SanDisk’s HBF Halves GPU Requirements for AI Inference by Outperforming HBM Capacity

SanDisk demonstrated at recent industry events that systems equipped with its High Bandwidth Flash (HBF) memory can match the AI inference performance of High Bandwidth Memory (HBM) systems while using half the number of GPUs, citing 10-100x higher capacity per die. SanDisk is advancing HBF, which stacks NAND flash silicon dies, as a higher-capacity, lower-cost, and lower-power alternative to HBM for AI systems, targeting the memory capacity bottleneck that large language model inference places on HBM. The company says HBF's NAND density could reduce the hardware footprint of AI infrastructure.

SanDisk’s HBF Halves GPU Requirements for AI Inference by Outperforming HBM Capacity SanDisk is advancing High Bandwidth Flash HBF memory, which stacks NAND flash silicon dies, as a high-capacity, lower-cost, and lower-power alternative to High Bandwidth Memory HBM for AI systems. AsiaAI Publisher · September 10, 2026 · 2 min read · Source: PC Watch Impress · Issue 92 East Asian Technology Intelligence Japan & China tech news — translated, contextualized, and delivered for Western readers. Free. Unsubscribe anytime. This story ran in Issue 92, alongside three other stories. Semiconductors & Hardware SanDisk is advancing High Bandwidth Flash HBF memory, which stacks NAND flash silicon dies, as a high-capacity, lower-cost, and lower-power alternative to High Bandwidth Memory HBM for AI systems. At recent industry events, SanDisk demonstrated that HBF-equipped systems can achieve comparable AI inference performance to HBM systems using half the number of GPUs, citing its 10-100x higher capacity per die. The demand for massive, high-speed memory in AI inference systems, particularly for large language models LLMs , is pushing HBM to its limits in terms of capacity. HBF, leveraging the higher density of NAND flash, offers a potential solution for overcoming this capacity bottleneck and reducing the hardware footprint of AI infrastructure.