Sandisk Fiscal 2026 Earnings Broke Records, but Pricing Drove the Surprise

Sandisk closed fiscal 2026 with record quarterly revenue, despite relying more on higher NAND pricing than rising shipment volumes. The company reported the results on August 5, following a quarter that ended July 3. That distinction defines the entire Sandisk fiscal 2026 earnings story.

Quarterly revenue reached $8.97 billion, rising 51% from the previous quarter and 372% from a year earlier. However, management attributed about two-thirds of the sequential increase to pricing. Higher volumes supplied only the remaining third.

The call was therefore more than a celebration of AI storage demand. Sandisk argued that multiyear customer agreements can turn a famously cyclical market into a more predictable business. Competitors including Micron, Samsung, and Kioxia are chasing the same demand while carefully controlling supply.

The numbers support Sandisk's near-term confidence. Whether they establish durable earnings power is a harder question. Investors must separate three forces: AI infrastructure growth, unusually favorable NAND prices, and contracts designed to soften the next downturn.

Sandisk Fiscal 2026 Earnings Set New Records

Sandisk delivered an exceptional quarter, but the composition of its growth matters as much as the headline numbers.

Revenue reached $8.965 billion in the fourth fiscal quarter. Non-GAAP gross margin climbed to 84.6%, up 6.2 percentage points sequentially. Non-GAAP diluted net income per share reached $39.25.

The company also generated $7.126 billion in operating cash flow. Adjusted free cash flow was $5.035 billion after excluding customer prepayments and deposits connected with new agreements.

These figures exceeded the high end of Sandisk's previous guidance. The company called the quarter a proof point for its strategy in its earnings presentation.

Sandisk's three end markets moved in different directions.

Datacenter revenue reached $2.977 billion, rising 103% from the previous quarter and 1,298% year over year. Its share of Sandisk's shipped bits increased from 12% a year earlier to 38%.

Edge revenue, which covers areas such as PCs, smartphones, tablets, vehicles, and robotics, reached $5.432 billion. That represented 48% sequential growth and 392% annual growth.

Consumer revenue fell 32% sequentially and 5% annually to $556 million. The decline exposes how differently the current memory cycle affects AI infrastructure and traditional retail products.

Across fiscal 2026, Datacenter produced $5.153 billion in revenue, up 437% from the prior year. Edge generated $12.160 billion, up 195%, while Consumer reached $2.935 billion, up 29%.

The mix shift is important. Sandisk no longer depends primarily on memory cards, portable storage, and other familiar consumer products. Enterprise SSDs now give the company direct exposure to hyperscale computing and AI infrastructure.

Yet pricing remains the largest immediate driver. Management said two-thirds of sequential quarterly revenue growth came from higher prices. Only one-third came from volume.

That split does not invalidate the results. It changes how investors should interpret them.

Volume growth indicates that customers need more physical storage. Pricing growth indicates that available supply is scarce relative to demand. Both help revenue, but pricing can reverse faster when the balance changes.

Sandisk also spent $4.5 billion repurchasing shares during the quarter. Its total authorization since becoming independent reached $20 billion, leaving $15.5 billion available.

The scale of those repurchases shows management's confidence in future cash generation. It also creates a capital-allocation question. Every dollar returned to shareholders is unavailable for capacity, technology transitions, or protection against a future downturn.

Fiscal first-quarter 2027 guidance extended the strong trajectory. Sandisk projected revenue between $10.3 billion and $10.8 billion. It also forecast a non-GAAP gross margin between 83% and 85%.

That outlook suggests the favorable environment continued beyond the July quarter. It does not establish how long current pricing can last.

The result created the article's central tension. Sandisk has entered an extraordinary profit period, while the underlying memory market remains defined by fluctuating supply and prices.

AI Inference Is Pulling NAND Into the Data Center

The strategic change is not simply higher NAND demand. It is the movement of NAND toward the most valuable part of AI infrastructure.

AI training receives much of the industry's attention because it requires expensive processors and high-bandwidth memory. Inference, which runs a trained model to answer requests, creates a different storage problem.

Each inference system must retrieve model assets, databases, documents, media, logs, and cached results. Those workloads require large storage pools with predictable latency and energy use.

Agentic applications add another layer. They repeatedly gather information, execute tasks, and preserve state across sessions. That activity creates data that must be written, retained, and retrieved.

Sandisk says these patterns are making flash storage a more important component of AI architecture. Its enterprise SSDs target both compute-intensive workloads and high-capacity data lakes.

The company began generating revenue from its QLC Stargate platform during the quarter. QLC stores four bits in each memory cell, allowing greater capacity at a lower cost per stored bit.

