Cerebras can't meet sky-high AI data center share hopes, so beating estimates still sells off
KV cache offloading turns NAND into AI memory, and the purest flash maker captures the demand surge directly.