SK Hynix, Sandisk Target AI With New 10x Capacity Memory Standard

SK Hynix and Sandisk Initiate HBF Standardization on Feb 25

SK Hynix and Sandisk formally launched a partnership on February 25 to establish a global standard for High-Bandwidth Flash (HBF) memory. At an event hosted at Sandisk's headquarters in California, the two companies announced the formation of a dedicated working group within the Open Compute Project (OCP) framework. This move signals a formal push to create an industry-wide specification for HBF, positioning it as a critical component for future AI infrastructure. The collaboration leverages both firms' extensive experience in NAND flash and memory packaging to drive a new standard.

HBF Designed to Offer 10x HBM Capacity for AI Inference

High-Bandwidth Flash is architected to fill a crucial gap between the ultra-fast, but capacity-limited, High-Bandwidth Memory (HBM) and the slower, high-capacity solid-state drives (SSDs). By vertically stacking NAND flash, HBF is designed to deliver roughly 10 times the storage capacity of HBM while maintaining high bandwidth. This positions it as an ideal solution for AI inference tasks, where applications serving millions of users require both rapid data access and massive storage capacity. By creating this new memory tier, SK Hynix expects to improve the scalability of AI systems and lower the total cost of ownership (TCO) for data centers.

Market Adoption Expected by 2028, May Outpace HBM by 2038

The commercialization timeline for HBF is accelerating, with industry experts projecting its integration into products from Nvidia, AMD, and Google between late 2027 and early 2028. The demand for HBF-based solutions is forecast to enter a rapid growth phase around 2030 as the AI industry shifts from model training to widespread inference deployment. Projections suggest the HBF market could eventually surpass HBM in scale by 2038. This positions companies that can provide a full stack of memory solutions, including both HBM and HBF, with a significant strategic advantage in the evolving AI hardware landscape.