Sandisk has put a price floor under more than half the bits it ships

A non-GAAP gross margin of 84.6% ^[4] leaves cost of revenue at 15.4% of the top line ^[10]. At unchanged volumes and unchanged unit costs, realized prices would have to fall by about 85% before gross profit reached zero ^[11].

The agreements work on the path of a decline. If spot NAND halved while the contracted bits held exactly at their minimum, blended revenue per bit would fall roughly a quarter, because half of one plus half of a half is three quarters ^[12].

How much of that protection is real depends on where the minimums sit, and the initiation does not state the contract length or the price level at which they are set ^[14]. A minimum a few points above cash cost and a minimum near fiscal 2026 realized pricing are described identically in a summary and are worth very different amounts. Revenue nearly tripled in fiscal 2026 ^[3] on pricing power the analyst attributes to AI infrastructure and datacenter demand ^[5], so the reference prices in any new negotiation were high ones.

If the floors sit close to cash cost, Sandisk's earnings behave the way NAND earnings have always behaved. Floors near recent pricing put the downturn on the customers instead, and the risk then arrives at renewal. A supply response from competitors, which the analyst lists among the risks ^[8], would land first on the half of the book with no minimum.

Sandisk is repurchasing shares aggressively from a net cash position ^[7]. Buying at peak margins fixes the price at which those shares are retired, and the analyst's Buy rating at a $2,160 target ^[2] implies he sees the stock below that level today ^[13]. He discloses long positions in MU, SNDK and KXIAY ^[9], and names Kioxia dependence as one of the risks to the thesis ^[8].

In my view the agreements change the variance of reported earnings. The average across a full cycle is unchanged. Before paying for better visibility I would want two numbers: the floor level and the renewal dates.