4.2 — The board

Sandisk’s board has a strong independent majority, but the CEO also chairs the board that oversees him.

Seven of eight directors are classified as independent under listing rules. That means the board has judged that their relationships should not compromise independent judgment. It does not guarantee that they will challenge management when doing so is difficult.

The board includes useful industry experience. Rick Cassidy previously led TSMC’s Arizona operation, and Tom Caulfield formerly ran GlobalFoundries. Alexander Bradley adds finance and semiconductor experience. Such knowledge helps directors assess manufacturing commitments, capital spending and a cyclical memory market.

Cassidy is the lead independent director. He chairs meetings of independent directors without management, helps coordinate their agendas and can speak with shareholders. Those powers provide a counterweight, but they do not make him the independent chair. Goeckeler still controls the board chair as well as the executive role.

The combined roles are common: in 2025, the current CEO also chaired the board at 42% of S&P 500 companies. The important issue is therefore not simply structure but conduct—whether directors challenge compensation, review major commitments and prepare seriously for succession.

The board table shows each director’s service date and independence classification.

The board

The next question is whether the owners behind those votes can use this machinery.

4.3 — Who owns it

Sandisk has substantial institutional ownership but no disclosed controlling shareholder.

The company has one class of common stock, with one vote per share. BlackRock and Vanguard Capital Management are among the largest reported holders, but neither controls the company. A beneficial owner may be an asset manager acting for clients, so a reported position does not necessarily mean the manager receives all the economic benefit.

Ownership filings also distinguish voting power from dispositive power, the ability to decide whether shares are bought or sold. FMR and Abigail Johnson reported interests in the same underlying shares, so adding their entries would count one block twice.

Goeckeler has a meaningful personal stake, which ties part of his wealth to the common stock but does not give him control. That is alignment, not ownership of the company.

Western Digital reduced its stake through secondary offerings and a debt-for-equity exchange. In a secondary offering, an existing shareholder sells shares to the public and keeps the proceeds; the company does not raise new money. Sandisk’s registration agreement required it to bear the offering costs, estimated at about $4.7 million for the June 2025 and February 2026 transactions. Western Digital held no Sandisk shares by July 3, 2026, removing that resale overhang.

A minority shareholder can vote for directors and express a view on executive pay, but cannot alone choose management or direct operations.

The ownership table lists principal holders, insiders, share classes, values, recent transactions and reporting dates.

Who owns it: insiders and principal holders

Those owners can vote, but trading by insiders offers another clue about alignment.

4.4 — Insider buying and selling

Sandisk insiders have sold substantial amounts and reported no open-market purchases in the measured period.

Officers and directors sold 88,895 shares for about $133.8 million in open-market transactions during the reported 12 months. Goeckeler accounted for 67,679 shares, sold for about $105.0 million. That is meaningful selling, but it does not by itself show that insiders expect the business to deteriorate.

A Rule 10b5-1 plan is a trading arrangement established in advance. It specifies future sales or gives a broker a formula for making them. Current rules impose a cooling-off period, so a sale under such a plan is weaker evidence of a last-minute bearish decision.

Goeckeler’s September sale filing identified a plan adopted on May 29. Visoso’s September sale also used a plan adopted earlier. The plan does not explain why either executive wanted to sell, but it separates the later trade from the earlier decision to arrange it.

Transaction type matters. When an award vests, Sandisk may withhold shares to cover the employee’s taxes. That is different from an open-market sale chosen for investment reasons.

The final test is what the board promises leaders—and what those promises cost owners.

4.5 — Pay, and promises kept

Sandisk’s pay plan rewarded an extraordinary first full year, but its largest stock awards could still dilute shareholders.

The company pays leaders through salary, an annual cash incentive and stock awards. In fiscal 2026, the annual incentive weighted profit at half, cash flow at a quarter and strategic goals at a quarter. Those goals included reducing debt and building the data-center business. The named executives received the maximum annual payout; Goeckeler’s was $4.725 million. His reported total compensation was about $15.2 million, an accounting measure that includes grant-date stock values.

Management exceeded the incentive plan’s profit target by a wide margin. The maximum target for adjusted operating income was $1.508 billion, while Sandisk achieved $12.7 billion under that measure. Sandisk also eliminated its debt by fiscal year-end. Strong memory pricing helped, so the result reflects both management and an unusually favorable market.

