Sandisk Could Be 32% Undervalued, but the AI Infrastructure Reset Faces a Test

Sandisk has entered google news with a striking claim: SNDK could be 32% undervalued after an AI infrastructure reset transformed its latest results. That conclusion rests on more than excitement about artificial intelligence. Sandisk reported that fiscal third-quarter revenue nearly doubled sequentially, while its data center business expanded 233%.

The numbers support a genuine operating change. They do not, however, settle the valuation debate. NAND flash remains a cyclical market, and the current earnings surge reflects higher pricing alongside stronger demand and a more favorable customer mix.

That distinction creates the central conflict. Bulls see Sandisk becoming a contracted supplier for long-lived AI infrastructure. Skeptics see unusually favorable memory conditions that competitors such as Samsung, SK hynix, Micron, and Kioxia will eventually challenge.

The 32% figure should therefore be treated as a model result, not an observable fact. A valuation model converts assumptions about revenue, margins, capital needs, and risk into an estimate. Small changes in those assumptions can materially alter the result.

What Changed Behind the Sandisk Google News Claim

Sandisk’s operating results changed faster than its familiar identity as a consumer flash-storage company.

The company reported fiscal third-quarter revenue of $5.95 billion, up 97% from the preceding quarter. GAAP net income reached $3.62 billion, compared with $803 million one quarter earlier. Diluted GAAP earnings rose to $23.03 per share.

Those results were also dramatically higher than the prior-year period. Revenue increased 251% from $1.70 billion, while Sandisk moved from a GAAP net loss to a substantial profit. The company’s quarterly results attribute the change to pricing and a mix shift toward higher-value customers.

Data center revenue provides the clearest evidence for the AI infrastructure thesis. It reached $1.47 billion, rising 233% sequentially and 645% year over year. This segment includes storage sold for cloud and enterprise computing environments.

Edge revenue, which covers storage used closer to where data is generated or processed, also grew sharply. It reached $3.66 billion, up 118% sequentially. Consumer revenue moved in the opposite direction, falling 10% from the previous quarter to $820 million.

The change in mix matters because enterprise solid-state drives generally address different requirements than memory cards or portable drives. Data center buyers prioritize endurance, predictable latency, capacity, power use, and qualified operation across large server fleets.

Sandisk’s gross margin illustrates how pricing and mix affected the quarter. GAAP gross margin reached 78.4%, compared with 50.9% in the preceding quarter and 22.5% one year earlier. That expansion was extraordinary for a NAND supplier.

Management described the quarter as an inflection point. It said Sandisk was moving toward high-value end markets and multi-year customer engagements supported by firm financial commitments. That statement is important, but investors still need evidence that the resulting earnings remain durable.

The company also said it ended the quarter with three signed agreements under its new business model. Two additional agreements were signed during the following quarter. Sandisk has not publicly disclosed every customer, commercial term, or pricing mechanism in those agreements.

That disclosure gap limits outside analysis. Multi-year commitments can improve demand visibility, but their value depends on minimum purchase requirements, pricing adjustments, cancellation protections, and the credit quality of each buyer.

The valuation headline emerged because these results make older assumptions look stale. Models built around modest data center sales and ordinary NAND margins would produce very different outcomes from models using the latest quarter.

Yet the same speed creates risk. A single exceptional quarter can dominate forward estimates, especially when margins move by more than 27 percentage points sequentially. Investors must decide which portion represents structural change and which portion reflects peak conditions.

AI Storage Demand Has Reset Sandisk’s Earnings Base

AI infrastructure is shifting storage from a supporting purchase into a constraint that can influence the design and economics of an entire computing cluster.

Graphics processors receive most attention in AI spending, but model training and inference also depend on moving large datasets through memory and storage systems. NAND flash provides nonvolatile storage, meaning it retains data without continuous power.

Enterprise SSDs use NAND to deliver faster access and lower latency than mechanical hard drives. They store training data, model checkpoints, retrieval indexes, logs, and information prepared for repeated inference workloads.

AI inference increases the importance of these systems. An inference service can consult model weights, cached information, retrieval databases, and user context across many simultaneous requests. Storage performance affects how efficiently expensive processors remain occupied.

Sandisk’s annual filing showed that the shift was developing before the latest quarter. Fiscal 2025 cloud revenue increased 195%, supported by a 153% increase in exabytes sold and higher average selling prices per gigabyte.

An exabyte is one billion gigabytes. It provides a measure of storage volume independent of the price attached to each unit. Rising exabytes and rising average prices together create a particularly favorable revenue combination.

