Sandisk Forecasts More Than $10 Billion in Quarterly Revenue as AI Storage Demand Surges

Sandisk raised its quarterly revenue outlook above $10 billion, despite lingering doubts that an AI-fueled NAND boom can escape the memory industry's old cycles.

The forecast pushed Sandisk across Google News after a fiscal quarter that transformed its financial profile. Revenue reached $8.97 billion, up 51 percent from the previous quarter and 372 percent from one year earlier. The company expects fiscal first-quarter 2027 revenue between $10.30 billion and $10.80 billion.

That guidance is not simply another strong semiconductor forecast. It tests whether AI inference has created lasting demand for flash storage, or only an unusually profitable phase in a familiar commodity cycle.

Sandisk argues that data center growth, higher-value customers, and longer agreements have made its earnings more durable. Competitors such as Micron, Samsung, SK hynix, Kioxia, and YMTC still give buyers alternatives. New capacity and weaker AI spending could eventually restore pricing pressure.

The central contest is therefore Sandisk's structural-growth claim against NAND's cyclical history. Its latest quarter provides strong evidence for the company, but pricing contributed more growth than shipment volume.

What Changed in the Sandisk Google News Forecast

Sandisk's outlook shows that AI infrastructure spending has moved beyond processors and into the storage systems feeding those processors with data.

Sandisk reported its fiscal fourth-quarter results on August 5, covering the period that ended July 3. The company's quarterly results put revenue at $8.97 billion.

That total exceeded Sandisk's earlier guidance of $7.75 billion to $8.25 billion. It also represented a 51 percent sequential increase from the fiscal third quarter.

The composition of that growth matters. Sandisk said approximately one-third came from higher volume, while two-thirds came from higher pricing. Demand strengthened, but a tight supply environment amplified its financial effect.

GAAP gross margin reached 84.6 percent, compared with 78.4 percent in the preceding quarter. GAAP operating income rose 71 percent sequentially to $7.04 billion.

Net income reached $6.90 billion, up from $3.62 billion during the previous quarter. Fiscal-year revenue totaled $20.25 billion, a 175 percent annual increase.

The next-quarter forecast extends that momentum. Sandisk expects revenue between $10.30 billion and $10.80 billion, alongside a non-GAAP gross margin between 83 and 85 percent.

At the midpoint, the forecast implies another sequential revenue increase of about 18 percent. That is slower than the latest quarter's 51 percent jump, yet still remarkable for a large memory supplier.

The result also marked a sharp reversal from fiscal 2025. Sandisk generated $1.90 billion in the comparable quarter and recorded a GAAP net loss of $23 million.

Its current performance is not spread evenly across every market. Data center revenue reached $2.98 billion, more than double the prior quarter's $1.47 billion.

Edge revenue, which includes storage used in connected devices and distributed computing systems, rose 48 percent sequentially to $5.43 billion. Consumer revenue fell 32 percent to $556 million.

That split makes the AI infrastructure connection more credible. The fastest-growing operation was data center storage, while the consumer business contracted.

For the full fiscal year, data center revenue climbed 437 percent to $5.15 billion. Edge revenue increased 195 percent to $12.16 billion, while consumer revenue rose 29 percent.

The Google News headline captured the optimistic forecast, but the underlying event is broader. Sandisk has become less dependent on consumer storage while directing more supply toward enterprise and cloud customers.

That change creates the article's tension. Enterprise buyers currently accept higher NAND costs because inadequate storage can delay expensive AI deployments. Their urgency gives Sandisk unusual pricing power.

However, pricing power based on scarcity differs from growth driven entirely by rising shipment volume. The distinction will determine whether current margins remain durable.

AI Inference Is Pulling NAND Into the Data Center Race

AI inference creates a storage problem because models must repeatedly retrieve large datasets, embeddings, checkpoints, and cached results without keeping everything in costly working memory.

Training builds a model, while inference uses that trained model to answer requests. At scale, inference systems must move data through processors, memory, networking equipment, and persistent storage.

NAND flash provides persistent storage without requiring power to preserve data. Enterprise solid-state drives use NAND to hold large datasets while providing faster access than conventional hard drives.

High-bandwidth memory, or HBM, performs a different role. It sits much closer to accelerators and offers far greater bandwidth, but it remains costlier and more capacity-constrained than NAND storage.

An AI server therefore cannot solve every data problem by adding HBM. It needs a storage hierarchy that places the hottest information near processors and colder information on larger, cheaper devices.

Sandisk benefits when AI operators expand that hierarchy. More inference traffic creates more model copies, vector databases, retrieval indexes, logs, and generated content.

This demand is arriving while NAND suppliers remain cautious about capacity. Memory manufacturers have repeatedly experienced periods when rapid expansion produced oversupply, collapsing prices and profits.

