February 26, 2026 /SemiMedia/ — SK hynix said it has teamed up with SanDisk to kick off a global effort to standardize High Bandwidth Flash (HBF), a new memory tier aimed at fast-growing AI inference workloads.

The two companies held the launch event at SanDisk’s headquarters in Milpitas, California, and formed a dedicated workstream under the Open Compute Project (OCP) framework to begin formal specification work.

The move comes as the AI market shifts from large-scale model training toward real-world inference services, where many users access AI systems at the same time. Industry players say traditional storage setups are struggling to balance throughput, capacity and power use in this phase.

HBF is designed to sit between high-bandwidth memory (HBM) and solid-state drives, helping bridge the gap between performance and capacity. In a typical AI system, HBM handles the highest bandwidth tasks, while HBF is expected to provide a larger and more efficient memory layer.

Supporters say the approach can improve system scalability and help lower total cost of ownership. Market watchers expect demand for HBF-based memory solutions to expand toward the end of the decade as AI inference deployments grow.

SK hynix and SanDisk said they plan to combine their experience in HBM and NAND flash design, packaging and mass production to speed up both standardization and commercialization of HBF.

Ahn Hyun, chief development officer at SK hynix, said competition in AI infrastructure is moving beyond single-chip performance toward full system optimization, and that open collaboration on HBF will help build a more efficient memory architecture for the AI era.