AMD selects Samsung as key HBM4 supplier for next AI GPUs

15:20

By: Dakir Madiha

AMD selects Samsung as key HBM4 supplier for next AI GPUs

AMD has named Samsung Electronics as its primary supplier of next generation HBM4 memory for its upcoming AI accelerators, deepening ties as competition intensifies in high performance semiconductor infrastructure.

Chief executive Lisa Su traveled to South Korea for the first time since taking the role in 2014, signing a memorandum of understanding with Samsung at its semiconductor facility in Pyeongtaek. Under the agreement, Samsung will supply HBM4 memory for AMD’s Instinct MI455X GPU, scheduled for release in the second half of 2026.

The deal also includes advanced DDR5 memory for AMD’s sixth generation EPYC server processors, code named Venice, as well as support for the company’s Helios rack scale data center platform. Samsung already provides HBM3E memory for AMD’s current MI350X and MI355 accelerators, but the new agreement expands the partnership as demand for AI memory surges globally.

The two companies also discussed potential collaboration in chip manufacturing, which could see Samsung produce future AMD processors. Such a move would extend their relationship beyond memory supply into contract chip fabrication, placing Samsung in closer competition with other foundries.

During the visit, Su met Samsung semiconductor division head Jun Young hyun and foundry chief Han Jin man, and was scheduled to meet chairman Jay Y Lee later in the day.

Earlier, Su signed a separate agreement with South Korean technology firm Naver at its headquarters in Seongnam. The partnership focuses on building next generation AI infrastructure, including high performance GPU computing systems optimized for Naver’s large language model HyperCLOVA X. The initiative aims to diversify Naver’s hardware base and reduce reliance on Nvidia.

Su said Naver’s AI capabilities and cloud platform make it a strong partner for deploying AMD’s next generation GPU technology. Naver chief executive Choi Soo yeon described the deal as a step toward strengthening infrastructure competitiveness and expanding technology options.

Su’s visit coincided with Nvidia’s annual GTC developer conference, a timing analysts viewed as a strategic effort by AMD to highlight its partnerships while attention remained on its main rival. She is also expected to meet South Korean government officials and visit AI startup Upstage to discuss data center solutions.

The trip underscores South Korea’s growing role in the global race to build AI infrastructure, driven by its leadership in semiconductor manufacturing and advanced computing technologies.