The annual filing also shows why pricing cannot be ignored. Sandisk said improved pricing, product mix, and lower manufacturing underutilization charges helped lift fiscal 2025 gross profit.

Underutilization charges arise when factories operate below efficient capacity. Lower charges can improve reported profitability as demand recovers, even without a permanent change in the underlying market structure.

Sandisk continued building its technical case in July 2026 by announcing samples of BiCS10 1-terabit triple-level-cell NAND. Triple-level-cell NAND stores three bits in each memory cell, balancing density, performance, endurance, and manufacturing cost.

The company says BiCS10 offers a NAND interface speed of up to 4.8 gigabits per second. It also claims a 59% improvement in bit density over BiCS8, its earlier generation. These figures describe company targets and have not yet been demonstrated through broad customer deployment.

Sampling is an early commercialization stage. Selected customers receive components for evaluation and qualification before volume adoption. It does not mean that the technology already contributes meaningful revenue.

Still, the timing supports the wider AI storage narrative. Sandisk is directing a new generation toward data-intensive workloads while its data center revenue is expanding. Product development and market demand are moving in the same direction.

Its manufacturing relationship with Kioxia also matters. The companies extended agreements covering their Yokkaichi joint venture through 2034. The arrangement gives Sandisk continued access to shared NAND manufacturing operations without requiring it to independently replicate every fabrication asset.

This model can support capital efficiency, but it also creates reliance on a strategic partner. Sandisk identifies that reliance among the risks in its public filings. Production problems, disagreements, or delayed technology transitions could affect supply.

The AI reset therefore has two connected parts. Demand is moving toward higher-value enterprise storage, while Sandisk is trying to secure products and manufacturing capacity suited to that demand.

Neither part guarantees a lasting valuation premium. Together, however, they explain why a company once associated mainly with removable storage now attracts attention as an AI infrastructure supplier.

The Real Contest Is Structural Demand Versus the NAND Cycle

The investment case turns on whether contracts and AI workloads can overpower the memory industry’s history of oversupply and falling prices.

NAND producers manufacture standardized bits at enormous scale. When demand exceeds available supply, prices and factory utilization can rise quickly. High margins then encourage producers to improve yields, expand output, or accelerate new capacity.

That response can eventually produce excess supply. Prices fall, inventory builds, utilization drops, and profitability contracts. The pattern has repeatedly made memory earnings more volatile than end-market demand alone would suggest.

Sandisk argues that its new business model can reduce this instability. Multi-year customer relationships backed by financial commitments should provide clearer volume visibility than ordinary short-term purchasing.

The claim is plausible, but contract duration alone does not remove cyclicality. A long agreement with market-linked pricing can still expose a supplier to falling prices. A fixed-price agreement can protect margins, but it can also limit upside when shortages become more severe.

Volume commitments can secure demand while creating customer concentration. If a few hyperscalers account for a growing share of planned output, changes in their capital budgets could affect Sandisk more sharply.

The company has not provided enough public detail to model those tradeoffs precisely. Investors therefore need to separate confirmed results from management’s expectations about the future benefits of the agreements.

Competition adds another constraint. Samsung has vast semiconductor manufacturing resources and sells both NAND and other memory products. SK hynix participates through Solidigm, while Micron operates across NAND and DRAM markets.

Kioxia is both a manufacturing partner and a separate supplier competing for customers. That relationship highlights the unusual structure of NAND production, where companies can share fabrication economics while competing in finished products and sales.

Western Digital provides another useful reference. Sandisk completed its separation from Western Digital on February 21, 2025. The transaction divided flash storage from Western Digital’s hard-drive business, giving investors cleaner exposure to two different storage technologies.

The separation also complicates historical comparisons. Results from periods before that date were prepared on a carve-out basis, meaning accounting records were allocated as if Sandisk had operated independently.

This does not make the figures unusable. It does mean that comparisons spanning the separation deserve more caution than a standard same-company time series.

Hard drives remain important for high-capacity storage where cost per unit matters more than latency. NAND is better suited to workloads that need faster access, but the technologies can coexist within the same AI data architecture.

AI clusters may use fast enterprise SSDs for active datasets and hard drives for colder information. The expansion of AI infrastructure can therefore benefit several storage categories without producing one permanent winner.

The key competitive question is not whether AI creates more data. It clearly does. The issue is whether Sandisk can retain favorable economics after rivals qualify comparable products and buyers gain more sourcing options.

Technology transitions could strengthen Sandisk temporarily. BiCS10’s claimed density and interface improvements would help if customers validate them on schedule. Delays or weaker-than-expected yields would give competitors more room.