The launch awards are more consequential for owners. They promised additional shares if Sandisk’s stock rose substantially after separation and reached the maximum payout of 300% of target. The awards still require continued service through 2028. Their reported potential value can rise with the stock, but it is not cash already received.

The recurring long-term plan uses revenue and adjusted earnings-per-share goals averaged over three years. Its first year reached 200% of target. For fiscal 2027, the board added performance against semiconductor peers, which may better separate company execution from an industry-wide rise.

The stock-pay table compares compensation expense, employee-award cash costs, taxes, buybacks and the changing share count.

Stock pay and the share count (US$ millions)

The evidence is mixed: Sandisk’s leaders are tied to shareholders through ownership and performance pay, but outside owners still depend on the board to turn strong results into durable returns without excessive dilution.

Sandisk Corporation makes NAND, the electronic storage used in solid-state drives, phones and memory cards. Its accounts describe a powerful recovery, but they also show why revenue, profit, cash and debt must be read together.

5.1 — Growth

Sandisk’s sales rebound is much larger than its increase in storage shipped.

Revenue rose 175.3% to $20.2 billion in fiscal 2026 after a severe downturn. Sandisk sells to data-center operators, device makers through its Edge business, and consumers. In fiscal 2026, Edge was the largest business at $12.2 billion, while Datacenter reached $5.2 billion.

The difference between dollars and physical output explains the apparent contradiction. Bits sold measure storage capacity; revenue per gigabyte measures the average dollars earned for that capacity. If Sandisk sells only modestly more capacity but earns much more for each gigabyte, revenue can surge without a comparable increase in factory output.

That is what the mix suggests. Total bits rose by a percentage in the mid-teens, while Edge bits rose by only a high-single-digit percentage and revenue per gigabyte rose almost 180%. Consumer bits fell by a percentage in the mid-teens even as consumer revenue increased. The industry forecast for calendar 2026 is bit growth in the low-20s percentage range, so Sandisk’s own output growth deserves separate attention from its sales growth.

The fiscal fourth quarter made the distinction especially clear: management attributed about one-third of the sequential revenue increase to volume and two-thirds to pricing. Higher prices may reflect stronger demand, but revenue per gigabyte also rises when customers buy more expensive enterprise products and fewer low-priced consumer products.

EPS adds another layer. It divides profit by shares, and Sandisk used an assumed 145 million shares for periods before separation. Later stock awards and repurchases changed the denominator, so EPS is a measure of profit per owner, not of factory output alone.

The five-year results table shows the movement in revenue, margins, profit, EPS and dividends.

Results (US$ millions)

Operating income — FY2024 included a $60 million property-sale gain; without it, the operating loss would have been $528 million.

The company’s own measures table separates Datacenter, Edge and Consumer across the same years.

The company's own measures (US$ millions)

5.2 — Profit

Sandisk’s margins show how a change in memory prices can produce a much larger change in profit.

Gross margin rose from 7.1% in fiscal 2023 to 71.5% in fiscal 2026. Gross margin is the share of sales left after the direct cost of the products sold. Because a semiconductor factory costs money even when it is not fully used, extra selling price can flow rapidly into profit once the factory is operating.

The weak years also show the reverse effect. In fiscal 2024, Sandisk recorded a $252 million charge for underused manufacturing capacity. It also recorded a $54 million pricing-related inventory write-down. An insurance recovery of $36 million increased that year’s reported result. These items show how excess supply can lower prices, reduce inventory values and leave factories underutilized.

Operating margin includes research, selling and administrative costs. Fiscal 2026 operating income reached $12.4 billion, equal to 61.2% of revenue, because sales grew faster than overhead. The comparison with fiscal 2025 is distorted by a $1.83 billion goodwill impairment. Without that charge, fiscal 2025 operating income would have been $453 million rather than a reported loss.

Net margin reached 56.5% in fiscal 2026. It included an $807 million unrealized gain on Sandisk’s Nanya investment. It also benefited from a $227 million tax benefit related to share awards. Those items affect reported income, but they do not show that Sandisk sold more NAND. The fiscal 2026 effective tax rate was 12.2%, a result of that year’s particular mix of income and tax items.

The adjusted table presents management’s alternative measures beside the tax rate; these are useful comparisons, but they depend on management’s chosen exclusions.

The company's own adjusted measures, and tax (US$ millions)

Effective tax rate — FY2026’s $227 million share-award tax benefit lowered the effective tax rate; without it, the rate was 13.9%, compared with 12.2% reported.

