Can flash memory help overcome AI’s growing memory bottleneck? An open standard is redefining data movement between memory and storage.
SK hynix has unveiled the first standard specifications for High Bandwidth Flash (HBF), a next-generation memory technology developed in collaboration with Sandisk to address growing bandwidth and capacity demands in AI infrastructure. Introduced at Future of Memory and Storage (FMS) 2026, the specification is designed as an open industry standard through the Open Compute Project (OCP), with participation from companies including Google and Tenstorrent.
As AI inference workloads continue to expand, conventional memory hierarchies face increasing challenges in balancing processing speed with storage capacity. HBF introduces an intermediate memory layer between High Bandwidth Memory (HBM) and solid-state drives (SSDs), combining NAND flash capacity with bandwidth characteristics closer to HBM. The approach is intended to reduce data movement bottlenecks while enabling more efficient memory architectures for AI training and inference systems.
The standard defines capacities of up to 512GB using 8-high and 16-high NAND die stacks, with bandwidth ranging from approximately 0.4TB/s to 3.0TB/s across three performance grades. It also adopts the Universal Chiplet Interconnect Express (UCIe) standard, allowing HBF devices to interface with heterogeneous processors, including CPUs and GPUs. In addition to electrical characteristics and connection interfaces, the specification covers packaging guidelines, reliability requirements and software I/O, providing a common framework for ecosystem development.
Alongside the HBF announcement, SK hynix also showcased its tenth-generation 375-layer 4D NAND technology, which the company says delivers 2.5 times higher performance per watt than its previous generation. The NAND platform is targeted at AI infrastructure such as data centres, with mass production of enterprise SSDs planned for next year.
“With the rapid spread of AI applications, we are at a point where overall data processing structures must be redesigned. Through HBF, SK hynix will expand the boundaries between memory and storage and contribute to building new architectures that enhance overall system efficiency,” say s Kim Chun-sung, Executive Vice President and Head of Solution Development at SK hynix.
Click here for the official announcement.