Yield measures the share of manufactured chips that meet required specifications. A dense design can look impressive technically while remaining expensive if too many dies fail testing.

Google news coverage tends to compress this contest into a single valuation percentage. The underlying business question is harder: how much of Sandisk’s current profitability survives when supply, competition, and customer bargaining power normalize?

Why the 32% Undervaluation Estimate Needs a Stress Test

A 32% discount can disappear when a model changes its assumptions about normalized margins, NAND pricing, or the cost of capital.

The Simply Wall St headline presents undervaluation as a possible outcome, not a verified market discrepancy. Its public methodology uses historical data and analyst forecasts, then applies valuation frameworks to estimate fair value.

Discounted cash flow models calculate the present value of expected future cash generation. They require forecasts for revenue, profit, reinvestment, taxes, long-term growth, and a discount rate reflecting risk.

Each input matters. A model extending the latest gross margin across several years will produce a much stronger result than one assuming margins return toward historical memory-sector levels.

The same applies to revenue. Sandisk guided fiscal fourth-quarter revenue to a range of $7.75 billion to $8.25 billion. That outlook indicates continued near-term momentum, but guidance is not a guarantee of multi-year growth.

Management also guided non-GAAP gross margin to between 79% and 81%. Non-GAAP measures exclude selected costs to present an adjusted view of operations. They should be considered alongside GAAP results and the company’s reconciliation disclosures.

Those ranges support the near-term bull case. They also raise the standard for future execution. When a valuation already assumes exceptional profitability, merely strong results can disappoint.

The Simply Wall St platform itself illustrates how different methods can conflict. Earlier analyses of Sandisk have produced conclusions ranging from large discounts to overvaluation, depending on the date and chosen framework.

That variability is not necessarily an error. Share prices, forecasts, and model inputs change. It is a warning against treating any one estimate as a precise measurement.

The 32% claim should pass at least three stress tests.

First, normalize gross margin. Sandisk’s latest 78.4% GAAP result should be compared with the 50.9% preceding quarter and 22.5% prior-year figure. A cautious model should test several levels rather than extending one quarter indefinitely.

Second, separate volume from pricing. Data center revenue expanded dramatically, but Sandisk explicitly attributed total revenue growth to both higher pricing and customer mix. Price-led growth can reverse faster than qualified enterprise deployments.

Third, test reinvestment requirements. Advanced NAND generations require development spending, equipment, manufacturing coordination, and successful yield ramps. Current cash generation does not mean future production arrives without capital or execution risk.

Investors should also examine share count assumptions. A per-share valuation can change when a company repurchases shares, issues equity, or uses stock-based compensation. Sandisk’s capital allocation therefore belongs inside the valuation, not outside it.

The discount rate creates another major sensitivity. Companies exposed to volatile pricing, concentrated buyers, or manufacturing partnerships generally warrant different risk assumptions from stable subscription businesses.

A lower discount rate increases the present value of distant cash flows. A higher rate reduces it. When much of the expected value depends on earnings several years ahead, the choice can materially change the conclusion.

The published estimate also arrives immediately before important new information. Sandisk scheduled its fiscal fourth-quarter and full-year 2026 earnings call for August 5, followed by an investor day on August 13.

That timing matters. New results, contract details, or long-term targets could quickly make the current model more useful or less relevant.

The cautious conclusion is not that SNDK must be overvalued. It is that the 32% figure represents a conditional argument. Its usefulness depends on whether readers understand the conditions.

What the Valuation Story Still Does Not Prove

Sandisk has demonstrated extraordinary recent earnings, but it has not yet demonstrated a complete cycle of durable AI-era profitability.

One unresolved issue is the source of the margin expansion. Sandisk cited higher pricing and a more favorable mix. Both can persist, but neither should automatically be treated as permanent.

The latest quarter’s gross margin exceeded the previous quarter by 27.5 percentage points. Changes of that size invite questions about inventory accounting, supply tightness, contract timing, and the sustainability of average selling prices.

Sandisk’s filings identify volatility in demand and average selling prices as material risks. They also cite competing products, technology transitions, manufacturing disruptions, and changes in customer relationships.

These are not generic legal warnings. Each risk maps directly to the current valuation debate.

Demand volatility matters because hyperscalers can revise infrastructure budgets. Pricing matters because NAND supply can respond. Technology execution matters because customers qualify components over long cycles and expect consistent performance.

Customer relationships matter because large buyers possess significant negotiating power. A supplier can gain visibility through multi-year agreements while conceding protections or pricing terms that outsiders cannot see.

