Samsung Electronics unveils CXL-based memory expansion technology developed with AMD

Samsung Electronics has unveiled the results of its Compute Express Link (CXL)-based collaboration with AMD to expand memory capacity in AI systems, and presented its open-collaboration strategy with global customers and partners to improve AI infrastructure performance.

Samsung presented its next-generation memory and storage technology strategy and partnership achievements at the 2026 OCP Global Summit, which opened Monday (local time) in San Jose, California.

A highlight was a joint demonstration with AMD of a CXL-based memory expansion solution, showcasing the two companies' technical collaboration to flexibly scale memory capacity in AI systems.

As the spread of AI agents drives growing data volumes and memory demands during inference, Samsung proposed using CXL technology to overcome the memory capacity limits of existing systems.

Samsung demonstrated CMM-D memory pooling technology that lets multiple systems efficiently share memory resources and flexibly secure the capacity needed in AI environments.

CMM-D is a CXL-based DRAM memory module. Memory pooling works by linking multiple CXL memory units into a pool from which systems can draw as much capacity as needed.

Through the joint demonstration with AMD, Samsung said it validated the technology in real system environments and is expanding cooperation to improve AI infrastructure performance.

AMD CEO Lisa Su visited South Korea on Wednesday and met with Samsung Electronics DS Division Vice Chairman Jeon Young-hyun to continue discussions on next-generation AI memory and computing technology cooperation.

Through its Samsung MAX Lab, Samsung evaluates memory and storage technologies in real AI environments and supports co-design tailored to customers' system configurations. Customers can validate pre-commercialization technologies — including processing-in-memory (PIM), CMM-D and next-generation solid-state drives (SSDs) — and incorporate their requirements from the earliest stages of development.

Samsung also presented storage software technology at OCP to improve data processing efficiency in AI systems, underscoring its contribution to building an open AI ecosystem.

Shim Ho-jun, senior vice president of the AE Team in Samsung's DS Division, unveiled a range of technologies in an executive session — from high-performance memory including eighth-generation HBM (HBM5) and custom HBM, to CXL-based memory expansion solutions and next-generation SSD-based data storage and management technology.

He also outlined a three-pronged technology strategy for improving AI infrastructure performance, covering speed, memory scalability and data persistence.