QLC historically involved performance and endurance tradeoffs. Controllers, firmware, and improved NAND designs can reduce those limitations for data-heavy workloads that prioritize capacity.

Sandisk is also shipping enterprise products built around TLC, which stores three bits per cell. TLC generally serves workloads that require a stronger balance of performance, endurance, and density.

This broader portfolio helps explain the rapid Datacenter expansion. Sandisk can address high-performance compute storage and capacity-focused AI repositories instead of betting on one application.

Its latest technology transition strengthens that position. The company started sampling BiCS10, a 332-layer, one-terabit TLC NAND design developed with Kioxia.

According to the BiCS10 specifications, the chip provides up to 4.8 gigabits per second of interface speed. Sandisk also claims a 59% bit-density improvement over BiCS8.

Those figures remain company claims until production products establish performance under real workloads. Sampling marks the beginning of customer evaluation, not full commercial adoption.

Sandisk and Kioxia share a manufacturing relationship through Flash Ventures. Sandisk owns 49.9% of that venture, while Kioxia owns 50.1%.

The partnership allows both companies to co-develop process technology and memory designs. It also means Sandisk depends on a strategic partner that competes in the same NAND market.

This arrangement differs from a conventional supplier contract. Both companies fund and use the venture while selling competing products to customers.

Sandisk also worked with SK hynix on High Bandwidth Flash, or HBF. HBF is a proposed NAND architecture intended to place much larger storage capacity near AI processors.

The companies released the first Open Compute Project technical specification for HBF on August 3. Their HBF specification targets AI inference systems that need more capacity than high-bandwidth memory can economically provide.

HBF remains an emerging route, not a material contributor to the reported quarter. Its relevance lies in where Sandisk expects future storage bottlenecks to appear.

The company estimates Datacenter will expand from roughly 30% of the NAND market in calendar 2025 to approximately 50% in 2026. This estimate comes from Sandisk's internal market model.

That qualification matters. Independent data confirms strong enterprise demand, but market size and growth forecasts vary widely by methodology.

TrendForce reported that revenue among the five largest NAND suppliers rose 83.7% sequentially during the first calendar quarter of 2026. It attributed the increase to enterprise SSD demand, limited supply, and higher prices.

The same NAND market survey found that major suppliers planned almost no new production capacity during 2026. Technology transitions would provide most additional bits.

Sandisk is therefore benefiting from two reinforcing forces. AI systems need more flash, while manufacturers have avoided the aggressive capacity expansion that damaged earlier cycles.

That combination pressures cloud operators and enterprise buyers. They need storage to support new services, but they face allocation, higher costs, and longer commitments.

It also pressures rival NAND manufacturers. Micron, Samsung, Kioxia, and SK hynix must decide how quickly to increase output without recreating an oversupplied market.

Long Contracts Are Sandisk's Answer to the NAND Cycle

Sandisk is trying to replace volatile spot demand with commitments that protect both supply access and future cash flow.

Management calls these arrangements new business models, or NBMs. They combine multiyear volume obligations, negotiated pricing structures, and financial guarantees.

By the end of the quarter, Sandisk had signed agreements with eight customers across Datacenter and Edge. Five agreements were added during the latest reporting period.

Three of those five involved new customers. Two expanded existing relationships. Three closed before the quarter ended, while two closed afterward.

The agreements covered $93.9 billion in minimum contracted revenue at floor pricing. Sandisk reported $59.8 billion of remaining performance obligations at quarter-end.

Including agreements signed after the quarter, remaining performance obligations reached $91.1 billion. This measure represents contracted revenue that the company has not yet recognized.

Sandisk also reported $16.5 billion in financial guarantees. These included cash deposits and other instruments supporting customer commitments.

The agreements had a weighted average duration exceeding four years. Individual arrangements can extend to five years.

About half of Sandisk's fiscal 2027 bits were committed under these models. The figure increased to approximately two-thirds for fiscal 2028.

Those commitments offer customers more reliable access to scarce NAND. They give Sandisk better visibility into demand, production needs, and cash collection.

The pricing structure is designed to divide risk. Agreements include fixed and variable components, with floors and ceilings applied to the variable portion.

A price floor protects Sandisk if market prices decline sharply. A ceiling protects customers from unlimited exposure during a shortage.

Management says the contracts provide attractive margins even at floor pricing. Investors cannot independently test that claim because Sandisk has not disclosed individual contract economics.

The mechanism nevertheless addresses a real weakness in the memory business. Suppliers traditionally add capacity during strong markets, only to face falling prices when demand slows.

Longer commitments can discourage speculative production. They also shift some forecasting responsibility to large customers.

