SanDisk's Contracts Turn NAND's Peak into a Floor — the Floor Is the Bet

A reformed NAND-investor habit is to assume the story now divides into two halves: before the AI supercycle and after it. In that telling, SanDisk — the pure-play flash memory maker Western Digital spun off in 2025 — is the beneficiary of a demand boom that finally broke memory's boom-bust cycle. The stock, near $1,777 after more than a 600% gain this year, has been the market's reward.

That conclusion is only half right. The driver of SanDisk's surge is not unit demand. It is a pricing supercycle — supply-constrained flash sold at record prices to hyperscalers — and the far more interesting development is that the company is trying to lock that peak in. Through a set of multi-year contracts it calls the "New Business Model," SanDisk is converting the top of a pricing cycle into guaranteed revenue with a price floor and a price ceiling. The mechanism attacks the historical cycle, but it does not end it. It relocates the bet: the question is no longer whether NAND prices can keep rising, but whether the floor holds.

The surge is a pricing story, not a volume story

Strip the AI narrative down to the reported numbers and the mechanism is plain. In the fiscal fourth quarter ended June 2026, revenue rose 51% sequentially to $8.97 billion, and SanDisk said roughly one-third of that sequential growth came from higher volumes and two-thirds from higher pricing. Fiscal 2026 exited at an 84.6% non-GAAP gross margin — a number the memory industry rarely touches outside a shortage, and one that is only meaningful because of what is behind it, not what it promises.

The market has split into two distinct sub-markets, and the economics sit on one side of the divide. Enterprise SSDs reached roughly 48% of global NAND bit shipments in the June quarter, up from 26% a year earlier, and datacenter now represents about a quarter of SanDisk's revenue, up from roughly 12%. The consumer and commodity flash that historically dragged the cycle down is a shrinking share of the mix. That enterprise weighting is what turns a cyclical pricing spike into what the company can argue is a structural one — but notice that the growth itself is still ASP-driven, exactly the kind of pricing-led move that has reversed in every prior memory cycle.

The contracts trade the ceiling for a floor

This is where the "New Business Model" earns its name. As of the August Investor Day, SanDisk had signed ten of these long-term agreements across eight datacenter and edge customers representing at least $93.9 billion of expected revenue at floor prices, covering more than half of its 2027 bit shipments and about two-thirds of 2028.

The structure is the point, because it addresses exactly where prior memory long-term agreements failed. The contracts carry committed volumes on both sides, enforceable minimum financial guarantees backed by billions in posted collateral and prepayments, and — crucially for the investor — a hybrid price with a floor and a ceiling. The pricing blend is weighted toward floating terms in the longer-dated agreements, so the company keeps some participation if spot prices keep climbing, but the ceiling means it is not capturing every dollar of an even more extreme spike. In exchange, the floor protects revenue on the way down.

The historical problem is not lost on management. CEO David Goeckeler acknowledged that memory customers have walked away from long-term deals in past downturns — which is why prior cycles always asserted themselves — and the new contracts are engineered with penalties so that a customer that abandons a commitment still pays. That is the mechanism designed to make this attempt stick where others did not.

The floor only holds if supply discipline holds

The honest read is that these contracts convert the downside of the cycle into a floor, at the cost of capping part of the upside. They are a rational trade to make at the top of a pricing cycle, but they are not free money. The floor is only as good as two things SanDisk does not fully control: its own supply restraint, and the supply restraint of everyone else.

On its own capacity, SanDisk is deliberately holding bit growth below what its technology would allow, growing output from node transitions rather than new fabs and keeping capex as a shrinking share of revenue. That is the supply-discipline playbook that has governed memory since 2022, and the contracts give it the confidence to stay disciplined. The counterpart risk is competitive supply. The Chinese producer YMTC has moved from roughly 8% to 13% of NAND revenue share, and past memory corrections have reliably arrived from below-cost capacity additions in a market that no single producer controls. The floor holds only if a downturn arrives and there is no flood of new supply to blow through the contracted price.

Valuation reflects the market's residual doubt. SanDisk trades in line with Micron on trailing earnings and, on a consensus forward basis, well below the 25–50x multiples that enterprise software companies with a comparable contracted backlog command. In other words, the market is still pricing SanDisk more as a volatile memory commodity than as a company with two-thirds of its 2028 output already sold. The gap between those two pricing frames is the entire argument.

The bet, stripped down, is whether the floor holds. If supply discipline throughout the industry holds and the enterprise/AI segment stays dominant, the contracts turn a once-volatile cycle into a stable, cash-returning annuity and the commodity multiple is too low. If competitive supply floods in and the floor price proves unsupportable, the contracts delay the reckoning rather than prevent it — and the market, which still treats this as a cyclical trade, will have paid for a floor that was never actually there.