SK hynix and Sandisk have agreed to jointly pursue global standardization of High Bandwidth Flash, a next-generation memory technology designed to improve artificial intelligence inference systems, the companies said Thursday.
The two chipmakers launched the HBF Spec Standardization Consortium at Sandisk’s headquarters in California, announcing plans to form a dedicated workstream under the Open Compute Project, the world’s largest open data center technology initiative.
The workstream will focus on developing industry specifications for HBF, a technology positioned between high-bandwidth memory and solid-state drives. While HBM delivers ultrafast performance and SSDs provide large storage capacity, HBF aims to combine elements of both to better meet the evolving demands of AI infrastructure.
A workstream is an industry collaboration framework in which companies jointly establish technical standards and specifications around emerging technologies.
SK hynix said HBF is intended to address growing requirements for memory capacity and power efficiency as AI workloads shift from model training to inference — the stage in which AI systems process user requests in real time.
Existing memory architectures often struggle to balance speed, scalability and energy efficiency in inference environments, the company noted.
By bridging the gap between HBM’s high-speed performance and SSDs’ storage density, HBF could improve system scalability and reduce the total cost of ownership for AI data centers.
“By making HBF an industry standard together with Sandisk, we aim to lay the foundation for the AI ecosystem to grow collectively,” SK hynix said in a statement.
Ahn Hyun, president and chief development officer at SK hynix, emphasized the need for broader collaboration in the AI era.
“The key to AI infrastructure is not simply competition over individual component performance, but optimization across the entire ecosystem,” Ahn said. “Through HBF standardization, we will establish a cooperative framework and introduce a memory architecture optimized for the AI era.”
The companies said they will pursue both standardization and commercialization of HBF, with industry demand for advanced memory solutions such as HBF expected to expand toward 2030.
herim@heraldcorp.com
