At the Future of Memory and Storage conference in Santa Clara, California, Samsung Electronics presented a series of next-generation 3D memory concept models and storage solutions to support artificial intelligence workloads.
During an opening keynote titled “Driving the Wave of AI Revolution: 3D Innovations in Memory & Storage Architecture,” Samsung executives Jin-Yub Lee and Kyungryun Kim outlined a vision to maximize AI system performance and power efficiency. The presentation focused on how advanced three-dimensional structures can assist the industry in overcoming existing hardware bottlenecks while meeting the escalating requirements of high-performance computing.
A central announcement was zHBM, an innovative memory architecture created specifically for advanced AI computing. Rather than placing high-bandwidth memory alongside a processor as seen in traditional setups, zHBM stacks memory directly above AI accelerators. By reducing the physical path that data must travel, the architecture is expected to offer eight times the performance of HBM5, ten times the memory density, three times the energy efficiency, and more than half the thermal resistance of conventional designs. It also gives customers the ability to integrate customized interlayer IP between the processor and memory.
Alongside zHBM, Samsung introduced zNAND-O, a specialized NAND solution currently in development for edge AI applications. Built on V-NAND technology, zNAND-O aims to provide low latency, superior space efficiency, and enhanced input-output capabilities for real-time edge processing.
The company also highlighted a major manufacturing milestone by revealing its V10 BV-NAND architecture. Utilizing a novel wafer bonding technology to stack memory cells, this solution features over 400 layers and increases memory density by roughly 58 percent compared to the prior generation. It delivers faster read, write, and input-output performance across enterprise environments.
To support the growing requirements of AI data centers, Samsung showcased several key hardware offerings. These included sampling progress for HBM4E and HBM5, alongside LPDDR5X-PIM, which integrates processing capabilities directly into memory to streamline data movement. The company also presented high-capacity enterprise storage solutions such as the PM1763 and BM1773.
As an integrated device manufacturer covering memory, foundry, and packaging, Samsung emphasized its turn-key model. This end-to-end approach allows customers to streamline development cycles and rapidly bring tailored AI semiconductor solutions to market.