Key Points
- NetworkWorld reports Samsung unveiled three wafer-bonded memory technologies at the FMS conference, led by BV-NAND with 400-plus layers and a 58 percent density gain over V9
- The zHBM concept stacks HBM on AI accelerators for up to ten times HBM5 density but remains a research project with no deployment timeline
- zNAND-O is an early 3D NAND extension in four- and eight-layer development, extending vertical stacking beyond current NAND
What is changing
Samsung announced three next-generation memory designs at the Future of Memory and Storage conference, all built on wafer bonding manufacturing. The lead product, BV-NAND, pushes flash past 400 layers and claims a 58 percent density increase over the current V9 generation while improving read, write, and I/O performance with lower power draw, making it better suited for AI servers that depend on high-capacity flash storage.
Two longer-term concepts were also shown. zHBM would stack high-bandwidth memory directly on top of AI accelerators instead of beside them. Samsung estimates the design could ultimately deliver more than ten times the memory density of conventional HBM5 while tripling energy efficiency and cutting thermal resistance by more than half, but stressed it is still a research concept. The other, zNAND-O, is an early-stage 3D NAND extension in four- and eight-layer development.
Why it matters
This matters most to system architects planning AI server clusters who need higher memory density and lower thermal budgets. The BV-NAND gains could translate to more storage in the same rack space, while the zHBM vision signals where HBM packaging may go if bonding yields improve. HBM has become critical for AI compute, but the industry has hit bandwidth and speed limits.
The practical takeaway is that BV-NAND is the only near-term technology here; zHBM and zNAND-O have no shipping timeline. IT buyers should watch for BV-NAND sampling dates but not budget for the conceptual designs yet.
Have you evaluated BV-NAND samples or tracked HBM stacking roadmaps? Share your experience in the comments.