Samsung and AMD are expanding collaboration to supply HBM4 memory for AMD Instinct MI455X AI accelerators and DDR5 memory for the sixth-generation AMD EPYC processors, with talks of a potential foundry partnership to manufacture next-generation AMD products.

A proposed foundry partnership would have Samsung provide contract manufacturing services, complementing its ongoing role as an HBM supplier for AMD.

The arrangement centers on high-bandwidth memory for data-center AI accelerators, signaling progress on advanced memory technology integration.

Naver’s CEO emphasized strengthening AI infrastructure competitiveness by deploying AMD platforms across Naver Cloud and AI services.

Plans include joint research initiatives and providing AI computing resources to support academic research, broadening impact across various infrastructure environments.

AMD’s AI strategy is gaining momentum with earlier commercial HBM4 shipments to Nvidia and AMD after qualification tests at 11.7 Gbps.

The collaboration aims to enable AI systems that fuse AMD Instinct GPUs, EPYC CPUs, and rack-scale architectures to advance large-scale model training and inference.

AMD and Naver signed an MOU to build a more integrated AI ecosystem, connecting infrastructure environments and improving deployment flexibility.

The discussions occurred during AMD CEO Lisa Su’s first Korea visit since 2014, with meetings with Samsung and Naver executives to advance AI infrastructure partnerships.

Samsung views HBM4 as a key lever to strengthen its AI leadership amid market pressure and ongoing expansion of its HBM supply capabilities.

The AMD-Naver agreement focuses on developing a high-performance GPU computing environment for Naver’s HyperClova X and expanding collaboration across AI infrastructure, data centers, and cloud services.

AMD has been pursuing major AI chip deals, including large-scale supply agreements with Meta and OpenAI, reinforcing its AI compute ambitions.