5.3 — Cash

Sandisk generated exceptional cash in fiscal 2026, but part of it arrived before the related revenue.

Cash from operations was $11.7 billion in fiscal 2026, while property and equipment spending was $177 million. That produced $11.5 billion of free cash flow. Deducting $232 million of stock-based pay gives a lower measure of cash economically available to owners, because stock awards transfer value even though they do not require a current-period cash payment.

Customer funding explains why operating cash needs careful reading. Under long-term agreements, customers can pay advances or security deposits so Sandisk will reserve capacity and supply them later. Sandisk records those payments as liabilities because it still owes performance, much as a builder holds a deposit before completing a house.

At fiscal year-end, contract and refund liabilities totalled $2.7 billion. Customer advances and deposits contributed $2.5 billion to fiscal 2026 operating cash flow. That cash is useful, but it includes funding for future deliveries and deposits that may eventually be returned.

Manufacturing spending is also broader than Sandisk’s own property-purchase line. In the October 2026 quarter, Sandisk reported $54 million of its own property purchases and $148 million of its share of Flash Ventures capital spending. Together, the company’s total cash capital-expenditure measure was $202 million.

The cash-flow table separates operating cash, working-capital effects, capital spending, stock pay and the company’s own free-cash-flow measure.

Follow the money: real free cash flow (US$ millions)

5.4 — Balance sheet and debt

Sandisk had strong liquidity and no reported borrowing at fiscal 2026 year-end, but its manufacturing obligations remain substantial.

Cash was $4.8 billion and reported debt was zero in fiscal 2026. The current ratio was 2.3 times, meaning current assets exceeded current liabilities. That is reassuring, but inventory is not cash: memory can lose value when prices fall.

On one calculation, inventory stood at roughly 153 days in fiscal 2026, up from about 143 days in fiscal 2025. Sandisk’s own measure gave 178 days and 135 days respectively, showing that the method matters. Higher inventory days may support customer commitments, but they also expose more capital to changing demand and prices.

Sandisk repaid its separation term loan on March 4, 2026 and recorded a $46 million debt-extinguishment charge. Its revolving facility was later refinanced and remains available. Fiscal 2026 interest expense was $73 million, easily covered by peak-cycle operating income, but that coverage would weaken quickly in a downturn.

Flash Ventures is the manufacturing partnership between Sandisk and Kioxia, which operates the Japanese wafer factories. Sandisk owns 49.9% of the venture entities, receives roughly half the output and pays half the fixed costs even when it orders less. It also funds its share of investment shortfalls and guarantees part of the venture’s leases.

Those guarantees totalled $923 million, including $378 million scheduled for fiscal 2027. Thus zero Sandisk debt does not mean zero manufacturing commitment. The balance sheet table shows the liquidity, liabilities, leases, equity and calculated ratios together.

Balance sheet (US$ millions)

5.5 — The latest quarter

The October 2026 quarter extended the recovery, with data-center sales providing most of the sequential increase.

Revenue reached $11.0 billion, compared with $9.0 billion in the preceding quarter. Gross margin rose to 85.0%. The margin reflects both strong pricing and the operating leverage of a factory system whose fixed costs are already in place.

Data-center revenue reached $4.8 billion, compared with $269 million in the year-earlier quarter. Edge revenue was $5.8 billion, while Consumer revenue fell to $397 million. Sandisk said it directed more supply toward data-center customers and that consumers were adjusting to higher prices.

Management attributed about 60% of the sequential revenue increase to volume and 40% to pricing. That is a healthier volume contribution than in the preceding fiscal quarter, but pricing remains central to the result.

Quarterly net income was $8.1 billion and diluted EPS was $52.60. Net income included $513 million of equity-security gains. Adjusted EPS was $49.51, reflecting the company’s specified adjustments. Sandisk also repurchased about 3.2 million shares, so per-share growth reflects both earnings and the changing share count.

The announced eight-for-one split means future EPS must be converted to the same basis before comparison. Management’s next-quarter revenue guidance of $12.2 billion to $12.6 billion supports the view that demand remained strong, but it is still a forecast.

The quarterly table makes the progression in revenue, margins, profit, EPS and free cash flow visible.

The last eight quarters (US$ millions)

Operating income — The March 2025 quarter included a $1.8 billion goodwill write-down; without it, the operating loss was $51 million.

