SEOUL, South Korea / RankWire.AI / – Samsung Electronics has introduced innovative approaches to memory and storage technology aimed at enhancing artificial intelligence infrastructure capabilities. These plans were announced at the 2026 OCP Global Summit in San Jose, California, which took place from October 12 to 15. The company’s roadmap emphasizes high-performance memory, expanding memory via Compute Express Link, and next-generation solid-state drives. Additionally, Samsung will collaborate with AMD on a joint CXL demonstration. These initiatives seek to address existing capacity limitations faced by current AI servers.
Samsung articulated its strategy around three core objectives: increasing speed, enabling memory scalability, and ensuring data persistence. To improve speed, the company showcased customized high-bandwidth memory alongside HBM5, the eighth generation of the technology. For scalability, CXL technology was promoted as a solution to extend capacity beyond the memory installed within each server. Regarding data persistence, Samsung introduced next-generation SSD-based storage and management innovations. The company also unveiled storage software designed to enhance data processing efficiency in AI systems.
The rising adoption of AI agents is increasing the demand for more memory during inference, according to Samsung. Larger models produce more data during response generation, which must remain accessible to processors. Traditional systems impose limits on the memory capacity per processor. Samsung presented CXL technology as a practical method to surpass these restrictions. The company also highlighted key-value cache optimization, allowing AI models to reuse previously processed data instead of repeating computations.
Partnership with AMD Expands to CXL Memory Expansion Solutions
Samsung and AMD will jointly showcase a CXL-based solution designed to increase memory capacity in AI systems. This technology enables processors to share memory resources rather than relying solely on local server memory. Samsung reports that engineers have validated this solution in real system environments. The demonstration builds upon a memorandum of understanding signed by both companies in March 2026 for next-generation AI memory and computing. On October 7, AMD CEO Lisa Su visited Korea and met with Jeon Young-hyun, Vice Chairman of Samsung’s Device Solutions division.
Memory pooling was another highlight of Samsung’s presentation. The company demonstrated CMM-D, a CXL-based DRAM module that links multiple memory units into a unified pool. This setup allows systems to draw capacity from the shared pool as workload demands increase. Samsung also exhibited processing-in-memory technology, which performs some calculations within memory chips to reduce data movement and energy consumption. Exhibits included 2-nanometer gate-all-around wafers, Groq LPU wafers, and Nvidia’s Groq 3 LPX system.
Early Co-Design Efforts Drive Samsung’s AI Memory Roadmap
In an executive session, Shim Ho-jun, Vice President and Head of Samsung’s Application Engineering Team, outlined the company’s plans. He emphasized that the growth of larger AI models and increased inference demand make greater memory capacity and bandwidth vital for system performance. Shim also mentioned that Samsung is expanding co-design and technical validation collaborations with global partners from the earliest development stages. The goal is to align memory and storage products more closely with customer system designs.
Samsung MAX Lab provides customers with a facility to evaluate memory and storage solutions under real-world AI workloads. The lab offers co-design services tailored to specific system configurations, allowing clients to test pre-commercial technologies, including PIM and next-generation SSDs, before market release. Samsung participated in the summit as a Diamond Sponsor. This event follows its appearances in 2026 at Nvidia GTC, COMPUTEX, and the Future of Memory and Storage conference, as the company continues to expand its presence in AI memory innovation.