SanDisk's Rally Was Never About AI Demand — That's Why Oracle's Boom Couldn't Lift It

On September 10, Oracle reported a blowout fiscal Q1, roughly 30% revenue growth and a contracted backlog that has swelled to roughly $638 billion — the kind of AI infrastructure number that is supposed to be rocket fuel for memory stocks. The next morning, the option was available: memory names rose in pre-market trading. By the close, SanDisk (SNDK) was down about 3.5% at roughly $1,633, and Micron and Western Digital fell alongside it.

That price path looks like a contradiction. An AI boom fails to lift the very company that sells the storage that boom consumes. It is not a contradiction. It is the tell that the market has been explaining SanDisk's move with the wrong driver all along. This rally never ran on AI demand. It runs on pricing, and the suppliers' willingness to keep NAND scarce. Oracle's demand news cannot lift a stock whose gains were never about a demand surprise in the first place.

Two-Thirds Price, One-Third Volume

SanDisk reports on a fiscal year with a fourth quarter ended in early July. In that quarter, revenue was $8.97 billion, up 372% year over year and up 51% sequentially, with non-GAAP earnings per share of $39.25. The company itself told investors how the sequential growth broke down: approximately one-third came from higher volumes, and two-thirds came from higher pricing.

That split is the entire story in one line. In a true demand-driven memory cycle, revenue growth would track unit shipments — more bits sold because more buyers need them. Here, volume is the minority contributor. Two-thirds of the growth is SanDisk simply charging more per gigabyte. This is a pricing phenomenon, not an AI-unit phenomenon, and it is the difference between a cyclical recovery and a supply-constrained oligopoly.

Pricing, not demand, is also the reason earnings have outrun revenue so violently. SanDisk's trailing earnings have accumulated to roughly $72.90 per share, and on a market capitalization near $239 billion the stock still trades around 21 times trailing earnings — a modest multiple for a name that has risen roughly 3,300% over the past year, precisely because the market is pricing in that extreme level of NAND price persisting.

Where the Pricing Power Comes From

NAND flash is a commodity produced by a handful of manufacturers — Samsung, SK Hynix, Kioxia, Micron, and SanDisk — and over the past year they have behaved like a functioning cartel of self-interest. The two big Korean suppliers have been throttling NAND output and diverting fab space and capital to the more lucrative AI memory lines, HBM and advanced DRAM. Industry projections put NAND capacity on a path roughly 40% below its 2022 peak by 2027. That is supply discipline, applied at the burner.

The result at SanDisk is visible in its own price book. The company raised NAND contract prices by 50% in November 2025 and was reported to be doubling prices on enterprise-class SSDs into early 2026, with NAND pricing rising 80–90% quarter over quarter at points along the way. When a supplier can raise a commodity's price 50% in a month, demand is not the binding constraint — capacity is.

This is the structural supply frame that supply-centric memory analysis keeps returning to: the cycle is no longer driven by a surge in units. It is driven by a persistent shortage that suppliers maintain on purpose. Demand is abundant, but the marginal driver of SanDisk's revenue — and of its share price — is price, and price is set by how tightly three or four companies choose to throttle output.

Why the Demand Confirmation Adds Nothing

Read Oracle's quarter the way the market treated it. The company reported a record backlog and reiterated its AI buildout, yet the memory complex could not hold its early gains. The reason is that the demand was already in the price. Nobody needed Oracle to confirm that AI buyers want more storage; hyperscaler spending has been the bull case for eighteen months and is fully discounted at a stock that has risen thousands of percent. Demand confirmation moves a stock only when demand was the uncertain, marginal variable. Here it is not; price is.

The more uncomfortable reading cuts underneath SanDisk's own revenue. If two-thirds of the growth is pricing, then the entire valuation rests on one continuous assumption: that NAND prices hold at historically extraordinary levels. That is a fragile fulcrum. The buyers on the other side are themselves stretched — Oracle spent $28.5 billion on capital expenditures in its latest quarter yet still produced roughly $5.4 billion in negative free cash flow, carrying a balance sheet analysts put at roughly $130 billion of debt, and the wider market has spent the late summer digesting reports that the largest tech companies hold trillions in off-balance-sheet AI commitments. A memory buyer's funding wobbles, or one supplier decides to add capacity to capture share, and the price that carries two-thirds of SanDisk's growth can snap far faster than demand ever moved it up.

The Condition That Matters

The prevailing story — that SanDisk is an AI demand story whose fortunes rise and fall with Oracle's backlog — is a demand-side explanation for a supply-side result, and it inverted on exactly the day it should have been confirmed. SanDisk's share price is not levered to how many AI data centers get built. It is levered to whether the NAND oligopoly keeps supply scarcer than an insatiable market wants. That is the variable to watch: NAND contract pricing and the discipline of three or four suppliers, not the next hyperscaler earnings report. As long as scarcity holds, the pricing that carries this earnings profile holds with it. The moment a supplier trades scarcity for share, the two-thirds-of-pricing growth unwinds in a hurry — and no Oracle backlog will be there to catch it.