SEOUL, South Korea / RankWire.AI / – Samsung Electronics has presented new memory and storage technology strategies designed to improve the performance of artificial intelligence infrastructure. The company revealed the plans at the 2026 OCP Global Summit in San Jose, California, held from October 12 to 15. Its roadmap focuses on high-performance memory, memory expansion through Compute Express Link and next-generation solid-state drives. Samsung will also stage a joint CXL demonstration with AMD. The plans address capacity constraints that limit current AI servers.
The company described its strategy through three goals: speed, memory scalability and data persistence. For speed, Samsung presented customized high-bandwidth memory alongside HBM5, the eighth generation of the technology. For scalability, it promoted CXL as a way to add capacity beyond the memory installed inside each server. For persistence, Samsung introduced next-generation SSD-based data storage and management technology. The company also presented storage software aimed at improving data processing efficiency in AI systems.
Growing use of AI agents is driving the need for more memory during inference, according to Samsung. Larger models create more data while generating responses, and that data must stay accessible to processors. Conventional systems limit how much memory each processor can reach. Samsung presented CXL technology as a practical way to overcome that limit. It also showed key-value cache optimization, which lets AI models reuse data they have already processed rather than repeating the same work.
AMD partnership extends to CXL memory expansion
Samsung and AMD will jointly demonstrate a CXL-based solution that expands memory capacity in AI systems. The technology lets processors share memory resources instead of relying only on local server memory. Samsung says engineers have validated the solution in real system environments. The demonstration builds on a memorandum of understanding the companies signed in March 2026 for next-generation AI memory and computing. AMD CEO Lisa Su visited Korea on October 7 and met Jeon Young-hyun, Vice Chairman of Samsung’s Device Solutions division.
Memory pooling formed another part of Samsung’s showcase. The company demonstrated CMM-D, a CXL-based DRAM module, connecting several memory units into one shared pool. Systems can draw capacity from that pool when workloads require it. Samsung also displayed processing-in-memory technology, which handles some calculations inside memory chips to reduce data movement and save energy. Its exhibits included 2-nanometer gate-all-around wafers, Groq LPU wafers and Nvidia’s Groq 3 LPX system.
Early co-design shapes Samsung’s AI memory roadmap
Shim Ho-jun, Corporate Vice President and Head of Samsung’s Application Engineering Team, outlined the plans in an executive session. He said larger AI models and growing inference demand make greater memory capacity and bandwidth increasingly important for system performance. Shim also said Samsung is widening co-design and technical validation work with global partners from the earliest development stages. The company aims to match its memory and storage products more closely with customer system designs.
Samsung MAX Lab gives customers a place to evaluate memory and storage under real-world AI workloads. The lab also offers co-design tailored to specific system configurations. Customers can test pre-commercial technologies, including PIM and next-generation SSDs, before they reach the market. Samsung took part in the summit as a Diamond Sponsor. The event follows its 2026 appearances at Nvidia GTC, COMPUTEX and the Future of Memory and Storage conference, as the company expands its AI memory presence.