SK Hynix announced on August 4 that it has unveiled the first standard specification for its next-generation storage technology, HBF, in collaboration with SanDisk.
At the 'FMS 2026' event held at the Santa Clara Convention Center in California, SK Hynix introduced HBF as a key technology to alleviate memory bottlenecks in the AI era.
HBF represents a new memory tier positioned between high-bandwidth memory (HBM) and solid-state drives (SSD). It maintains fast data processing speeds similar to HBM while significantly increasing capacity based on NAND flash, making it a crucial technology to address the slowdown caused by the surge in AI data.
This announcement marks the first achievement since the consortium's launch six months ago in February. The disclosed specification allows for NAND chips to be stacked in 8-layer and 16-layer configurations, achieving a maximum capacity of 512 gigabytes (GB). The bandwidth can reach up to 3.0 terabytes per second (TB/s).
By adopting the next-generation connection standard UCIe, HBF is designed to integrate flexibly with various semiconductors, including graphics processing units (GPUs) and central processing units (CPUs).
The specification was revealed through the Open Compute Project (OCP), the world's largest open data center collaboration, and is expected to serve as an industry-wide standard. Currently, companies like Google and Tenstorrent are participating in the consortium.
At FMS 2026, SK Hynix will also showcase a range of next-generation technologies. Kim Cheon-seong, head of solution development, and Kang Wook-seong, in charge of next-generation product planning, will jointly deliver a keynote address on 'tiered memory' for optimizing AI infrastructure.
Notably, SK Hynix will present the world’s first 10th generation (V10) 375-layer 4D NAND wafer and products that improve power efficiency by 2.5 times. The company plans to begin mass production of enterprise SSDs (eSSD) based on this NAND technology early next year, aiming to strengthen its leadership in the AI memory market.
Kim Cheon-seong emphasized, "As the use of AI rapidly expands, there is a need to redesign the entire data processing structure. Through HBF, we will extend the boundaries between memory and storage and propose a new architecture that enhances overall system efficiency."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.