5.6 — Manufacturing costs and technology transitions

Sandisk’s long-term test is whether new technology turns density gains into cheaper, saleable storage.

Cost per bit means the manufacturing cost of each unit of storage capacity. It falls when Sandisk produces more usable bits from the same manufacturing resources, but the complete calculation must include equipment, depreciation, venture costs and early-production inefficiencies.

A technology generation, or node, is a new design for building NAND cells. Sandisk said BiCS8 became the majority of its bit production in fiscal 2026. It also said a BiCS10 QLC design achieved 60% greater bit density than BiCS8. A separately announced BiCS10 TLC chip showed 59% greater density than the eighth-generation comparison.

Density is promising, but it is not the same as a cost reduction. A new process may initially produce too many defective chips. Yield—the share of chips that meet quality standards—must improve before theoretical capacity becomes saleable output. Customer qualification also takes time, especially for demanding data-center products.

That makes BiCS10 sampling encouraging but incomplete evidence. Sampling means customers are evaluating the product; it does not prove broad qualification, high yield or profitable volume production.

The specialist therefore watches for four things: saleable output, quantified cost per bit, customer qualification and evidence of a smooth manufacturing ramp. Sandisk has shown technological progress, but the accounts do not yet establish how much of the density gain will become durable economic advantage.

Sandisk is a flash-storage company whose shares now price in an extraordinary recovery. The useful question is not whether the stock looks cheap on recent earnings, but how much of those earnings, cash flow and factory economics the price assumes.

6.1 — What an investor pays today

Sandisk’s apparent cheapness rests on unusually strong earnings and a manufacturing system that is not fully visible in its own accounts.

Sandisk makes NAND, a type of semiconductor memory that stores data even when power is turned off. Market capitalization is the value of shareholders’ shares. Enterprise value adds debt and subtracts cash because a buyer would inherit the debt but receive the cash. Sandisk’s latest condensed balance sheet showed no borrowing balance and held $5.136 billion of cash on October 2, 2026.

The price-to-earnings ratio, or P/E, compares the share price with earnings per share. The forward P/E is much lower than the trailing P/E, which means analysts expect earnings to rise. That estimate can flatter a cyclical NAND producer: Sandisk’s non-GAAP gross margin, using management’s adjusted accounting, rose from 29.9% in fiscal Q1 2026 to 85.1% in fiscal Q1 2027.

EBITDA means earnings before interest, taxes, depreciation and amortization. It helps compare businesses, but depreciation represents the gradual use of expensive equipment. Removing it does not remove the need to replace and upgrade a factory.

That limitation is especially important because Sandisk owns 49.9% of Flash Ventures, its manufacturing partnership with Kioxia. Flash Ventures makes wafers, which Sandisk generally buys at cost plus a small markup. Sandisk pays its share of variable costs, shares fixed costs even when it takes less output and may have to fund roughly half of new investment. Because Sandisk uses equity accounting, it records its share of the ventures’ earnings rather than all their equipment, debt and depreciation.

Book value, cash-flow yield and the absence of a dividend offer additional lenses, but none settles the question. Sandisk also held $2.289 billion of marketable equity securities on October 2, 2026, including its Nanya stake, and recorded a $513 million gain on those securities in the latest quarter. That gain came from owning another company’s shares, not from selling NAND.

The table collects the current share price, market value, enterprise value, earnings multiples, book value, cash-flow yield and dividend yield.

What an investor pays today (US$ millions)

6.2 — Against its own past

Sandisk’s revaluation is real, but its short independent history cannot establish a normal valuation range.

Sandisk became an independent public company on February 21, 2025. Before then, Western Digital, its former parent, carved Sandisk’s accounts out of shared records. Those statements are useful, but allocated costs and assets are not the same as a long standalone trading history.

The comparison is also distorted by a $1.8 billion goodwill impairment in fiscal 2025. Goodwill is the accounting value assigned to past acquisitions; impairing it reduces reported assets and can make price-to-book comparisons look unusually severe. The move from a discount to book value to a large premium is therefore genuine, but it compares different accounting circumstances.

The missing evidence is a complete NAND downturn. Sandisk has not yet traded independently through a full weak pricing period, so today’s earnings may show a powerful cycle rather than a permanent earning level.

The table shows fiscal-year-end share price, P/E and price-to-book for fiscal 2025.

