SanDisk Stock Surges 649%: How AI Turned Storage Into a Hot Investment

AI needs fast processors to run, but it also needs somewhere to put the information those processors produce. That second requirement has made SanDisk one of 2026’s most dramatic stock stories. The shares rose about 649% year to date through the last completed US session, while the company’s NAND flash revenue and margins surged. The question for investors is whether AI has changed the economics of storage for years to come or whether today’s share price assumes that an unusually favourable memory cycle will last.

Let's break down the rally, how AI uses SanDisk’s products, what the latest accounts actually show and how Micron, Kioxia, SK hynix, Western Digital and Seagate compare.

How much has SanDisk stock risen in 2026?

SanDisk shares (SNDK) closed at $1,777.80 on 25 September 2026 versus $237.38 on 31 December 2025. The calculation is ($1,777.80 ÷ $237.38 − 1) × 100 = 648.9%. The 649% headline is therefore a year-to-date price return through Friday’s close, not a one-year return or a live 28 September price. The stock was also about 24.5% below its $2,354.39 intraday peak on 22 June. Those two observations can both be true after a volatile rally.

SanDisk became a separate publicly traded company after Western Digital spun off its flash business in February 2025. Today Western Digital (WDC) is mainly a hard disk drive business, while SanDisk focuses on NAND flash and products built with it. Treating the two tickers as the same storage exposure would miss the different technologies and economics.

The rally has more than one driver. SanDisk’s reported earnings have risen sharply, investors expect AI servers to use more flash storage and NAND prices have climbed as supply has stayed tight. The company has also signed longer-term customer agreements that may make demand more visible. A share price gain of 649%, however, is a market return; it cannot be attributed precisely to AI demand alone.

Why do AI systems need NAND flash when they already have HBM?

The terms sound similar but the jobs are different. High-bandwidth memory, or HBM, is a specialised form of DRAM placed close to AI processors. It feeds active calculations at high speed. NAND flash keeps information after power is switched off and is used in solid-state drives, or SSDs, to store large amounts of data. Hard disk drives, or HDDs, provide another, generally cheaper way to store very large data sets, though their access patterns and speeds differ from SSDs.

Think of a busy kitchen: HBM and DRAM are the ingredients at the cook’s fingertips while SSDs are nearby shelves and HDDs are the larger storeroom. All three can become more valuable when the kitchen serves more meals, but no single product automatically replaces the others.

AI training needs data sets and saved model checkpoints. AI services also generate queries, outputs and temporary information that may need to be kept or fetched quickly. These uses support demand for storage near compute systems. Yet more AI activity does not translate one-for-one into NAND sales: software design, data retention policies, product pricing and the split between SSDs and HDDs all matter. SanDisk is also working with SK hynix on a proposed High Bandwidth Flash standard for AI inference. A technical specification is a step toward possible products, not evidence that HBF already contributes material revenue. 

Sources: SanDisk fiscal 2026 Form 10-K; SanDisk and SK hynix, August 2026 HBF announcement.

What do SanDisk’s latest results say about demand and pricing?

SanDisk’s fiscal fourth quarter ended 3 July 2026. It reported $8.965 billion in revenue, up 51% from the preceding quarter and 372% from a year earlier. Management said roughly two-thirds of the sequential revenue increase came from higher pricing and one-third from higher volumes. This is the key distinction in the entire investment case: AI-linked demand is rising, but a large part of the near-term earnings explosion comes from what customers pay per unit of memory. 

Source: SanDisk fiscal Q4 2026 earnings release, filed with the SEC on 5 August.

Source: SanDisk fiscal Q4 2026 earnings release. Dollar values rounded to the nearest million. GAAP means results prepared under US accounting rules.

For the full fiscal year, revenue reached $20.248 billion, up 175%, and GAAP operating profit was $12.389 billion, compared with an operating loss in fiscal 2025. Operating cash flow rose to $11.671 billion from $84 million. These are real reported improvements, though the 2025 comparison spans the period when SanDisk became independent and memory prices were far lower. A cash flow figure should also be read alongside manufacturing joint-venture commitments, rather than assuming modest directly reported property spending captures every future capacity investment. 

