Sandisk: The Stock Is Riding ASP, Not Units. The $42 Billion Backlog Is the Real Story.

Wall Street has a clear explanation for Sandisk's run. AI data centers need more NAND flash memory for storage, demand is exploding, and the stock - up more than 4,000% since its February 2025 spinoff from Western Digital - is simply reflecting that demand surge. Analysts have piled on, with Goldman Sachs boosting its price target to $2,200 and Bernstein lifting its target to $3,000 in July.

That explanation gets the direction right but misses the mechanism. The NAND recovery is not being driven by a surge in unit demand. It is being driven by constrained supply following the 2023 NAND downturn, when manufacturers collectively cut production and have yet to replace that capacity. Sandisk is riding ASP (average selling price) expansion, not volume growth. The distinction determines how long this cycle lasts, what breaks it, and whether the stock's current valuation is defensible.

What Actually Carries the Numbers

Sandisk's fiscal third quarter of 2026 closed with $5.95 billion in revenue, up 97% sequentially and 251% year over year. Gross margin expanded to 78.4% from 50.9% a quarter earlier and 22.5% a year earlier. Non-GAAP EPS reached $23.41. The company that was losing money in the same quarter a year ago earned $3.6 billion in net income.

Revenue growth in NAND has two components: unit shipments and ASP. In the 2023–2024 downturn, the NAND industry oversupplied, prices collapsed, and manufacturers collectively cut production. That supply discipline was the first driver of recovery. As Samsung, Kioxia, and others restrained output, ASPs recovered even before unit demand returned to normal. Sandisk's fiscal Q2 and Q3 revenue nearly doubled sequentially - from $3.0 billion to $5.95 billion - while gross margin expanded by 27 percentage points in a single quarter. That margin explosion is pricing, not volume. A unit shipment surge does not compress cost of goods sold by 27 percentage points in one quarter.

Samsung publicly forecast that the memory supply shortage will worsen in 2027 and persist into 2028. TSMC is expected to raise contract manufacturing prices by up to 10% in 2027, with some products facing increases of up to 20%. The supply-constrained environment that is supporting Sandisk's ASP trajectory is not transient. It is the baseline assumption for the next two years.

The Structural Shift: Multi-Year Agreements Replace the Spot Market

Where the market has been less attentive is the business model change Sandisk has executed since the spinoff. The company calls it the "New Business Model" - multi-year customer agreements backed by firm financial commitments.

As of the end of fiscal Q3, Sandisk had three signed agreements. It signed two additional agreements during the fiscal fourth quarter, bringing the total to five. The remaining performance obligations and contracted backlog stood between $41.6 billion and $42 billion. Sandisk's entire 2026 enterprise AI storage capacity is sold out under long-term contracts.

This matters because NAND has historically been traded largely on a spot-market basis. Prices swung with quarterly supply-demand imbalances. Companies had no pricing floor and no revenue visibility beyond a single quarter. The long-term agreements flip that structure. Bernstein estimates a pricing floor of roughly $0.29 per gigabyte across these contracts, broadly in line with Q2 2026 average selling prices. Even in a worst-case scenario where ASPs collapse to $0.11/GB - a decline worse than the 2010 downturn - Bernstein estimates Sandisk's fiscal 2030 EPS would still be $214 per share, because 60% of volumes would be covered by the agreements.

That is the bold move the analyst community is reacting to, even if the headlines have not fully captured it. Sandisk is no longer a cyclical NAND play. It is transitioning into a contracted-storage business with pricing floors and five-year visibility.

The Valuation Problem

The numbers above do not resolve the valuation question. At a current price near $1,385, Sandisk trades at 46 times trailing earnings and 37 times EV/EBITDA. For comparison, Micron - the other major beneficiary of the AI storage cycle - trades at 20 times earnings and 14 times EV/EBITDA. Western Digital, Sandisk's former parent, trades at 30 times earnings and 48 times EV/EBITDA.