Fiscal 2025 valuation (share price in US$ per share; valuation multiples in times)

6.3 — Against rivals

Rival multiples provide context, but differences in products and cycles make them rough guides rather than precise answers.

Kioxia is the closest public comparison: it is Sandisk’s partner in Flash Ventures and also makes NAND. Its higher trailing P/E may reflect different customers, products, accounting and expectations rather than a simple judgment that Kioxia’s factories are worth more.

Micron and SK hynix make NAND but also have large DRAM businesses. DRAM is another kind of memory, used for fast temporary data storage, with its own pricing cycle. Samsung is broader still, so its valuation reflects many businesses beyond NAND.

Western Digital and Seagate focus mainly on hard disk drives, not flash memory. ADATA and Transcend sell memory products and modules but occupy different positions in the supply chain. Private companies such as Yangtze Memory and Kingston matter commercially, yet their lack of public share prices means they offer no observable stock-market P/E.

Sandisk buys wafers, adds controllers, firmware and other components, then sells storage solutions to data centers, device makers and consumers. NAND prices swing because producers must build costly capacity before they know demand. If several expand together, supply exceeds demand and prices fall; if demand outruns capacity, margins surge.

The table compares each rival’s listing, latest fiscal-year revenue, net margin, market value and trailing P/E.

The rivals

6.4 — How the price moved

Sandisk’s share-price path shows that improving business expectations and temporary market forces can operate at the same time.

The stock rose dramatically over the year but also suffered a sharp recent decline. Management’s long-term margin model helped support the rise, while a short seller argued that investors were valuing a cyclical commodity business like a protected technology franchise. The criticism is worth testing: Sandisk has design expertise, manufacturing technology and long-term customer agreements, but those advantages do not guarantee that competitors cannot pressure prices.

Index membership can add another source of demand. Sandisk entered the S&P 500 on November 28, 2025, the Nasdaq-100 on April 20, 2026 and the S&P 100 on September 21, 2026. Funds that track those indexes may need to buy the shares. Such buying can move the price without changing Sandisk’s earnings power.

The eight-for-one share split announced on October 29 changes the number of shares and the price of each share, but not the company’s value or an investor’s ownership. Historical prices and per-share earnings must be adjusted together when comparing periods.

The table shows the current share price against the 52-week high and low.

The share price today and over 52 weeks (US$ per share)

The table shows share-price change over one month, three months, six months and one year.

Share price performance

6.5 — What the price implies

Sandisk’s valuation works only if unusually strong cash generation survives beyond the current NAND upswing.

At today’s market value, an investor seeking a 10% annual return from a business producing the same cash indefinitely would need about $19.2 billion a year in cash available to shareholders. This simple illustration assumes no growth, no change in share count and no separate value for cash or the Nanya stake.

Management’s target is an adjusted free-cash-flow margin of approximately 50% for fiscal 2028–2030. At that margin, producing the required cash would take about $38.3 billion of annual revenue. Sandisk reported $11.031 billion of revenue in fiscal Q1 2027. Repeating that quarter four times would produce about $44.1 billion, but one strong quarter does not establish a durable annual result.

Customer agreements make future volume more predictable. Sandisk says agreements with eight customers cover about half of expected fiscal 2027 bits. They cover about two-thirds of fiscal 2028 bits. A bit is one unit of storage capacity, so these commitments describe volume rather than guaranteed profit.

The model also assumes approximately 80% non-GAAP gross margin and 75% non-GAAP operating margin for fiscal 2028–2030. It assumes revenue growth in the mid-to-high teens. The latest quarter’s adjusted gross margin exceeded the model, while the comparable quarter a year earlier was only 29.9%.

Contracts can protect against cancellations, but they cannot guarantee profit per unit. Costs may rise, products may become obsolete, customers may weaken and new NAND supply may push prices down. The price therefore assumes not merely that customers will buy, but that Sandisk can keep earning exceptional margins while serving them.

Sandisk Corporation makes NAND flash memory, the data-storage material used in devices and data centers. Its next chapter depends on whether exceptional prices and customer commitments can finance factories, taxes and obligations when conditions become less favorable.

7.1 — What management has guided

Management’s plan is a strong near-term quarter followed by an unusually ambitious profit and cash-generation model through fiscal 2030.

