Samsung Electronics and AMD have signed a new Memorandum of Understanding to significantly expand their strategic collaboration. Focused on the rapidly growing artificial intelligence sector, the agreement guarantees that Samsung will supply its next-generation AI memory technologies to power AMD’s upcoming high-performance processors. This deal highlights a shared push to create more efficient and powerful AI data centers.
The historic signing ceremony occurred on March 18, 2026, at Samsung’s highly advanced chip manufacturing complex located in Pyeongtaek, South Korea. AMD Chair and Chief Executive Officer Dr. Lisa Su traveled to the facility to meet with Young Hyun Jun, the Vice Chairman and Chief Executive Officer of Samsung Electronics. Together, the two leaders formalized a plan to integrate AMD’s computing hardware with Samsung’s cutting-edge memory and manufacturing capabilities.
Primary Supply for Advanced AI Hardware
A central piece of the new agreement is the supply of high-bandwidth memory. Samsung will serve as the primary supplier of its groundbreaking HBM4 technology for the highly anticipated AMD Instinct MI455X graphics processing unit. This specialized AI accelerator is designed to handle the heavy computational demands required for artificial intelligence model training and complex inference tasks.
The partnership also targets traditional computing power within modern data centers. Samsung and AMD will collaborate closely to develop high-performance DDR5 memory solutions. These advanced memory chips will be fully optimized to work alongside AMD’s 6th Generation EPYC central processing units, a server processor product line currently operating under the internal codename Venice.
Both the specialized HBM4 and the optimized DDR5 memory are essential components for the AMD Helios platform. This advanced rack-scale computing architecture combines AMD’s processors to deliver the massive scale and performance needed for modern AI infrastructure. As artificial intelligence workloads become more complex, combining reliable memory bandwidth with strict power efficiency is increasingly vital for maximizing overall system-level performance.
Industry-Leading HBM4 Innovation
Samsung’s new HBM4 memory represents a major technological leap for the global semiconductor industry. According to the company, it is the industry’s first HBM4 memory to successfully reach the mass production stage. The chips are constructed using Samsung’s most advanced sixth-generation 10-nanometer-class DRAM process, which is then combined with a 4-nanometer logic base die to maximize physical efficiency.
The performance metrics of this new memory technology push well beyond current industry standards. Samsung’s HBM4 features processing speeds that reach up to 13 gigabits per second. Additionally, it offers a maximum bandwidth of 3.3 terabytes per second. This immense speed ensures that massive amounts of data can flow continuously to AMD’s artificial intelligence accelerators without creating system bottlenecks.
By utilizing this industry-leading performance, energy efficiency, and high reliability, the AMD Instinct MI455X graphics processing unit is expected to become an optimal hardware solution for the world’s most demanding high-performance systems.
Exploring Future Foundry Services
While memory supply is the immediate focus, the memorandum also paves the way for a deeper manufacturing alliance. Samsung and AMD are actively discussing new opportunities to form a dedicated foundry partnership. Under such an arrangement, Samsung’s specialized foundry business would directly manufacture next-generation AMD silicon products, expanding Samsung’s traditional role from a memory supplier to a core manufacturing partner.
Young Hyun Jun emphasized the strategic benefits of this comprehensive approach. “Samsung and AMD share a commitment to advancing AI computing, and this agreement reflects the growing scope of our collaboration,” he said. He added that Samsung is uniquely positioned to deliver unrivaled turnkey capabilities, ranging from next-generation memory architectures and advanced packaging to cutting-edge foundry services.
Dr. Lisa Su echoed the pressing need for comprehensive industry teamwork to power the future of computing. “Powering the next generation of AI infrastructure requires deep collaboration across the industry,” she stated. Su explained that seamless integration across the entire computing stack—from the base silicon to the system and up to the server rack—is absolutely essential to translate AI innovation into real-world impact at scale.
Building on Decades of Collaboration
This latest memorandum is not the beginning of a new relationship, but rather the expansion of a highly proven partnership. Samsung and AMD have actively collaborated for nearly two decades, working closely together on a wide variety of graphics, mobile, and computing technologies.
Most recently, this ongoing relationship saw Samsung operating as the primary HBM3E memory partner for AMD. Samsung’s previous generation of high-bandwidth memory currently powers the latest AMD Instinct MI350X and MI355X AI accelerators. The transition to the new HBM4 standard marks the next logical step as both technology companies seek to support the rapidly expanding global artificial intelligence market.