This approach is not unique to Sandisk. Kioxia has also emphasized multiyear agreements as it redirects its business toward AI infrastructure.

At its 2026 investor event, Kioxia described long-term contracts as a way to improve revenue visibility and earnings quality. Its AI growth strategy confirms that Sandisk's partner is following a similar path.

Micron has likewise reported demand exceeding available supply across major product categories. Its fiscal third-quarter results showed how broadly the AI memory cycle extends beyond one company.

Micron's record quarterly results reinforce the bullish demand picture. They also show that Sandisk faces capable competitors seeking the same strategic customers.

The contest is therefore not Sandisk against weak rivals. It is long-term contracting against the old spot-driven memory cycle.

That opponent matters because the industry has experienced repeated shortages and gluts. Strong demand can encourage new investment, faster technology ramps, and customer inventory accumulation.

Each response appears rational in isolation. Together, they can eventually push supply beyond consumption.

Sandisk forecasts mid-to-high-teens supply growth as it ramps BiCS8 and BiCS10. Higher inventory will reduce sellable bit growth to the mid-teens, according to management.

The company expects capital investment to represent about 6% of annual revenue. Its joint-venture structure and external funding reduce the cash burden of manufacturing expansion.

During the July quarter, Sandisk's share of joint-venture gross capital spending was $519 million. Total Sandisk gross capital expenditure was $562 million.

External funding, wafer purchases, and other mechanisms reduced total cash capital expenditure to $153 million. That equaled 1.7% of quarterly revenue.

This capital efficiency supports free cash flow, but it does not remove execution risk. Sandisk still depends on successful node transitions, equipment availability, customer qualifications, and Kioxia's manufacturing cooperation.

The contracts introduce their own risks. Customers can misjudge future requirements. Product architectures can change. Financial guarantees can become difficult to enforce during a severe downturn.

Long agreements can also limit Sandisk's upside if spot prices exceed negotiated ceilings. Stability requires giving up part of the benefit available at the top of a cycle.

That tradeoff is intentional. Sandisk is trying to exchange some theoretical upside for a more dependable earnings base.

The fourth-quarter numbers do not yet prove that strategy works through a downturn. They show that customers will sign commitments during a shortage.

A true test will come when available supply begins catching demand. Contract performance under weaker pricing will reveal whether NBMs changed the cycle or merely documented its strongest phase.

What the Record Margin Does Not Prove

An 84.6% gross margin shows extraordinary pricing power, but it cannot distinguish structural improvement from a market peak.

Sandisk presented fiscal 2026 as a fundamental inflection point. Datacenter became a major business, BiCS8 reached most of bit production, and long-term contracts expanded.

Those changes are tangible. However, the company also benefited from a supply environment that lifted prices faster than volumes.

The fourth-quarter gross margin increased 58.2 percentage points from the prior year on a non-GAAP basis. Such movement deserves scrutiny precisely because it is so large.

A durable margin improvement normally comes from better products, lower production costs, richer customer mix, or stronger contractual protection. Sandisk benefited from all four, according to management.

Cyclical scarcity added another force. When customers compete for limited bits, suppliers can raise prices faster than their underlying costs.

Sandisk estimates the global NAND market will exceed $300 billion in calendar 2026, tripling from the prior year. It forecasts a $500 billion market during 2027.

These are aggressive projections sourced to a TechInsights report cited in Sandisk's presentation. They should be treated as forecasts, not established outcomes.

The company further expects customer demand to grow faster than its supply. Management believes NAND bits will remain allocated beyond calendar 2027.

Competitor statements broadly support tight conditions. They cannot establish the exact duration or scale of the shortage.

Demand forecasts depend heavily on hyperscaler capital spending. A slower pace of AI infrastructure deployment would affect enterprise SSD orders, contract expansions, and future price negotiations.

Inference efficiency presents another uncertainty. Better caching, compression, model design, and data management can reduce storage required for a given workload.

Those improvements do not necessarily shrink total demand. Lower costs can stimulate greater usage, creating more applications and stored data. The final balance remains uncertain.

Technology competition also matters. NAND serves capacity storage, while DRAM and high-bandwidth memory handle workloads requiring much lower latency.

Hard disk drives remain cost-effective for large data sets that do not require flash performance. Cloud operators can adjust their mix among these technologies.

HBF attempts to give NAND a larger role closer to accelerators. Yet the approach needs processor support, controller development, software integration, and broad industry adoption.

Sandisk and SK hynix releasing a specification is an early coordination step. It does not ensure that major chip designers or cloud operators will deploy the format.

BiCS10 faces a more familiar challenge. Sampling must lead to successful qualification, high manufacturing yields, competitive costs, and volume production.