For the quarter ending around January 2027, Sandisk expects revenue to rise from the latest quarter and adjusted earnings to remain strong. The forecast uses the share count before the scheduled eight-for-one stock split. A split gives each owner seven additional shares for every old share, so the share price and earnings per share become roughly one-eighth while the business and each owner’s percentage remain unchanged. Thus, the pre-split per-share forecast would become roughly one-eighth as large after the split.

Sandisk’s longer-range model calls for mid-to-high-teen revenue growth, approximately 80% non-GAAP gross margin, approximately 75% non-GAAP operating margin and a 50% adjusted free-cash-flow margin. Gross margin is sales left after direct product costs; operating margin also subtracts operating expenses. These targets describe a highly favorable memory market, not a guaranteed outcome.

GAAP is the standard accounting result. Non-GAAP is management’s version after removing items it considers unusual or less useful for judging continuing operations. Stock-based compensation is a real employee cost even when Sandisk pays with shares rather than immediate cash. In the latest quarter, GAAP earnings also included a $513 million gain on equity securities, which Sandisk removed from its adjusted result. Because that gain exceeded the $91 million stock-compensation add-back, GAAP earnings per share of $52.60 exceeded non-GAAP earnings per share of $49.51.

Sandisk’s forecasting record is encouraging but brief. Across six comparable quarters, revenue and adjusted earnings per share exceeded guidance every time, while adjusted gross margin exceeded guidance in most cases. The strong memory market helped produce those results, so the record supports confidence without proving that the long-range targets are durable.

The guidance table sets out the forecast ranges, margins, expenses, tax assumptions, share count and adjustment items.

Guidance in force (US$ millions)

The promises table summarizes how often management’s principal forecasts came in above, within or below their ranges.

Promises kept: 4/5

The promise is ambitious; analysts’ estimates show how much uncertainty remains around it.

7.2 — What analysts expect

Analysts expect rapid growth, but their average becomes much less reliable as the forecast horizon lengthens.

A consensus is the average of estimates from analysts who publish forecasts. An October 29 snapshot puts fiscal 2027 revenue at about $49.6 billion and adjusted earnings at $216.01 per share. For fiscal 2028, the averages rise to about $59.2 billion and $267.11 per share. These earnings estimates use analysts’ adjusted basis, which need not match Sandisk’s GAAP result.

Near-term expectations have also left room for positive surprises. In the latest quarter, adjusted earnings per share of $49.51 exceeded consensus of $47.15. Revenue also exceeded its consensus estimate, and one earnings-history provider records four beats in the preceding four reported quarters.

The ranges reveal more than the averages. For fiscal 2028, revenue estimates run from $43.0 billion to $69.5 billion. By fiscal 2029, they run from $35.5 billion to $85.9 billion, while only four earnings estimates remain. Fiscal 2030 has just five revenue estimates and one earnings estimate.

That thinning matters because a few analysts can move the average substantially. The wide spread reflects uncertainty about demand, factory capacity and memory prices. The outer years therefore describe possible paths rather than a settled destination. All published per-share estimates must also be converted to the post-split basis before comparison.

Those possible paths require substantial spending.

7.3 — The spending ahead

Sandisk’s investment plan combines advanced technology spending with a long commitment to its Japanese manufacturing partnership.

Sandisk receives flash wafers from manufacturing ventures with Kioxia, its Japanese partner. The companies extended that relationship on January 29, 2026, through December 31, 2034. Under the agreement, Sandisk committed $1.2 billion to Kioxia over 2026–2029 for manufacturing services and continued supply.

Flash Ventures is the joint manufacturing arrangement behind the partnership. Sandisk owns 49.9% of the ventures and generally receives about half their output, while paying its share of variable production costs and fixed costs. Factory capacity matters because semiconductor plants take years and billions of dollars to build. Newer BiCS generations stack memory cells more densely, but the transition requires new equipment, engineering and successful production yields.

The immediate technology priority is BiCS8, with BiCS10 also being ramped. Management expects fiscal 2027 capital spending to rise in dollars while falling to approximately 6% of revenue, with most of that spending supporting BiCS8. In the quarter ended October 2, 2026, Sandisk’s share of joint-venture gross capital expenditure was $324 million, while its share of cash capital expenditure was $148 million. Direct equipment and office purchases added $54 million, bringing total cash capital expenditure to $202 million.

Customer contracts make that expansion easier to plan. Sandisk has signed New Business Model agreements with eight customers that specify volumes, minimum financial guarantees and structured pricing. They cover about half of fiscal 2027 bits and about two-thirds of fiscal 2028 bits, improving visibility into demand.

