The signing ceremony was held at Samsung’s most advanced chip manufacturing complex in Pyeongtaek, Korea, attended by Dr. Lisa Su, Chair and CEO of AMD, and Young Hyun Jun, Vice Chairman & CEO of Samsung Electronics.

“Samsung and AMD share a commitment to advancing AI computing, and this agreement reflects the growing scope of our collaboration,” said Young Hyun Jun, Vice Chairman & CEO of Samsung Electronics. “From industry-leading HBM4 and next-generation memory architectures to cutting-edge foundry and advanced packaging, Samsung is uniquely positioned to deliver unrivaled turnkey capabilities that support AMD’s evolving AI roadmap.”

“Powering the next generation of AI infrastructure requires deep collaboration across the industry,” said Dr. Lisa Su, Chair and CEO of AMD. “We are thrilled to expand our work with Samsung, bringing together their leadership 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 essential to accelerating AI innovation that translates into real-world impact at scale.”

Under the MOU, Samsung and AMD will align on primary HBM4 supply for the next-generation AMD AI accelerator, the AMD Instinct MI455X GPU, as well as advanced DRAM solutions for 6^th Gen AMD EPYC CPUs, codenamed “Venice.” These technologies will support next-generation AI systems combining AMD Instinct GPUs, AMD EPYC CPUs and rack-scale architectures such as the AMD Helios platform.

Samsung and AMD are closely collaborating on advanced memory technologies for AI and data center workloads. As memory bandwidth and power efficiency become increasingly critical to system-level performance, this collaboration will help deliver more optimized AI infrastructure for customers.

An industry-first to enter mass production, Samsung’s HBM4 is built on its most advanced 6th-generation 10-nanometer (nm)-class DRAM process (1c) and a 4nm logic base die, featuring processing speeds of up to 13 gigabits-per-second (Gbps) and maximum 3.3 terabytes-per-second (TB/s) bandwidth that exceeds industry standards.

Powered by Samsung HBM4’s industry-leading performance, reliability and energy efficiency, the AMD Instinct MI455X GPU is expected to be the optimum solution for high-performance systems handling AI model training and inference.

The MI455X GPU will serve as a key building block for the AMD Helios rack-scale architecture, designed to deliver the performance and scalability required for next-generation AI infrastructure.

As part of their collaboration, Samsung and AMD will also work together on high-performance DDR5 memory optimized for the 6^th Gen AMD EPYC CPUs. The companies aim to deliver industry-leading DDR5 memory solutions for systems built on the AMD Helios rack-scale architecture.

The two companies will also discuss opportunities for foundry partnership, through which Samsung would provide foundry services for next-generation AMD products.

Samsung and AMD have collaborated for nearly two decades across graphics, mobile and computing technologies, including Samsung serving as the primary HBM3E partner to AMD, powering the latest AMD Instinct MI350X and MI355X AI accelerators.

About AMD
AMD (NASDAQ: AMD) drives innovation in high-performance and AI computing to solve the world’s most important challenges. Today, AMD technology powers billions of experiences across cloud and AI infrastructure, embedded systems, AI PCs and gaming. With a broad portfolio of AI-optimized CPUs, GPUs, networking and software, AMD delivers full-stack AI solutions that provide the performance and scalability needed for a new era of intelligent computing. Learn more at www.amd.com.

The centerpiece of Samsung’s showcase at NVIDIA GTC 2026 will be the new sixth-generation HBM4, which is now in mass production and is designed for the NVIDIA Vera Rubin platform. Samsung’s HBM4 is expected to help accelerate the development of future AI applications, delivering consistent processing speeds of 11.7 gigabits-per-second (Gbps), which exceeds the industry standard of 8Gbps, and can be enhanced to 13Gbps.

By leveraging the most advanced sixth-generation 10-nanometer (nm)-class DRAM process (1c), Samsung has achieved stable yields and industry-leading performance. The company’s next-generation HBM4E, delivering 16Gbps per pin and 4.0 terabytes-per-second (TB/s) bandwidth, will be on display as well for the first time at GTC 2026.

Visitors will also be able to catch a glimpse of Samsung’s hybrid copper bonding (HCB) technology, a new method that will enable next-generation HBM to achieve 16 or more layers while reducing heat resistance by more than 20 percent, compared to thermal compression bonding (TCB).

An Alliance Taking the AI Era to the Next Level

The strong collaboration between Samsung and NVIDIA will be highlighted in the booth’s separate ‘NVIDIA Gallery,’ specifically featuring a broad lineup of Samsung’s cutting-edge technologies, such as HBM4, SOCAMM2 and PM1763 SSD that are designed for NVIDIA AI infrastructure.

Addressing the need for maximum efficiency and scalability in AI systems, Samsung’s SOCAMM2 based on low-power DRAM is an optimum server memory module that offers high bandwidth and flexible system integration for next-generation AI infrastructure. Samsung’s SOCAMM2 is currently in mass production, an industry-first to reach that milestone.

Designed for next-generation AI storage solutions, Samsung’s PM1763 SSD is based on the latest PCIe 6.0 interface offering fast data transfers and high capacities. The industry-leading PM1763 performance will be demonstrated on servers working with the NVIDIA SCADA programming model.