Sandisk's premium is justified only if two conditions hold simultaneously. First, the supply-constrained ASP environment must persist long enough for the backlog to generate sustained earnings. Sandisk guides Q4 revenue to $7.75 billion to $8.25 billion and non-GAAP EPS to $30 to $33, with gross margins near 80%. Annualized, the guided quarter implies roughly $126 in per-share earnings. At $1,385, the stock is trading at roughly 11 times that annualized pace - still rich for a hardware business, but not unsustainable if the earnings trajectory continues.

Second, the company must actually deliver on that trajectory. The capex number tells part of the story here. Sandisk's trailing twelve-month capital expenditure is only $179 million, or roughly 1% of its current market capitalization and a fraction of its $4.5 billion in trailing free cash flow. The company generates far more cash than it reinvests. That is a sign of how lightly capitalized the flash business is relative to its revenue scale - NAND is sold to fabless customers who own the fabs - but it also means Sandisk is not building supply capacity that would extend the pricing cycle. It is fully dependent on industry-wide supply discipline at Samsung, Kioxia, and SK Hynix.

The Two-Market Split

The NAND market has bifurcated into two distinct sub-markets, and the split determines which players capture value.

One market is the data center AI storage segment, where Sandisk's BiCS8 chips - which pack 15% to 19% more data into a smaller footprint while consuming roughly 13% less power - are in structural shortage. Data center revenue at Sandisk surged 645% year over year in Q3 to $1.467 billion. This segment is what the multi-year agreements cover.

The other market is the consumer and edge segment, where pricing is still more cyclical and volumes are larger but margins are lower. Edge revenue grew 295% year over year in Q3 to $3.663 billion, now the largest end market by dollar value. Consumer revenue grew 44% year over year to $820 million. These segments are not locked in by multi-year pricing floors.

The implication is fairly straightforward. Sandisk's earnings power over the next two years is structurally underpinned by the data center backlog. But the consumer and edge segments - which still account for most of the company's revenue - remain exposed to spot-market pricing. If the broader NAND cycle turns, the agreements protect a meaningful portion of earnings but not the entire business.

The Zero-Debt Factor

Sandisk carries no net debt, has $3.7 billion in cash, and has authorized a $6 billion share repurchase program. In a semiconductor business that just completed a historic profit swing, a debt-free balance sheet gives management options that cyclical competitors do not have. The buyback, if executed, would compound per-share earnings in a business that is already expanding margins at an extraordinary rate.

The company also trades on roughly 158 million diluted shares - a small float for a company of this revenue scale. Any further reduction in share count through buybacks amplifies the per-share trajectory.

The August 5 Earnings Test

Sandisk reports fiscal Q4 results on August 5. The company has already guided for revenue of $7.75 billion to $8.25 billion and non-GAAP EPS of $30 to $33. The question is not whether it will miss - the backlog makes that unlikely - but whether the guidance for fiscal 2027 justifies the current valuation.

At $1,385, the stock is already 94% below its all-time high of $2,354 and has pulled back roughly 30% from that June peak. A recovery to the high would require annualized earnings near $245 per share at current multiples - nearly double the guided Q4 annualized pace. That would need NAND pricing to keep climbing well into 2027, and for the multi-year agreements to generate the earnings compounding the market is pricing in.

The stock's YTD return of 478% and rolling annual return of nearly 2,900% mean the margin for error has evaporated. Every quarter now has to confirm the structural thesis, not just beat estimates.

Investor Takeaway

The key issue is not whether AI demand for NAND storage remains healthy. It has already been validated by the $42 billion backlog and the five multi-year agreements. The more important question is whether the supply-side conditions that are driving ASPs higher will persist once the agreements lock in pricing floors.

Sandisk has made a structural transition from a cyclical spot-market business to a contracted storage company with multi-year revenue visibility. That is a genuine competitive upgrade, and it provides meaningful downside protection compared to peers without similar agreements. But the valuation still requires the cycle to hold. If Samsung, Kioxia, and the rest of the industry maintain supply discipline - as Samsung has publicly committed to doing through 2028 - the earnings trajectory supports the current price over time. If supply restraint fractures and ASPs fall, the agreements protect 60% of volumes but leave the consumer and edge segments exposed. That is the binary the market has not yet priced in.