AI Inference Reshaping NAND Cycle; Sandisk’s Role in Storage Stocks

The interesting claim here isn't the demand figure but the cycle mechanics: long-term agreements of this scale, tied to inference rather than training, could shift NAND from a spot-driven boom-bust market toward contracted, more visible demand. If that holds, the read-across for storage names is about earnings predictability rather than unit growth. The open question is whether these agreements genuinely dampen oversupply risk or simply defer it, since committed volumes can become a glut if inference economics change.