AMD is set to acquire Samsung's HBM4 high-bandwidth memory for its upcoming "AI accelerator", the Instinct MI455X GPU, alongside the 6th generation AMD Epyc CPUs, codenamed Venice. AI data center operators can integrate Instinct GPUs with Epyc CPUs and advanced architectures such as the AMD Helios platform to facilitate next-generation AI systems.
This agreement was formalized in a memorandum of understanding during a signing ceremony held today at Samsung's cutting-edge chip manufacturing facility in Pyeongtaek, Korea. The event was attended by AMD's CEO, Dr. Lisa Su, and Samsung Electronics' CEO and Vice Chairman, Young Hyun Jun.
Young Hyun Jun remarked:
"Samsung and AMD share a commitment to advancing AI computing, and this agreement reflects the expanding scope of our collaboration. From industry-leading HBM4 to next-generation memory architectures and sophisticated foundry and advanced packaging capabilities, Samsung is uniquely equipped to deliver unmatched turnkey solutions that align with AMD’s evolving AI roadmap."
Dr. Lisa Su added:
"Powering the next generation of AI infrastructure requires deep collaboration across the industry. We are excited to enhance our partnership with Samsung, combining their expertise in advanced memory with our Instinct GPUs, Epyc CPUs, and rack-scale platforms. Integration across the full computing stack—from silicon to system to rack—is vital to accelerating AI innovation and achieving substantial real-world impact at scale."
Samsung's HBM4 is produced using a 10nm process with a 4nm logic base die, offering processing speeds up to 13Gbps and a maximum bandwidth of 3.3TBps. Furthermore, Samsung and AMD will collaborate on high-performance DDR5 memory optimized for the 6th generation Epyc CPUs. The two companies are also exploring potential foundry partnerships, which would entail Samsung providing foundry services for next-generation AMD products.
Source
