SK hynix and SanDisk Release First High-Bandwidth Flash Standard for AI Era

SK hynix and SanDisk have jointly introduced the world’s first High Bandwidth Flash (HBF) technical standard, marking a major milestone in the evolution of AI infrastructure. Unveiled at the Future of Memory and Storage (FMS) 2026 conference, the new specification is designed to create a new category of memory that sits between ultra-fast High Bandwidth Memory (HBM) and traditional solid-state drives (SSDs). The companies said HBF is intended to solve one of the biggest challenges facing artificial intelligence today: moving massive amounts of data quickly enough to keep powerful AI processors operating at full capacity.

Unlike conventional flash storage, High Bandwidth Flash is optimized for AI inference workloads, where enormous AI models constantly access large datasets. The open specification supports capacities of up to 512GB with bandwidth ranging from approximately 0.4 terabytes per second to 3 terabytes per second, delivering dramatically faster data access than today’s enterprise storage solutions. The standard also supports UCIe (Universal Chiplet Interconnect Express), allowing HBF to be integrated more efficiently with next-generation AI processors and accelerator chips.

The project has quickly gained support from major players across the AI ecosystem. Companies including Google DeepMind and Tenstorrent have joined the initiative, helping shape the standard through the Open Compute Project (OCP). SK hynix and SanDisk believe that establishing an open industry standard will encourage wider adoption, improve compatibility between hardware vendors, and accelerate innovation as demand for AI infrastructure continues to grow worldwide.

The introduction of HBF comes as AI developers face what experts describe as the “memory wall”—a growing bottleneck where processors become so powerful that memory and storage systems struggle to deliver data quickly enough. While High Bandwidth Memory has significantly improved AI performance, it remains expensive and limited in capacity. HBF is designed to complement HBM by providing a larger, more cost-effective layer of high-speed flash storage, enabling AI systems to process larger models while reducing latency and improving overall efficiency.

Industry analysts believe High Bandwidth Flash could become one of the most important technologies for the next generation of AI computing. As cloud providers, hyperscalers, and enterprise customers build increasingly sophisticated AI systems, the need for faster and more efficient memory architectures will continue to rise. By introducing the first open HBF standard, SK hynix and SanDisk are positioning themselves at the forefront of this emerging market, potentially reshaping how future AI servers store and access data while opening the door to a new era of high-performance memory solutions.