Lee Jae-yong and Lisa Su forge ‘AI alliance’... Samsung secures priority HBM4 supply rights to AMD after Nvidia

[Financial News] On the 18th, Samsung Electronics was selected as the priority supplier of High Bandwidth Memory 4 (HBM4) to be used in AMD’s next-generation MI455X data center AI accelerator. Following its initial HBM4 supply deal with Nvidia last month, Samsung is now seen as shaking up an HBM market that had been led by SK hynix. Lee Jae-yong, Chairman of Samsung Electronics, is expected to use a dinner meeting with Lisa Su, CEO of AMD, who is visiting Korea for the first time in 12 years, to strengthen an AI semiconductor alliance between the two companies, centered on HBM and other AI chips.
■Samsung to supply AI rival chips to both Nvidia and AMDJun Young-hyun, who oversees semiconductors at Samsung Electronics as head of the Device Solutions (DS) Division and vice chairman, signed an MOU with Lisa Su at the Samsung Electronics Pyeongtaek Plant that day. The agreement aims to expand cooperation in AI memory and computing technologies.
At the heart of the deal is that Samsung Electronics has been designated the priority supplier of sixth-generation HBM4 for AMD’s AI accelerators. Samsung will begin full-scale adoption of its industry-leading HBM4 in AMD’s next-generation AI accelerator, the AMD Instinct MI455X GPU. Previously, Samsung had been the exclusive supplier of HBM3E high-bandwidth memory for AMD’s MI350X and MI355X AI accelerators. By securing priority supply status for HBM4 in the next-generation accelerator as well, Samsung has further solidified its position as a key supplier to AMD.
Under the agreement, Samsung Electronics and AMD will also collaborate on high-performance Double Data Rate 5 (DDR5) memory solutions. The goal is to maximize the performance of Helios, AMD’s rack-scale AI data center platform, and its sixth-generation AMD Epyc server CPUs.
“We are very pleased to combine Samsung’s leadership in advanced memory technologies with AMD’s Instinct GPUs, Epyc CPUs and rack-scale platforms,” Lisa Su said. Lee Jae-yong is reported to have held an in-depth dinner meeting with Su, a core partner in the AI semiconductor market, to discuss ways to deepen cooperation between the two companies. This marks the first official meeting between Lee and Su.
The two sides are also believed to have discussed cooperation in foundry, or contract chip manufacturing. Leveraging its strength as a comprehensive semiconductor company spanning memory, foundry and packaging, Samsung Electronics plans to continue expanding partnerships with global big tech firms. On the 19th, Lisa Su is also scheduled to meet Roh Tae-moon, head of the Device eXperience (DX) Division at Samsung Electronics, as part of broader cooperation with the company.
■Relay meetings with Korean IT players like Naver and UpstageLisa Su’s visit to Korea is also widening the scope of collaboration between AMD and the domestic AI industry. On the morning of the 18th, she visited NAVER 1784, Naver’s headquarters building in Seongnam City, and met with Choi Soo-yeon, CEO of Naver. Naver and AMD signed an MOU for “expanding the AI ecosystem and cooperating on next-generation infrastructure.” The two companies will strengthen technical collaboration to build a GPU computing environment optimized for Naver’s HyperCLOVA X large language model (LLM), and jointly advance infrastructure technologies that can operate AI models stably.
AMD is preparing to launch its AMD Instinct MI450 series GPU, and is expected to supply high-performance, cost-efficient AI chips to Naver’s AI data center. Industry observers say that through this agreement, AMD will deepen cooperation with Naver, one of Korea’s major GPU customers, and thereby counter the Nvidia-dominated market structure. Naver, in turn, is expected to reduce its dependence on Nvidia for AI infrastructure development and significantly cut costs.
On the 19th, Lisa Su is also scheduled to meet Ha Jung-woo, Senior Presidential Secretary for AI Future Planning at Cheong Wa Dae, and Sung Kim, CEO of Upstage. Upstage is working to expand its presence in the enterprise artificial intelligence (AI) market with its own Solar large language model. During these meetings, AMD and its Korean counterparts are expected to explore potential cooperation focused on AI semiconductor-based infrastructure and AI model operating environments.

soup@fnnews.com Im Soo-bin, Joo Won-gyu, Cho Yoon-joo, Lee Dong-hyuk Reporter