The present shortage reflects both demand and that restraint. A NAND market analysis from TrendForce said the five leading suppliers increased combined revenue 83.7 percent sequentially during the first quarter of 2026.

TrendForce also expected suppliers to add virtually no new production capacity during 2026. It linked the shortage to strong AI-related demand and expected tight conditions throughout the year.

Sandisk has another constraint because it does not independently manufacture every NAND wafer. It obtains supply through Flash Ventures, its long-running manufacturing partnership with Kioxia in Japan.

That partnership offers scale without requiring Sandisk to own the entire fabrication network. It also creates reliance on joint investment decisions, manufacturing execution, and a strategic partner.

Sandisk and Kioxia extended key venture agreements through 2034 earlier this year. The extension gives Sandisk longer visibility into production at Japanese facilities.

Yet manufacturing agreements cannot add mature capacity instantly. New tools, buildings, process transitions, and yield improvements require time before producing commercially useful output.

Sandisk is responding by changing its customer relationships. It calls the approach its New Business Model, or NBM, built around multi-year commitments and firmer financial terms.

The company announced five NBM agreements during its April earnings update. By August, it had signed five more, including three agreements with new customers.

Two other deals expanded previously signed arrangements. Sandisk has not publicly disclosed every customer's identity or the complete economics behind those contracts.

Longer commitments can reduce the uncertainty surrounding future demand. They can also prevent buyers from abandoning orders when market prices reverse.

For buyers, securing supply protects scheduled data center deployments. A delayed storage shipment can strand accelerators, networking equipment, power capacity, and other costly infrastructure.

That imbalance explains a remark Sandisk CEO David Goeckeler made earlier in 2026. He told Reuters that customers preferred securing supply over receiving a lower price.

The comment described a market where availability became the central purchasing concern. Sandisk's latest results show that this behavior continued through the end of its fiscal year.

Still, contracts do not eliminate the cycle. They redistribute some risk between supplier and buyer while potentially reducing the speed of future price declines.

If demand stays strong, Sandisk gains visibility and attractive margins. If demand weakens, contract details will determine how much committed revenue remains protected.

Sandisk Is Betting Against NAND's Cyclical History

The bullish case says AI has created sustained enterprise demand, while the skeptical case says scarcity and pricing still explain most of Sandisk's profit expansion.

Sandisk's fourth-quarter growth offers evidence for both readings. Data center revenue doubled sequentially, supporting the claim that AI-related storage demand is real.

However, the company attributed about two-thirds of sequential revenue growth to higher pricing. That means scarce supply, not only expanding consumption, powered the quarter.

NAND has historically behaved like a commodity market. Suppliers invest during profitable periods, new output arrives, inventories increase, and prices eventually fall.

That pattern can reverse earnings quickly because semiconductor factories carry high fixed costs. A relatively small supply imbalance can produce a much larger change in margins.

Sandisk's new agreements aim to weaken that pattern. Multi-year customer commitments should make purchasing behavior more predictable than short-term orders from consumer electronics companies.

Its end-market shift also matters. Consumer revenue represented only about 6 percent of fourth-quarter sales, down from more than 30 percent one year earlier.

Data center and edge customers now dominate the business. Their storage requirements often connect to longer infrastructure plans rather than discretionary device upgrades.

Even so, Sandisk's latest gross margin deserves careful interpretation. An 84.6 percent margin is exceptionally high for a NAND supplier and reflects unusually favorable market conditions.

The company guided for a slightly lower midpoint next quarter. That forecast still suggests strong pricing, but it does not establish that such profitability has become permanent.

Sandisk also expanded its share repurchase authorization by $14 billion. The remaining authorization reached $15.5 billion after the board's decision.

Buybacks can increase earnings per share by reducing the share count. They do not create additional customer demand or protect NAND prices from future oversupply.

The scale of the authorization signals management's confidence in cash generation. It also means investors should separate operational improvement from per-share gains caused by repurchases.

Sandisk ended the fiscal year with $4.76 billion in cash and no long-term debt. That balance sheet gives it more flexibility during future market downturns.

Its independence is relatively new. Sandisk separated from Western Digital in February 2025 and returned to public trading as a standalone flash storage company.

The separation created a focused NAND investment. It also removed the diversification that Western Digital's hard-drive business once provided.

A focused structure magnifies upside during a flash shortage. The same structure exposes shareholders more directly when NAND prices, enterprise orders, or manufacturing yields weaken.

This makes Micron an important comparison. Micron sells NAND, but it also participates in DRAM and HBM markets that serve different parts of the AI computing stack.

Micron's broader portfolio can spread risk across memory categories. Sandisk offers more concentrated exposure to NAND storage and its current pricing environment.

Samsung and SK hynix also operate across multiple memory categories. Kioxia resembles Sandisk more closely, although the two companies share manufacturing investments and remain commercial competitors.