The customer mix tells an equally useful story:

Shares calculated from SanDisk’s $8.965bn quarterly total; segment revenue and sequential changes from its fiscal Q4 2026 earnings release.

Datacenter revenue grew 437% across fiscal 2026, reaching $5.153 billion for the year. It is a genuine growth engine. But edge generated most of the latest quarter’s sales and also grew rapidly. Describing every dollar of SanDisk’s revenue growth as a direct sale into AI data centres would exaggerate the evidence. Price increases across other NAND uses have helped too. 

Can long-term customer contracts make NAND profits less cyclical?

Memory markets have a familiar problem: when supply runs short, prices rise, producers make extraordinary profits and new capacity eventually pressures prices. SanDisk is trying to reduce that exposure with what it calls New Business Model agreements, or NBMs. These use customer volume commitments, financial guarantees and pricing arrangements across multiple years. At its August investor day, the company said agreements with eight customers covered roughly half of expected fiscal 2027 bit output and two-thirds of expected fiscal 2028 bit output. Bit output refers to the quantity of memory, not a share of revenue or profit. 

This can improve planning: a customer that commits to volumes before a fabrication plant expands gives the supplier better visibility. It does not make margins risk-free. Floor prices can limit downside on covered units, while uncovered supply, product mix, capacity costs and the strength of customer demand still affect results. Some customers may ultimately want to renegotiate when market conditions change. The quality of these agreements will be judged by cash collected and profitability through the next weak pricing period, not only by the value announced when the market is strong.

SanDisk and Kioxia also share a major manufacturing and technology relationship in Japan. Their August announcement described plans to invest more than $31 billion in NAND production over a multiyear period. That is a combined strategic investment plan, not $31 billion of SanDisk’s own immediate capital expenditure. The partnership improves access to manufacturing scale but creates investment obligations and links SanDisk’s execution to its partner. 

Management’s longer-range model calls for mid-to-high-teens annual revenue growth during fiscal 2028–2030, around 80% non-GAAP gross margin, about 75% non-GAAP operating margin and an adjusted free cash flow margin near 50%. These are management targets, not reported results or a guaranteed floor. They imply that today’s unusually high profitability could become relatively durable. Investors need evidence across multiple quarters to assess that claim. 

Is SanDisk stock expensive after a 649% rise?

Using the $1,777.80 close and the 146.419 million shares disclosed as outstanding on 7 August gives an indicative market capitalization of about $260.3 billion. Dividing that by reported fiscal 2026 revenue of $20.248 billion gives 12.9 times historical annual sales. Dividing it by fiscal 2026 GAAP net income of $11.433 billion gives 22.8 times historical earnings. The share count may have changed because of repurchases, so these are dated approximations. 

The earnings multiple looks much lower if the latest quarter is projected forward, but that is exactly where a cyclical stock can mislead. SanDisk guided fiscal Q1 2027 revenue to 10.3billion–10.8 billion, GAAP gross margin to 83.0%–84.9% and non-GAAP diluted EPS to 44–46. Four quarters at the $45 EPS midpoint would make an illustrative annualised $180 per share and a roughly 9.9 times price-to-earnings ratio. It is not fiscal 2027 guidance: it assumes an entire year looks like one exceptionally profitable guided quarter. 

One further accounting detail helps keep the comparison honest. Fiscal Q4 GAAP net income of $6.903 billion included an approximately $804 million gain on equity securities below operating income. That gain is not the profit from selling more SSDs. Operating profit and cash generation are more useful for understanding the flash business than treating every cent of quarterly EPS as repeatable. 

Here is a deliberately simple margin test, holding quarterly revenue at the $10.55 billion guidance midpoint and holding operating expenses at an illustrative $594 million. Each ten-percentage-point drop in gross margin reduces quarterly operating profit by about $1.055 billion, before tax, changes in costs and volume or any investment gains.

