Samsung Electronics introduced its latest advancements in memory and storage architecture at the Future of Memory and Storage (FMS) 2026 conference in Santa Clara, California. The company unveiled its V10 BV-NAND, an industry-first architecture utilizing wafer bonding technology to stack more than 400 layers of memory cells. This new NAND solution offers a 58% increase in memory density compared to the previous V9 generation, alongside improvements in read, write, and I/O performance.
Beyond NAND, Samsung presented concept models for zHBM and zNAND-O, which represent the company's vision for next-generation 3D memory structures. The zHBM architecture is designed to vertically stack high-bandwidth memory directly above AI accelerators, rather than placing it alongside the processor. By reducing the physical distance data must travel, Samsung aims to significantly increase bandwidth and power efficiency for large-scale AI training and inference tasks.
Technical projections for the zHBM architecture suggest it could deliver approximately eight times the performance of HBM5. Samsung also anticipates that this design will provide over 10 times the memory density of HBM5, while simultaneously reducing thermal resistance by more than 50% and improving energy efficiency threefold. The architecture is intended to support customer-specific configurations, allowing for the integration of custom intellectual property between the memory and the accelerator.
The company also detailed its broader AI memory roadmap, which includes HBM4E samples and HBM5 models. Samsung noted that it began mass production of HBM4 using 1c DRAM and 4-nanometer base die technologies in February, followed by the shipment of HBM4E samples to customers in May. Additionally, the company showcased LPDDR5X-PIM, which incorporates processing-in-memory technology to perform data operations directly within the memory to lower power consumption.
For edge AI environments, Samsung introduced zNAND-O, a high-performance solution available in four- and eight-layer configurations. This technology is optimized for low-latency, data-intensive applications that require real-time processing. These memory innovations are complemented by new enterprise storage offerings, including the PM1763 and BM1773, which are designed to meet the high-capacity storage demands of modern AI data centers.
Samsung emphasized its role as an integrated device manufacturer, offering a one-stop turnkey strategy that combines memory, foundry, and advanced packaging capabilities. By managing the entire development cycle from design to mass production, the company aims to shorten development timelines and optimize performance for specialized AI semiconductor solutions.