Weekly recap cover, week ending October 30, 2026A 3D NAND word line staircase with channel holes through it, the lead topic of the week ending October 30, 2026; below it, one drawing for each topic the week covered: NAND flash (3), DRAM and HBM (2), SSDs and storage systems (1), Standards (2), Market and earnings (4), Research papers (3), New on Eyes on Flash (2).THE WEEK IN MEMORY AND STORAGEWEEK ENDINGOctober 302026LEAD · NAND FLASH3212432

Week endingFriday, October 30, 2026Covers October 24 to October 30, 2026 · 17 items

Record Memory Earnings, Solidigm Weighs US Capacity, and Kioxia Folds In a Fab

Earnings season set the tone. SK hynix reported record quarterly results and Seagate reported its fiscal first quarter, both leaning on AI demand. On the NAND side, SK hynix told investors that Solidigm is reviewing US capacity but has decided nothing, and Kioxia said it will fold its Iwate fab subsidiary into the parent company.

On the technology side, Kioxia described stacking several NAND array wafers on one logic wafer, Micron and Lenovo showed LPCAMM2 in a mobile workstation, and NVM Express finished its live migration feature set. Eyes on Flash added a UFS guide that starts a new series and a hands-on guide to running language models locally.

Top stories

The rest of the week

NAND flash

DRAM and HBM

Standards

Market and earnings

Research papers

  1. Paper proposes 3D ferroelectric NOR flash as main memory (opens in a new tab)

    The authors built back end of line ferroelectric NOR arrays that hold multilevel and fast rewritable cells together. The system level speedups come from simulation, not hardware, and the work is a preprint.

    arXiv ·

  2. Paper moves fully homomorphic encryption work into the storage device (opens in a new tab)

    Rosemary argues that encrypted computing spends most of its time moving data between storage and compute, and proposes accelerating it inside the storage device to cut that movement.

    arXiv ·

  3. Paper tiers idle AI agent sessions into high bandwidth flash (opens in a new tab)

    The letter proposes placing session state for agent style language model serving in high bandwidth flash, so an idle session does not have to be restored from an SSD into GPU memory. The reported gains come from the authors' own evaluation.

    IEEE Computer Architecture Letters ·

New on Eyes on Flash