Rumor
Nvidia lowers HBM4 specifications amid supply constraints
Tuesday, March 3, 2026 at 07:05 AM
Nvidia has reportedly lowered the required technical specifications for HBM4 memory due to ongoing supply chain challenges within the HBM4 segment.
Context
Nvidia is reportedly relaxing technical specifications for its upcoming sixth-generation high-bandwidth memory, HBM4, to address persistent yield and production constraints. While the company initially targeted a peak performance benchmark of 11.7Gbps, it is now expected to adopt a "dual-bin" procurement strategy. This approach includes sourcing slightly lower-tier 10.6Gbps modules alongside top-tier chips to ensure supply stability and prevent manufacturing bottlenecks for its next-generation Rubin AI platform.
The strategic shift prioritizes volume over extreme performance as major suppliers like SK Hynix, Samsung, and Micron struggle to reach high yields at the most aggressive specifications. Samsung has reported yield rates of approximately 60% for relevant nodes, while SK Hynix is projected to maintain a dominant 70% share of the HBM4 market. By lowering the entry barrier, Nvidia aims to keep its 2026 rollout for the Rubin architecture on track, safeguarding its lead in the AI accelerator market despite tightening global memory supply.
Related Companies
Nvidia
NVDA