Sandisk Corporation (Nasdaq: SNDK) and SK hynix held an ‘HBF Standardization Kick-Off’ event at Sandisk Global Headquarters, announcing a joint effort to standardize High Bandwidth Flash (HBF™), the next-generation memory solution designed for the AI inference era. The companies will establish a dedicated workstream under the Open Compute Project (OCP) to advance the HBF™ standardization work.

Recently, the AI industry is shifting from training which focuses on creating Large Language Models (LLMs) to inference, which delivers actual AI services to users. Unlike training, inference runs continuously as models generate answers and predictions across millions of devices, making these workloads memory bandwidth and capacity hungry and increasingly power constrained. As this demand grows, this HBF™ solution addresses these challenges as a new memory category purpose-built for inference at scale. It is being designed to deliver high bandwidth with industry-leading capacity, while offering the persistence and thermal stability needed for large-scale AI deployments.

“We’re not just defining a new industry standard – we are setting the bar for the next era of AI computing,” said Alper Ilkbahar, Chief Technology Officer, Sandisk. “Through our workstream under the Open Compute Project, we’re laying the foundation for the HBF™ solution to become a core building block of next‑generation AI systems.”

By formalizing an HBF™ standard, Sandisk and SK hynix are working to ensure that this new memory category is ready for broad industry adoption and deployment, based on the companies’ experience in NAND design, SoC, and packaging.