[Edaily Reporter SOYEON KIM ] SamsungElectronics has partnered with AMD to unveil a Compute Express Link (CXL)-based solution—a next-generation technology designed to expand memory capacity in artificial intelligence (AI) data centers. The company also announced the results of its collaborations with global AI firms, including NVIDIA. SamsungElectronics’ strategy is to collaborate with global clients from the early stages of product development to provide customized memory solutions and strengthen its leadership in the AI infrastructure market.

On the 12th (local time), at the “OCP Global Summit 2026” that opened in San Jose, California, SamsungElectronics unveiled its strategy for next-generation memory and storage technologies designed to enhance AI infrastructure performance.

At the event, the company highlighted the results of its technological collaboration with AMD by jointly demonstrating a CXL-based memory expansion solution. It also showcased NVIDIA Grok’s Language Processing Unit (LPU) wafer and NVIDIA’s “Grok 3 LPX” AI inference system, emphasizing its collaboration with global AI companies.

As AI technology advances, the volume of data to be processed during inference is surging, and the memory capacity required to support this is becoming increasingly critical. SamsungElectronics proposed a solution to address the memory capacity limitations of existing systems by leveraging CXL technology. In particular, the company showcased memory pooling technology using CMM-D—a CXL-based memory module—that enables multiple systems to efficiently share memory resources and flexibly secure the memory capacity required for AI environments.

The company is focusing not only on enhancing the performance of individual memory products but also on optimizing the entire AI system, including the central processing unit (CPU), graphics processing unit (GPU), memory, and storage. To this end, its strategy involves conducting joint design and technical validation with global AI customers from the early stages of product development to provide solutions optimized for each customer’s specific system environment.

The joint demonstration with AMD is also part of this collaboration strategy. SamsungElectronics is validating CXL-based memory expansion technology in real-world system environments, thereby solidifying its collaboration to improve AI infrastructure performance. Lisa Su, AMD’s CEO, who recently visited South Korea, stated that the company is continuously exploring opportunities for collaboration with SamsungElectronics in the memory and foundry sectors. Previously, the two companies signed a memorandum of understanding (MOU) in March to expand their collaboration on next-generation AI memory and computing technologies.

SamsungElectronics is also strengthening its collaboration to build an open AI ecosystem. Through the “Samsung MAX Lab,” the company introduced an open collaboration strategy that involves jointly designing AI systems with global customers and partners and validating technologies in real-world operating environments. The Samsung MAX Lab is a research facility that enables customers to evaluate memory and storage technologies in actual AI system environments and incorporate system requirements from the early stages of development.

At this facility, customers can validate pre-commercial technologies—such as Processing-in-Memory (PIM), CMM-D, and next-generation solid-state drives (SSDs)—and participate in the development of products tailored to their systems.

During the executive session that day, Shim Ho-jun, Senior Vice President of SamsungElectronics’ AE Team, presented three technological directions for enhancing AI infrastructure performance. These include: △ high-performance memory, such as High Bandwidth Memory (HBM) 5 and customized HBM; △ CXL-based memory expansion technology; and △ data storage and management technology utilizing next-generation SSDs. He emphasized that the competitiveness of AI infrastructure depends not only on high-performance memory such as HBM but also on memory capacity expansion and the efficiency of data storage and management.