The AI infrastructure reset also does not eliminate consumer exposure. Edge remained Sandisk’s largest reported end market during the fiscal third quarter. Consumer revenue still contributed $820 million despite its sequential decline.

This diversification can provide reach across several markets. It can also expose Sandisk to smartphones, personal computers, and retail storage cycles that follow different demand patterns from AI data centers.

Another uncertainty involves the new product roadmap. Sandisk’s July BiCS10 announcement describes sampling, not mass production at mature yields. Customer qualification and volume ramping remain ahead.

The claimed performance figures are relevant, but buyers will evaluate endurance, thermal behavior, error rates, controller integration, consistency, and total system cost. A fast interface alone does not determine enterprise adoption.

Competitive responses remain difficult to predict. Samsung, Micron, SK hynix, Solidigm, and Kioxia can adjust output or product priorities when enterprise SSD economics become unusually attractive.

An aggressive supply response would weaken the scarcity component of Sandisk’s margins. A disciplined industry response would strengthen the case that profitability can remain above prior-cycle levels.

Geopolitics and trade rules create additional uncertainty. NAND supply chains cross the United States, Japan, South Korea, China, and other markets. Tariffs or export restrictions can change costs and customer access.

Sandisk specifically cites trade policies, tariff regimes, currency movements, and regulatory changes among its risks. These forces can affect both manufacturing economics and the global demand available to each supplier.

The company’s new independence adds one more variable. Operating separately from Western Digital can sharpen strategic focus, but it also means Sandisk must establish its own capital allocation and public-market record.

Its latest results are encouraging. A complete judgment requires several reporting periods that show how the independent company behaves through changing supply and demand conditions.

For that reason, this analysis should not be read as a recommendation to buy or sell SNDK. The google news valuation claim is best understood as a testable thesis about future cash flows.

Three Signals That Will Decide the Sandisk Thesis

The next two weeks will provide more useful evidence than another round of percentage-based valuation headlines.

The first signal is Sandisk’s fiscal fourth-quarter report on August 5. The company previously guided revenue between $7.75 billion and $8.25 billion, with a non-GAAP gross margin between 79% and 81%.

Meeting those ranges would strengthen the view that the third quarter was not an isolated spike. Missing them would force investors to reconsider how quickly pricing, product mix, or customer demand changed.

The composition of revenue will matter as much as the total. Continued data center expansion would support the AI infrastructure reset. A sharp slowdown masked by other segments would weaken it.

Investors should also examine cash flow, inventory, average selling prices, and factory utilization. Revenue growth is more convincing when it converts into cash without requiring an unsustainable buildup of inventory or working capital.

The second signal is the investor day scheduled for August 13. Sandisk can use that event to explain its new business model, manufacturing strategy, product roadmap, and long-term financial framework.

The most valuable disclosure would address contract mechanics. Investors need to understand how minimum commitments, pricing adjustments, customer concentration, and cancellation protections influence future revenue.

Management does not need to reveal confidential customer terms. It can still provide aggregated measures that let outsiders distinguish firm demand from optimistic pipeline estimates.

Details about BiCS10 qualification and production would also help. A timeline showing customer sampling, validation, yield improvement, and expected volume contribution would connect the technology announcement to financial forecasts.

The third signal is the NAND supply response over the next one to three months. Watch capacity plans, inventory commentary, and enterprise SSD launches from Samsung, Micron, SK hynix, Solidigm, and Kioxia.

Disciplined output and persistent enterprise demand would support Sandisk’s margin durability. Rapid capacity growth, weaker contract pricing, or rising inventories would undermine the assumption that AI has neutralized the traditional cycle.

Investors should pay particular attention to the difference between unit demand and price. AI storage deployments can keep expanding while NAND prices fall if industry supply grows even faster.

That scenario would validate the technology trend but weaken a valuation model built on unusually high margins. It is the most important distinction hidden inside the 32% headline.

The Sandisk google news story is therefore not simply about whether one stock looks inexpensive. It asks whether AI has changed the economic structure of flash storage faster than competitors can respond.

Sandisk has supplied compelling early evidence. Revenue, data center sales, profit, and margins all moved sharply in its favor. Its contract strategy and next-generation NAND roadmap provide a coherent explanation for that change.

The remaining question is durability. Readers should compare the August results, investor-day disclosures, and competitor supply signals against the assumptions behind the valuation claim.

Treat the 32% estimate as a hypothesis with measurable checkpoints. If Sandisk sustains its guidance, clarifies firm customer commitments, and faces disciplined supply, the thesis gains credibility. If any of those conditions fail, the apparent discount can close without the stock moving at all.