At FMS 2026, Samsung showcased its latest AI memory roadmap, featuring advanced NAND, HBM and enterprise storage solutions for next-generation data centres.
Samsung Electronics has unveiled a new portfolio of AI-focused memory and storage technologies at the Future of Memory and Storage (FMS) 2026 conference in Santa Clara, California, highlighting innovations designed to support the rapidly growing demands of artificial intelligence, high-performance computing (HPC) and next-generation data centres.
Among the major announcements was V10 Bonding V-NAND (BV-NAND), which Samsung described as the industry’s first NAND flash architecture built using advanced wafer bonding technology. The new memory features more than 400 layers, delivering higher storage density, improved read and write speeds, and better power efficiency than the previous generation. According to the company, the architecture increases memory density by approximately 58% over V9 V-NAND while enhancing overall performance for AI applications.
Samsung also introduced two concept technologies—zHBM and zNAND-O—that outline its vision for future AI memory architectures. The zHBM concept vertically stacks High Bandwidth Memory (HBM) directly above AI accelerators to reduce data transfer distances, enabling higher bandwidth, improved energy efficiency and lower thermal resistance. Samsung said future systems using the technology could deliver significantly higher performance than current HBM5-based designs.
The zNAND-O concept is aimed at edge AI applications, combining low latency, improved input/output performance and space-efficient storage to support real-time AI workloads.
In addition, Samsung showcased its expanding AI memory portfolio, including HBM4E, HBM5, LPDDR5X-PIM, and enterprise storage solutions such as PM1763 and BM1773 for AI data centres. The company noted that it has already started shipping HBM4E samples to customers following the mass production of HBM4 earlier this year.
Samsung also highlighted its end-to-end semiconductor capabilities, spanning memory, foundry and advanced packaging, allowing it to offer integrated manufacturing solutions for AI chips. The company said this approach will help customers accelerate product development while improving system performance, power efficiency and scalability for future AI infrastructure.