YMTC adds another source of competitive pressure, particularly in China. Its expanding technical capabilities could influence global supply even when trade restrictions limit some customer relationships.

Sandisk's argument does not require NAND to stop being cyclical. It requires future downturns to become shallower because AI systems consume more persistent storage.

That remains plausible but unproven. One exceptional year cannot establish how customers will behave after supply catches demand.

The company's fiscal 2027 performance will provide a stronger test. Stable volumes and contract-backed revenue during softer pricing would support the structural-growth case.

A sharp revenue decline following weaker pricing would support the traditional cycle explanation. Investors should watch both units and prices instead of treating headline revenue as a single signal.

What Sandisk's Numbers Do Not Yet Prove

The results confirm extraordinary current demand, but they do not prove that customers will accept rising prices after supply conditions normalize.

Sandisk's official outlook contains several uncertainties. It cites demand volatility, average selling prices, customer deployment timing, competition, manufacturing transitions, and reliance on Kioxia.

Those are standard financial disclosures, but they map directly onto the current debate. Each could interrupt the path implied by the upbeat Google News coverage.

The first uncertainty concerns pricing elasticity, meaning how much demand changes when prices rise. AI operators accepted higher costs while storage remained scarce, but that tolerance has limits.

Customers can delay deployments, redesign storage tiers, improve data management, or negotiate with alternative suppliers. They can also use hard drives for workloads that prioritize capacity over access speed.

The second uncertainty concerns the timing of AI capital spending. Hyperscalers have announced large infrastructure programs, yet quarterly purchasing rarely follows a smooth line.

A data center delay can move storage orders between quarters. Power availability, accelerator shipments, construction schedules, and networking equipment can all change deployment timing.

The third uncertainty involves capacity. TrendForce expected limited additions during 2026, but strong profits create an incentive for suppliers to invest.

New capacity does not affect the market immediately. Once enough output arrives, however, a shortage can turn into balance or oversupply faster than long-term forecasts imply.

The fourth uncertainty sits inside Sandisk's customer agreements. The company says NBMs include firm financial commitments, but public disclosures provide limited detail about volume floors and adjustment mechanisms.

Investors therefore cannot fully measure the protection those agreements provide. Contract duration alone does not reveal pricing formulas, cancellation terms, or customer concentration.

The fifth uncertainty concerns product execution. Sandisk must keep advancing NAND density and enterprise SSD performance while improving manufacturing yields.

Its BiCS10 sampling began with a one-terabit, triple-level-cell device. Triple-level-cell NAND stores three bits in each memory cell.

Sandisk says BiCS10 provides a 59 percent bit-density improvement over BiCS8 and an interface speed reaching 4.8 gigabits per second. Sampling is not the same as high-volume production.

Customers must qualify the technology inside complete storage systems. Sandisk and Kioxia must then reach yields that support acceptable cost, reliability, and delivery volumes.

The sixth uncertainty is technological substitution. NAND is well suited to large persistent datasets, but AI storage architectures remain in flux.

Software can reduce storage requirements through compression, caching, data deduplication, and smaller models. New memory technologies can also change how systems divide work across storage tiers.

None of these risks negates Sandisk's results. They explain why one should not translate a strong quarterly forecast directly into a permanent earnings assumption.

The consumer decline offers an immediate warning. Consumer revenue fell both sequentially and year over year, even while total revenue reached a record level.

Higher input costs can weaken demand for laptops, phones, memory cards, and retail SSDs. Suppliers may prioritize data center buyers, leaving consumers with fewer affordable options.

That tradeoff can continue while enterprise demand remains urgent. It becomes harder to sustain if consumer weakness combines with delayed cloud orders.

The next several quarters should reveal whether data center growth can offset those pressures. Sandisk needs volume expansion to take a larger role from pricing.

A healthier long-term pattern would pair moderate price growth with rising shipments. Continued dependence on steep price increases would keep the cyclical concern alive.

Competitors Are Racing to Redesign the AI Storage Tier

Sandisk's advantage depends on converting current supply tightness into products and standards that remain valuable after competitors add capacity.

The company is developing two notable technologies for that task. BiCS10 increases NAND density, while High Bandwidth Flash targets a new position closer to AI processors.

High Bandwidth Flash, or HBF, combines stacked NAND with a high-speed interface. Its designers want to offer much greater capacity than HBM at lower cost per stored bit.

Sandisk and SK hynix released the first HBF specification through the Open Compute Project in August. The specification provides a common technical foundation for potential adoption.

HBF does not replace HBM in every task. NAND remains slower and has different endurance characteristics, making the technologies suitable for different positions in a system.

The intended use involves AI inference workloads with large models or datasets. HBF could keep more information physically close to processors without relying entirely on expensive HBM capacity.