The Sandisk and Kioxia partnership complicates competitive analysis. The companies share foundational technology and manufacturing assets while pursuing overlapping customers.

That structure can improve scale and research efficiency. It can also expose Sandisk to operational decisions made inside a closely connected competitor.

Consumer weakness provides another warning. Sandisk's Consumer revenue fell to $556 million, even as the company reported record overall results.

Management presents Consumer as a source of diversification and channel reach. During this quarter, diversification did not prevent a sharp sequential decline.

Edge demand also deserves careful interpretation. The category produced most of Sandisk's revenue and includes several markets with different economic drivers.

PC and smartphone buyers face higher component costs when NAND prices rise. Device makers can increase storage in premium products, but mass-market demand remains price-sensitive.

Sandisk expects PCs and smartphones to return to growth during calendar 2027. That remains a forecast subject to replacement cycles, trade policy, and consumer spending.

The company also disclosed risks from tariffs, currency movements, product transitions, cybersecurity incidents, and customer concentration. These are standard disclosures, but several directly affect the current strategy.

Long-term contracts concentrate significant value among a limited number of strategic buyers. Losing one customer or renegotiating one large agreement could have an outsized effect.

The financial results were also preliminary when presented. Sandisk noted that its annual filing could include adjustments after final closing and audit procedures.

None of these issues erase the quarter. They define the evidence needed before calling current profitability structural.

Sandisk has demonstrated demand, pricing power, and contracting progress. It has not yet demonstrated how those advantages behave when NAND availability improves.

Three Signals Will Test Sandisk's 2026 Thesis

The next test is whether Sandisk can preserve its economics as supply grows, customers deploy contracted bits, and competitors respond.

The first signal is Sandisk's investor day on August 13, 2026. Management scheduled the event eight days after the earnings call.

Investors should look for more detail on contract pricing, customer concentration, production commitments, and capital requirements. Clear disclosure would make the claimed durability easier to evaluate.

The event should also explain how Sandisk separates ordinary purchase commitments from genuinely protective arrangements. Contracted revenue matters most when customers cannot easily reduce obligations during weaker conditions.

If management provides measurable targets for margins and free cash flow across a cycle, the structural thesis will strengthen. Vague long-term optimism would leave pricing as the clearest explanation for current results.

The second signal is fiscal first-quarter 2027 performance. Sandisk expects revenue between $10.3 billion and $10.8 billion, with non-GAAP gross margin between 83% and 85%.

Results within those ranges would confirm that favorable conditions continued after the fiscal year ended. They would also show that higher quarterly revenue did not immediately dilute margins.

The mix will matter more than a simple guidance beat. Investors should track how much growth comes from volume, pricing, and Datacenter shipments.

Another quarter led mostly by pricing would still generate significant earnings. It would provide less evidence that physical demand can sustain the revenue trajectory independently.

Performance below guidance would weaken the Sandisk fiscal 2026 earnings thesis. It would raise questions about customer timing, price assumptions, or the speed of product ramps.

The third signal is competitor supply behavior through early 2027. Samsung, Kioxia, Micron, SK hynix, and other producers all benefit from the shortage.

If they keep capacity disciplined, Sandisk's contracts and allocation outlook gain credibility. Moderate bit growth would let AI demand absorb new supply without a sudden price correction.

Aggressive expansion would create a different scenario. New capacity, improving yields, and faster node transitions could increase available bits before contracted demand fully develops.

Watch enterprise SSD qualifications as well as factory announcements. Supply only affects the competitive market after products achieve acceptable yields and pass customer testing.

Sandisk's own BiCS10 ramp will be part of that equation. A smooth transition can lower costs and increase output, supporting margins even if prices stop rising.

A delayed ramp would restrict supply but create operational problems. Scarcity is helpful only when Sandisk can deliver enough qualified products to satisfy commitments.

The Sandisk fiscal 2026 earnings call established a credible strategic direction. The company has moved toward enterprise storage, secured multiyear commitments, and generated exceptional cash flow.

It also entered this new phase during one of the strongest pricing environments the NAND market has experienced. That makes every durability claim harder to isolate.

Developers and AI product teams should care because storage economics shape inference costs. Faster and denser flash can support larger retrieval systems, longer histories, and richer agent workflows.

Enterprise buyers should care because allocation contracts can change procurement. Guaranteed access may require longer commitments and reduced flexibility.

Investors should focus on the mechanism, not the superlatives. Datacenter bit share, contracted volumes, gross margins at price floors, and disciplined industry supply will decide whether the change lasts.

The next question is straightforward: can Sandisk deliver growing physical volume while holding margins after pricing stops doing most of the work? The coming investor disclosures and quarterly results should provide the first real answer.