But contracted demand is not the same as earned revenue. The agreements can include advances, deposits and refundable amounts. Sandisk must still manufacture and deliver the storage, and customers must perform. The contracts reduce demand uncertainty while creating delivery and repayment obligations.

The next question is whether Sandisk’s cash can carry both the obligations and the expansion.

7.4 — Can they pay for it

Sandisk appears able to fund its disclosed near-term commitments, but that conclusion depends on continued operating strength.

The company ended the latest quarter with substantial cash and no total debt. Its debt schedule shows no borrowings coming due in the five years after the latest fiscal year-end. That creates breathing room, but commitments for factories, equipment, leases and supplies still require future cash.

In the latest quarter, operations generated $6.363 billion. Sandisk also spent $5 billion repurchasing shares and $755 million on taxes associated with vested employee stock awards, yet cash still increased. The company’s adjusted free cash flow was $5.696 billion after accounting for $465 million of New Business Model prepayments and deposits.

That adjusted figure is lower than operating cash flow because free cash flow subtracts capital spending, while Sandisk’s adjusted measure also removes unusual customer-payment and joint-venture effects. Early customer cash helps liquidity, but it relates to products Sandisk still must deliver and deposits that may be refundable.

From the July 2026 year-end, Sandisk listed known material cash requirements of $3.391 billion for fiscal 2027 and $5.625 billion across fiscal 2028–2029. These are snapshots of known obligations, not one immediate bill. The company also had a $1.5 billion revolving credit facility, which is borrowing capacity rather than cash already owned.

The remaining $10.5 billion buyback authorization is discretionary and can be suspended. Management expects operating cash flow to fund repurchases after business investment. The real test is whether cash generation remains strong if prices fall, deposits return or customers pay more slowly.

The cash-and-debt table shows the company’s recent liquidity and debt position.

Cash and debt, last eight quarters (US$ millions)

The debt-maturity table shows when scheduled borrowings come due.

Debt coming due after the latest fiscal year end (US$ millions)

The next results will show whether current strength is translating into durable performance.

7.5 — What to watch

The decisive test is whether Sandisk converts customer commitments and high prices into lasting deliveries and cash.

The first checkpoint is the January 2027 fiscal-quarter result. Compare sales, margins and adjusted earnings with management’s outlook, but separate volume from price. In the October quarter, approximately 60% of sequential revenue growth came from volume and 40% from higher pricing. Volume-led growth would support the investment case; price-led growth would leave Sandisk more exposed to a memory-price reversal. One industry forecast expects NAND prices to rise another 30% in 2027, but that forecast cannot establish what Sandisk will actually earn.

During fiscal 2027, watch BiCS8 spending, customer deliveries, advances and refundable deposits. A large contract improves visibility only if Sandisk can produce the promised storage and collect the related cash.

Taxes will also claim more attention. Sandisk expects the U.S. corporate alternative minimum tax, or CAMT, to apply in fiscal 2027. CAMT is a 15% minimum tax calculated on an adjusted financial-statement income base. At July 3, 2026, Sandisk recorded $323 million of unrecognized tax benefits, meaning tax positions authorities might later reject. About $259 million could eventually require cash payment, and a separate $128 million tax-indemnity liability connects to former parent Western Digital.

Finally, separate unusual items from the continuing business. Sandisk recognized $93 million related to legal settlements in fiscal 2026. Management also plans samples of its first high-bandwidth flash products in calendar 2027, but samples show customer evaluation rather than meaningful sales.

The central question is therefore not whether Sandisk can fund the next quarter. It is whether manufacturing execution, customer performance, taxes and memory prices allow today’s exceptional cash generation to finance the much larger plan ahead.

The forecast points to substantial probability-weighted upside on a consistent post-split basis, and SanDisk remains profitable in the base case. The weighted price is $369.91 in October 2027, a 122% gain including dividends on a consistent post-split basis. It falls to $277.22 in October 2028, a 67% gain including dividends on a consistent post-split basis. By October 2029 it is $270.32, a 63% gain including dividends on a consistent post-split basis.

The last-date range is $106.06 to $814.92, against today’s split-adjusted $166.26. The bear case has the greatest downside because NAND prices can reverse quickly; the bull case needs unusually durable margins.

The weighted row is the key comparison: it combines the three paths rather than presenting the base case as a certainty.