That approach addresses a real constraint. Model sizes and working datasets can grow faster than the memory capacity attached to accelerators.

However, a published specification does not guarantee commercial adoption. Chip designers, system vendors, cloud operators, and software developers must agree that HBF solves a worthwhile problem.

Packaging complexity also matters. Stacked components require reliable thermal management, high manufacturing yields, and interfaces that work across multiple vendors.

Sandisk and SK hynix provide important storage expertise, but broader support remains necessary. Accelerator companies can choose different memory architectures or proprietary approaches.

BiCS10 follows a more conventional path. Greater density lets Sandisk store more information on a given wafer area, potentially reducing production costs per bit.

The technology uses 332 active memory layers. Layer count alone does not determine efficiency, but denser designs can improve economics when manufacturing yields reach target levels.

Kioxia and Sandisk share development and production responsibilities across their Japanese partnership. Their cooperation gives both access to advanced NAND technology while preserving separate commercial businesses.

Samsung remains the largest NAND supplier and has extensive data center relationships. Micron combines enterprise SSDs with DRAM and HBM, while SK hynix owns Solidigm's enterprise storage portfolio.

Kioxia competes directly for cloud and enterprise SSD demand. YMTC's growing scale adds further pressure to a market that has rarely maintained scarcity indefinitely.

Hard-drive suppliers also remain relevant. Seagate and Western Digital serve bulk data storage where capacity and cost matter more than the lowest access latency.

AI data centers can use both flash and hard drives. Flash handles latency-sensitive datasets, while disks can retain archives, checkpoints, and less frequently accessed information.

That makes the competitive landscape more complex than Sandisk versus one rival. The main contest remains Sandisk's durability claim against the behavior of the wider storage market.

Product advances can strengthen that claim if customers pay for differentiated performance. Commodity expansion can weaken it if buyers view suppliers as interchangeable.

Sandisk's current margins suggest buyers have limited negotiating leverage. The duration of that imbalance depends on qualification cycles, capital spending, and available alternatives.

The company now has cash, customer commitments, and a growing data center operation. It must use that position before the supply response catches up.

Three Signals Will Decide Whether the Forecast Holds

The next test is not another optimistic headline, but whether Sandisk can preserve demand, margins, and customer commitments as the NAND market adjusts.

The first signal is Sandisk's fiscal first-quarter revenue mix. The company expects total revenue between $10.30 billion and $10.80 billion.

Readers should focus on data center revenue and the balance between volume and pricing. Another quarter driven mostly by higher prices would extend the boom without resolving the durability question.

A larger contribution from shipment volume would strengthen Sandisk's case. It would show that customers are deploying more storage rather than mainly paying more for constrained supply.

The second signal is the detail Sandisk provides about its New Business Model agreements. Additional customers, disclosed commitments, or better contract visibility would make future revenue easier to assess.

Investors should also watch contract liabilities, which represent payments or obligations tied to products not yet recognized as revenue. These reached $1.24 billion at fiscal year-end.

That total included $849 million in current contract liabilities and $393 million in non-current liabilities. Growth would suggest that longer-term customer arrangements are becoming financially meaningful.

The opposite pattern would weaken the thesis. Limited additions or vague commitments would leave more revenue exposed to spot-market conditions and customer deployment changes.

The third signal is the industry's supply response. Production plans from Samsung, Kioxia, Micron, SK hynix, and YMTC will shape pricing beyond 2026.

Capacity announcements alone deserve caution because usable output arrives after equipment installation and yield improvements. Actual bit shipment growth provides a more reliable measure.

If suppliers remain disciplined while AI storage consumption expands, Sandisk's margins can stay elevated. If output accelerates faster than demand, NAND's traditional cycle will reassert itself.

BiCS10 qualification and HBF adoption belong inside this third signal. Successful commercial ramps could let Sandisk compete on density and architecture rather than scarcity alone.

Failures or delays would leave pricing as the dominant profit driver. That would make the company more vulnerable when industry supply improves.

The latest forecast deserves attention because the underlying figures are unusually strong. Sandisk converted a $1.90 billion quarterly business into an $8.97 billion operation within one year.

Yet the same figures demand skepticism. Two-thirds of sequential growth came from pricing, while the consumer operation contracted and enterprise supply remained tight.

The most useful reading of the story sits between celebration and dismissal. AI inference has clearly expanded the market for flash storage, and Sandisk has captured that demand.

What remains uncertain is whether customer commitments and product differentiation have permanently changed the cycle. Google News readers should track shipment volume, contract liabilities, and industry output.

Those three indicators will reveal more than the next headline. Watch whether volume replaces pricing as the primary growth engine, then compare Sandisk's progress with competing supply additions.

If that shift occurs, Sandisk's structural-growth argument gains credibility. If pricing falls before volume takes over, the company's extraordinary year will look more like a cycle peak.