As part of the new NVIDIA BlueField-4 STX reference architecture for accelerated storage infrastructure in NVIDIA’s Vera Rubin platform, Samsung’s PM1753 SSD will show how it helps to enhance energy efficiency and system performance for inference workloads.

Memory Architecture To Scale Intelligent Manufacturing

Samsung will showcase its collaboration with NVIDIA on AI Factory development at GTC 2026, including plans to implement NVIDIA accelerated computing to scale Samsung’s AI Factory and fast-track digital twin manufacturing that leverage NVIDIA Omniverse libraries. This collaboration powers one of the world’s most comprehensive chip manufacturing infrastructures ― spanning memory, logic, foundry and advanced packaging.

Separately, Yong Ho Song, Executive Vice President and Head of AI Center at Samsung Electronics, will illustrate the two companies’ strategic collaboration in further detail through his speaker session on March 17, 2026. Titled ‘Transforming Semiconductor Manufacturing with Agentic AI from Design and Engineering to Production,’ the session will elaborate on the company’s AI Factory and share innovative end-to-end real-life use cases where AI and digital twins are reshaping semiconductor manufacturing with breakthroughs in electronic design automation (EDA) and computational lithography to designing and operating advanced manufacturing facilities powered by NVIDIA.

Efficient Memory for Local Intelligence

Samsung’s memory solutions also offer maximized efficiency for local AI workloads on personal devices. During GTC 2026, Samsung will showcase tailored and efficient solutions for personal AI supercomputers, including the Samsung PM9E3 and PM9E1 NAND for NVIDIA DGX Spark.

Additionally, Samsung will display DRAM solutions, LPDDR5X and LPDDR6, that are designed for seamless integration into premium smartphones, tablets and wearable devices, offering faster data throughput and lower latency. LPDDR5X delivers speeds of up to 25Gbps per pin, while cutting power consumption by up to 15 percent, enabling ultra‑responsive mobile experiences, high‑resolution gaming and AI‑enhanced applications without sacrificing battery life.

Building on that foundation, LPDDR6 pushes bandwidth further to a scalable 30‑35 Gbps per pin and introduces advanced power‑management features such as adaptive voltage scaling and dynamic refresh control, which together provide the performance needed for next‑generation edge‑AI workloads.

By proactively leveraging its most advanced 6th-generation 10 nanometer (nm)-class DRAM process (1c), the company achieved stable yields and industry-leading performance from the outset of mass production — all accomplished seamlessly and without any additional redesigns.

“Instead of taking the conventional path of utilizing existing proven designs, Samsung took the leap and adopted the most advanced nodes like the 1c DRAM and 4nm logic process for HBM4,” said Sang Joon Hwang, Executive Vice President and Head of Memory Development at Samsung Electronics. “By leveraging our process competitiveness and design optimization, we are able to secure substantial performance headroom, enabling us to satisfy our customers’ escalating demands for higher performance, when they need them.”

Setting the Bar for Maximum Performance and Efficiency

Samsung’s HBM4 delivers a consistent processing speed of 11.7 gigabits-per-second (Gbps), exceeding the industry standard of 8Gbps by approximately 46% and setting a new benchmark for HBM4 performance. This represents a 1.22x increase over the maximum pin speed of 9.6Gbps of its predecessor, HBM3E. HBM4’s performance can be further enhanced up to 13Gbps, as well, effectively mitigating data bottlenecks that intensify as AI models continue to scale up.

Also, total memory bandwidth per single stack is increased by 2.7x compared to HBM3E, to a maximum of 3.3 terabytes-per-second (TB/s).

Through 12-layer stacking technology, Samsung offers HBM4 in capacities ranging from 24 gigabytes (GB) to 36GB. The company will also keep its capacity options aligned with future customer timelines by utilizing 16-layer stacking, which will expand offerings to up to 48GB.

In order to address power consumption and thermal challenges driven by the doubling of data I/Os from 1,024 to 2,048 pins, Samsung has integrated advanced low-power design solutions into the core die. HBM4 also achieves a 40% improvement in power efficiency by leveraging low-voltage through silicon via (TSV) technology and power distribution network (PDN) optimization, while enhancing thermal resistance by 10% and heat dissipation by 30%, compared to HBM3E.

By bringing outstanding performance, energy efficiency and high reliability to tomorrow’s datacenter environments, Samsung’s HBM4 enables customers to achieve maximized GPU throughput and effectively manage their total cost of ownership (TCO).

Comprehensive Yet Agile Production Capabilities

Samsung is committed to advancing its HBM roadmap through its comprehensive manufacturing resources — including one of the largest DRAM production capacities and dedicated infrastructures in the industry — ensuring a resilient supply chain to meet the projected surge in HBM4 demand.

A tightly integrated Design Technology Co-Optimization (DTCO) between the company’s Foundry and Memory Businesses allows it to secure the highest standards of quality and yield. Additionally, extensive in-house expertise in advanced packaging allows for streamlined production cycles and reduced lead times.

Samsung also plans to broaden the scope of its technical partnership with key partners, based on close discussions with global GPU manufacturers and hyperscalers focused on next-generation ASIC development.

Samsung anticipates that its HBM sales will more than triple in 2026 compared to 2025, and is proactively expanding its HBM4 production capacity. Following the successful introduction of HBM4 to market, sampling for HBM4E is expected to begin in the second half of 2026, while custom HBM samples will start reaching customers in 2027, according to their respective specifications.