SK海力士与闪迪在FMS 2026上发布首个HBF标准
SK海力士于8月4日宣布，与闪迪共同在FMS 2026上发布了高带宽闪存（HBF）的首个标准规范，该技术基于NAND闪存，面向下一代存储应用。
原文深度解读
SK海力士于8月4日宣布，与闪迪共同在FMS 2026上发布了高带宽闪存（HBF）的首个标准规范。HBF是一种基于NAND闪存的新一代存储技术，旨在比HBM存储更多数据，同时比SSD处理数据更快。该规范定义了最大512GB的容量（基于8层和16层堆叠）和0.4TB/s至3.0TB/s的带宽，并采用UCIe作为连接标准。规范通过OCP公开，谷歌和Tenstorrent已参与HBF联盟。SK海力士还计划展示第10代375层4D NAND晶圆，并于明年初量产企业级SSD。
- SK海力士于8月4日宣布，与闪迪在FMS 2026上发布了HBF的首个标准规范。
- HBF基于NAND闪存，通过垂直堆叠NAND闪存来增加容量和数据传输速度，旨在减少AI推理带来的内存瓶颈。
- 规范定义了最大512GB的容量，基于8层和16层两种堆叠配置。
- 带宽分为三个等级，支持从0.4TB/s到3.0TB/s。
- 连接方式采用行业标准UCIe，以便灵活连接GPU和CPU等不同处理器。
- 规范通过OCP公开，谷歌和Tenstorrent已参与HBF联盟。
供应链影响
- HBF标准的发布可能推动NAND闪存堆叠技术的进一步应用，影响相关设备和材料供应链。
- SK海力士计划明年初量产基于V10 375层技术的企业级SSD，可能影响企业存储市场的供应格局。
- HBF联盟的扩展可能吸引更多参与者，从而影响存储生态系统的合作模式。
- UCIe标准的采用可能促进不同处理器与HBF的互操作性，进而影响AI硬件供应链的集成方式。
- HBF的容量和带宽规格可能取决于实际应用需求，对存储系统设计产生潜在影响。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
SK海力士在FMS 2026展会的鸟瞰图
SK海力士于8月4日宣布，与闪迪共同发布了高带宽闪存（HBF）的首个标准规范，这是一种基于NAND闪存的新一代存储技术。
SK海力士于8月4日表示，在参加于美国加利福尼亚州圣克拉拉会议中心举行的全球最大内存会议“FMS（未来内存与存储）2026”期间，发布了这些规范。会议于当地时间8月4日至6日举行。
HBF是一种新型内存技术，旨在比高带宽内存（HBM）存储更多数据，同时比固态硬盘（SSD）处理数据更快。与堆叠多层DRAM的HBM类似，HBF垂直堆叠NAND闪存以增加容量和数据传输速度。它被认为是减少因人工智能（AI）推理导致数据量激增时内存瓶颈的新一代技术。
这些规范是自今年2月SK海力士成立联盟以来，约六个月后发布的首个成果。此前，SK海力士于去年8月与闪迪合作开展HBF标准化工作。
容量定义为最大512GB，基于8层和16层两种堆叠配置。带宽分为三个等级，支持从每秒0.4太字节（TB）到3.0TB。
作为连接HBF和处理器的连接方式，采用了行业标准UCIe。这为灵活连接和使用HBF与不同类型的处理器（如图形处理单元（GPU）和中央处理单元（CPU））奠定了基础。
这些规范通过开放计算项目（OCP）公开，OCP是全球最大的开放数据中心技术协作组织，以便整个行业都能利用这些规范。
谷歌和AI半导体公司Tenstorrent目前正在参与HBF联盟。SK海力士计划扩大参与公司数量，以提高技术成熟度和市场接受度，加速HBF生态系统的建立。
SK海力士还将在本次活动中首次展示第10代（V10）375层4D NAND晶圆和正在开发的产品。明年初，它计划开始量产应用该技术的高性能、高容量企业级SSD（eSSD）。
“随着AI应用的快速普及，现在需要重新设计整个数据处理结构，”SK海力士解决方案开发负责人Cheonseong Kim表示。“通过HBF，我们将扩展内存和存储之间的边界，并有助于实现提高整体系统效率的新架构。”
SK海力士于8月4日宣布，与闪迪共同通过OCP公开了HBF（一种新一代存储技术）的首个标准规范。
Bird's-eye view of SK Hynix exhibition at FMS 2026
SK Hynix announced on August 4 that, together with SanDisk, it unveiled the first standard specifications for High Bandwidth Flash (HBF), a next-generation storage technology based on NAND flash memory.
SK Hynix stated on August 4 that it presented the specifications while participating in 'FMS (Future of Memory and Storage) 2026,' the world's largest memory conference held from August 4 to 6 (local time) at the Santa Clara Convention Center in California, U.S.
HBF is a new memory technology developed to hold more data than High Bandwidth Memory (HBM) while processing data faster than Solid State Drives (SSDs). Like HBM, which stacks multiple layers of DRAM, HBF stacks NAND flash vertically to increase capacity and data transfer speed. It is considered a next-generation technology to reduce memory bottlenecks in situations where data volume pours out due to artificial intelligence (AI) inference.
These specifications are the first result released about six months after SK Hynix launched a consortium in February this year, following its collaboration on HBF standardization with SanDisk in August last year.
The capacity was defined up to a maximum of 512GB based on two stack configurations that stack NAND dies in 8 layers and 16 layers. The bandwidth was divided into three tiers to support from 0.4 terabytes (TB) to 3.0TB per second.
As the connection method linking HBF and processors, the industry standard UCIe was adopted. This established a foundation to flexibly connect and use HBF with different types of processors, such as Graphics Processing Units (GPUs) and Central Processing Units (CPUs).
The specifications were disclosed through the Open Compute Project (OCP), the world's largest open data center technology collaboration organization, so that the entire industry can utilize them.
Google and AI semiconductor company Tenstorrent are currently participating in the HBF Consortium. SK Hynix plans to expand participating companies to increase technology maturity and market acceptance, accelerating the establishment of the HBF ecosystem.
SK Hynix will also reveal for the first time at this event the 10th-generation (V10) 375-layer 4D NAND wafer and products under development. Early next year, it plans to begin mass production of high-performance, high-capacity enterprise SSDs (eSSDs) applying this technology.
“As AI adoption spreads rapidly, it is a point where redesigning the entire data processing structure is required,” said Cheonseong Kim, head of solution development at SK Hynix. “Through HBF, we will expand the boundary between memory and storage and contribute to implementing a new architecture that improves overall system efficiency.”
SK Hynix announced on August 4 that, together with SanDisk, it disclosed the first standard specifications for HBF, a next-generation storage technology, through OCP.