What the shares could be worth (USD per post-split share at each date; in brackets the return since today's split-adjusted $166.25625, dividends included)

8.2 — The model

The forecast prices SanDisk on normal operating profit because its NAND earnings are too cyclical for the latest quarter to stand as a permanent run rate. The calculation applies an enterprise-value-to-normal-operating-profit multiple, subtracts net debt, and divides by projected shares.

This fits the company because operating income moved from a $2.0 billion loss to a $12.4 billion profit across the recent history, while fixed venture costs remain when NAND prices weaken. Net cash raises equity value, whereas buybacks reduce the denominator. The model therefore links valuation, cash generation and capital returns in one framework.

8.3 — The drivers

Normal operating profit is the main driver. The base case starts at $34 billion and eases to $26 billion as supply responds. It reaches $22 billion by 2029, reflecting a more ordinary NAND cycle. The latest quarter produced $8.8 billion of operating income on an 79.6% margin, while the next-quarter revenue guide is $12.2–$12.6 billion. But annual operating income was negative in three of the previous five years, so the latest margin is not treated as normal.

The bear case uses $30 billion in 2027 and then only $8 billion as prices reset; it recovers to $12 billion. The bull case rises from $37 billion to $43 billion and then $45 billion. Data-center revenue grew from $960 million to $5.2 billion, and customer agreements cover about half of fiscal 2027 bits and two-thirds of fiscal 2028 bits. The bull case nevertheless requires exceptional margins through 2029.

Shares begin at a post-split 1,152 million, based on the latest basic count and the announced split. The base case falls to 1,110 million after $10.5 billion of buybacks. It reaches 1,068 million after further repurchases and assumed stock issuance. The bear case reaches 1,013 million in 2028 but then rises to 1,021 million as buybacks stop. The bull case ends at 1,079 million because stronger cash generation supports larger repurchases. These are assumptions: the authorization is discretionary, although $5.0 billion was spent in the latest quarter.

Dividends are zero in every case and at every date. SanDisk paid no dividend in any of the five reported fiscal years, and its cash-return programme does not promise dividends.

The base multiple declines from 12x to 10x and then 9x as peak margins normalize. The bear case uses 8x, 5x and 7x because fixed costs magnify a downturn. The bull case uses 18x, 20x and 18x because durable enterprise demand could make earnings more credible. Peer P/E figures provide context but are not directly comparable to this measure.

Net debt starts at negative $5.1 billion, meaning net cash, after the latest quarter. The base case reaches negative $20.7 billion and then negative $39.0 billion as operating cash exceeds capex and buybacks. The bear case reaches only negative $15.7 billion in 2028 because cash generation weakens. The bull case reaches negative $69.3 billion by 2029, requiring high profits and customer advances.

The forecast’s table shows every selected input and its scenario range.

The forecast's table (base case, then bear / bull in brackets; USD millions, per share in USD, unless shown)

8.4 — Bear and bull

The bear case requires data-center construction to slow, NAND supply to catch demand and customer purchases to weaken. SanDisk would still carry Flash Ventures’ fixed costs, while enterprise qualifications failed to become sufficient volume. Its price is $241.98 in 2027 and $106.06 in 2029. The probability is 35% because operating income was negative in three recent fiscal years before the latest recovery.

The base case assumes strong near-term deliveries, followed by lower prices and margins. Its price is $386.20 in 2027 and $221.92 in 2029. The probability is 50% because guidance and contracted bits support growth, but the six-quarter record of beating guidance partly reflects a strong memory market.

The bull case needs durable AI-storage demand, recurring enterprise volume and successful technology ramps that lower cost per bit. Its price is $614.09 in 2027 and $814.92 in 2029. The probability is 15% because data-center growth is real, but sustained margins above 65% have not been proven through a full standalone cycle.

8.5 — What would change the view

January 2027 results are the first checkpoint. Revenue near the guided $12.2–$12.6 billion range would support the base and bull paths; a miss would move the forecast toward the bear case.

Through 2027 and 2028, renewed NAND oversupply, slower data-center construction or lower customer purchases would reduce margins and the valuation multiple. Successful enterprise qualifications becoming recurring production would support the bull case, while delayed deployment would weaken it.

By 2028 and 2029, evidence of lower cost per bit and smooth manufacturing ramps would justify stronger profits. Cash generation that funds buybacks would also improve per-share value, but management’s excess-cash commitment does not guarantee a schedule.