Sandisk Just Made the Next Memory Crash a Lot Less Scary

Sandisk (NASDAQ:SNDK) earned $6.9 billion of net income in its latest quarter, largely because memory prices went on an extraordinary run. The market clearly doubts the run can last.

The growth stock still sits more than a quarter below its 52-week high. And the stock costs only about 8 times expected fiscal 2027 earnings. A price like that assumes much of today's profit won't survive the cycle.

The flash memory specialist's answer is written into contracts. It now has 10 long-term supply agreements with eight data center and edge customers, and they are expected to produce at least $93.9 billion of revenue over their lives -- assuming prices settle at their contractual floors. For scale, fiscal 2026 revenue, up 175% year over year, was $20.25 billion.

How much downside protection does a floor like that buy?

Image source: Getty Images.

A $93.9 billion minimum

The agreements (Sandisk calls them New Business Model agreements) commit the company to deliver, and its customers to buy, set volumes of flash memory over multiyear terms -- more than four years on a weighted-average basis, and up to five. Pricing combines fixed and variable elements, and the variable part is subject to floors and ceilings. The $93.9 billion is the minimum those terms produce if every variable price lands at its floor. It isn't an annual figure or a conventional backlog -- it's contracted revenue spread across the agreements' lives. The agreements also carry financial guarantees (customer cash deposits and other instruments totaling $16.5 billion) in case a buyer walks away. And on the company's August earnings call, chief financial officer Luis Visoso said Sandisk expects them to cover more than half of its bits (the volume of memory shipped) in fiscal 2027 (the fiscal year that began in July), and about two-thirds the following year.

Notably, the floor assumption cuts only one way. If market prices hold above the floors, revenue comes in higher, up to the contracts' ceilings.

The contracted book is still building, too. Remaining performance obligations (contracted product not yet delivered) went from $41.6 billion in early April to $59.8 billion by July 3. And two agreements signed after the fiscal year closed, with a combined contract value the annual report puts at $31.3 billion, aren't in that total.

How bad could the next bust be?

Sandisk's recent history shows what an unprotected downturn looks like. In the final quarter of fiscal 2025, the company generated just $1.9 billion of revenue, ran a 26.2% gross margin, and posted a small net loss. Four quarters later, revenue was $8.97 billion, gross margin was 84.6%, and net income came to $6.9 billion.

Most of that swing came from price, not volume. Management said higher pricing accounted for about two-thirds of the quarter's growth from the prior quarter. And its outlook asks for more of the same: fiscal first-quarter 2027 revenue of $10.3 billion to $10.8 billion, with non-GAAP gross margin expected to hold between 83% and 85%.

The floors are aimed at the reverse trip. In fiscal 2025, nothing stood between Sandisk's revenue and a falling spot price.

If the cycle turns now, more than half of this fiscal year's volumes can't reprice below their contractual minimums, whatever the spot market does. That, I'd argue, is the biggest change in Sandisk's story.

"We expect attractive margins even at floor pricing," Visoso said on the August call.

A price floor isn't a profit floor

However, it's worth noting what that promise covers. Attractive margins at the floor make a case for staying profitable -- not a case that an 84.6% gross margin survives a downturn. In fact, the multi-year model management presented at its August investor day assumes non-GAAP (adjusted) gross margin settles near 80% for fiscal 2028 through fiscal 2030. And management hasn't said how far below today's prices the floors sit.

The rest of the business has no floor at all. Nearly half of this year's bits still sell at whatever the market pays. And no downturn has tested the structure, or customers' willingness to keep paying above-market minimums through one.

Ultimately, the downside case shrinks, but it doesn't go away. A memory crash would still hit nearly half of Sandisk's volumes at full force, and it would still pull contracted pricing down toward the floors.

What it arguably can't do anymore is drag the company back to $1.9 billion quarters and a net loss, as long as customers honor their agreements.

At about 8 times expected fiscal 2027 earnings, I think the stock is priced for a steep decline in earnings, and the contracts make the harshest versions of that decline harder to reach. Still, I'd like to see one quarter where memory pricing falls and margins hold before treating the floors as proven. Until then, I'm not a buyer.

Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has no position in any of the stocks mentioned. The Motley Fool has a disclosure policy.