Samsung Electronics has outlined their vision for the future of AI infrastructures with the unveiling of several next-generation memory solutions at the Future of Memory and Storage (FMS) 2026 in Santa Clara, California. They have presented novel memory architectures and storage solutions to cater to the growing needs of artificial intelligence (AI), high-performance computing (HPC), and hyperscale data centers.
This is in light of Samsung’s attempt to address the growing need for better AI system performance, improved memory bandwidth, storage capacity, and power efficiency as companies continue to invest in generative AI and large AI models.
zHBM: Next-Generation Memory Architecture for AI Workloads
Samsung introduced zHBM (Vertical High Bandwidth Memory), an innovative memory architecture for AI workloads, as one of the key announcements at FMS 2026.
As compared to traditional High Bandwidth Memory (HBM) chips placed next to AI processors, zHBM chips are located above the accelerators through a 3D packaging technology. Through reducing the data transfer path, Samsung hopes to boost the bandwidth, cut down the power usage and enhance thermal performance of AI workloads.
zNAND-O Is Suitable for Edge AI in Real Time
Samsung has also unveiled another product called zNAND-O, an architectural concept suitable for edge AI products that need low-latency and fast storage.
Samsung developed zNAND-O to combine the speed of NAND flash with efficient I/O operations, enabling real-time AI inference for intelligent devices, robotics, autonomous systems, and industrial applications.
V10 BV-NAND Is Over 400 Layers Technology
One of the most notable products launched by Samsung is V10 BV-NAND which is the new generation of NAND flash memory featuring more than 400 memory layers.
This technology is based on the wafer bonding technology and helps in increasing storage density, as well as increasing read/write and I/O operations speeds. According to Samsung, this technology increases memory density by about 58% compared to its predecessor.
Expanded 3D Memory Roadmap
Aside from their conceptual technologies, Samsung has shown an expanded range of AI memory devices, which include:
- HBM4E
- HBM5
- LPDDR5X-PIM (Processing-In-Memory)
- PM1763 enterprise SSD
- BM1773 enterprise storage solutions
Such products are aimed to provide improved performance and capabilities for AI servers, cloud computing, and upcoming data center infrastructure through increased bandwidth, efficiency, and data processing speeds.
Satisfying Increasing AI Infrastructure Requirements
AI loads are developing not only from training tasks but also to inference ones, increasing the pressure on memory performance and storage capacity requirements. The company’s current roadmap is reflecting the trend of vertically integrating memory architecture, reducing latency and improving overall system efficiency.
The company has also shown around 30 memory and storage technologies at its AI cloud server-themed exhibition during FMS 2026.
Intel and Lens Technology Partner to Advance Semiconductor Packaging for AI
Industry Outlook
With the recent moves of Samsung, it can be seen that there is an increasingly tough battle among manufacturers of semiconductor memory technology that are competing to develop new AI infrastructures. Through its zHBM and zNAND-O technologies, as well as the recently released V10 BV-NAND technology, Samsung strives to consolidate its dominance in the ever-growing AI semiconductor industry.
With enterprises increasingly focusing on generative AI, HPC, and hyper-scale cloud services, it is anticipated that advancements in memory technology will prove crucial for achieving better performance, lower power usage, and more sophisticated AI applications.
Reference: Samsung Newsroom
