Google Backs SK hynix's HBF as Next-Gen AI Memory

Google DeepMind has identified High Bandwidth Flash (HBF) as the memory technology that will lead the next-generation AI era. HBF, currently under development by SK hynix, is a next-generation memory that stacks NAND flash to increase storage capacity and bandwidth.

At FMS 2026 in Santa Clara, California, on Aug. 6 (local time), Google DeepMind's Xiaoyu Ma pointed out that the need for HBF will grow as large language models (LLMs) require significantly more data for inference tasks. This surge in data processing demand makes it difficult to achieve fast response speeds with existing memory systems.

Ma stated that HBF, equipped with high capacity and bandwidth, can secure high power efficiency at the same time. To address concerns about NAND cell damage during frequent data recording (writing) during AI computation, SanDisk Vice President Rajiv Nagabirava proposed that HBF takes on a predominantly data-reading role.

SK hynix's Eui-cheol Lim also suggested using HBF jointly with High Bandwidth Memory (HBM) in the reading-centric AI inference stage. Google and SanDisk formed a consortium with SK hynix to announce standard specifications for HBF at the event.