Illustration only: $10.55bn revenue × assumed gross margin − $0.594bn expenses. The $0.594bn expense assumption is the midpoint of SanDisk’s Q1 FY2027 GAAP operating-expense guidance. This is neither a company forecast nor a share-price target.

That calculation explains both sides of the investment debate. SanDisk could remain profitable even if margins eased considerably, yet billions of quarterly profit would disappear while the stock’s valuation might also fall. Conversely, sustained demand, contract protection and technological gains could make a higher margin more durable than in past NAND cycles. Our view is that the duration of pricing power, rather than the 649% chart itself, is the decisive issue.

SanDisk vs Micron, Kioxia, SK hynix, Western Digital and Seagate

The comparison works best by asking which part of the data chain each company sells. A simple price-to-earnings ranking across them would mix different accounting periods, currencies and products at a moment when memory profits are unusually high.

Sources: SanDisk, Micron, Western Digital and Seagate earnings filings; Kioxia first-quarter results; SK hynix Q2 release. Periods and currencies vary, so revenue and margins are descriptive rather than directly comparable valuations.

For US-listed memory exposure, Micron (MU) is the closest large semiconductor comparison. Its latest filing said NAND sales increased 99% sequentially in fiscal Q3 2026, driven mainly by an approximately mid-80% rise in NAND average selling prices and a smaller rise in bits shipped. That independently supports the conclusion that sector pricing, not only SanDisk-specific technology, is driving the boom. Micron’s much larger DRAM and HBM businesses also mean its stock is exposed to another part of the AI infrastructure budget. Its next fiscal results were scheduled for 30 September, so using them in an article dated 28 September would be premature. 

Sources: Micron fiscal Q3 2026 Form 10-Q and investor relations calendar.

Kioxia deserves special attention because its joint factories and NAND research overlap with SanDisk’s supply base. This is a deeper operating relationship than an ordinary competitor comparison. SK hynix and Samsung are broader Korean memory competitors with DRAM, HBM and NAND businesses. Their profitability also surged in Q2 2026, consistent with a broad memory upswing, but SanDisk’s 84.6% quarterly gross margin should not be read as proof that it alone has a permanent moat.

On the storage side, Seagate (STX) and Western Digital show why AI data growth can benefit more than flash. Their HDDs are useful when a customer needs huge capacity at a different cost and performance point. An AI data centre can buy HBM, SSDs and HDDs simultaneously. For investors seeking a broader overview of listed US businesses, INDmoney’s semiconductor stocks page offers a separate starting point; a sector page is not a substitute for examining each company’s product mix.

What should Indian investors watch next?

SanDisk’s next earnings release needs to show whether guided Q1 FY2027 sales and gross margin were achieved, and how much growth came from shipments rather than further price increases. Track datacenter revenue alongside edge revenue, the share of output covered by NBMs, realised cash collection and spending on the Kioxia manufacturing partnership. The most useful warning sign would be NAND prices weakening while the market continues to value the stock as if current margins are permanent.

The India connection is mainly through global investment and the memory supply chain. Indian investors can compare US-listed SNDK with MU, WDC and STX; Kioxia, SK hynix and Samsung trade in their home markets and require different access routes. India’s semiconductor assembly and testing ambitions, including Micron’s Indian facility for DRAM and NAND products, are related to the industry but do not turn an Indian stock into a direct equivalent of SanDisk’s NAND manufacturing economics. Currency movements also affect an Indian investor’s rupee return separately from the US dollar stock return. 

Our view: The improvement in SanDisk’s accounts is too large to dismiss as hype. Fiscal Q4 revenue, margins and cash generation show how much buyers currently value NAND capacity. But management itself attributes most of the latest sequential revenue increase to higher prices. The stock is therefore a concentrated bet that contracts, product improvements and AI-driven storage demand can sustain unusually high returns when the supply cycle turns. A careful comparison with Micron’s memory mix and the HDD suppliers’ different cost structure is more informative than treating every AI storage stock as the same trade.