The race to build the infrastructure for artificial intelligence just gained momentum. This week, Samsung Electronics and AMD announced a deepened collaboration, signing a Memorandum of Understanding (MOU) to accelerate the development of next-generation AI technologies. The agreement, signed at Samsung’s advanced chip manufacturing complex in Pyeongtaek, Korea, focuses heavily on High-Bandwidth Memory (HBM) – a critical component for training the increasingly complex AI models powering everything from chatbots to image recognition software – and potentially extends to Samsung manufacturing AMD’s chips.
The partnership represents a significant expansion of a relationship spanning nearly two decades, but its timing is particularly noteworthy. It coincides with Nvidia’s GTC conference, a major event in the AI world, and follows recent praise from Nvidia CEO Jensen Huang for Samsung’s HBM4 technology, underscoring the Korean firm’s growing importance in the AI ecosystem. The deal highlights the intensifying competition to provide the specialized hardware needed to power the AI revolution, a market currently dominated by Nvidia but increasingly contested by AMD, and others.
At the heart of the agreement is Samsung’s commitment to supply its next-generation HBM4 to AMD. This advanced memory will be used in AMD’s upcoming Instinct MI455X AI accelerators, designed to handle the massive data processing demands of modern AI workloads. Samsung’s HBM4 boasts impressive specifications, built on a 6th-generation 10nm-class process, achieving processing speeds up to 13 Gbps and a bandwidth of 3.3 TB/s, according to the companies. This increased bandwidth is crucial for accelerating both AI inference – using trained models to produce predictions – and the computationally intensive process of AI training.
Beyond Memory: A Potential Foundry Partnership
Even as the initial focus is on HBM4, the MOU also signals a potential shift in the semiconductor manufacturing landscape. Both Samsung and AMD are exploring a “foundry partnership,” meaning Samsung could potentially manufacture AMD’s future chips on a contract basis. Reuters reported that this would be a significant win for Samsung’s foundry division, which is seeking to secure major clients in a market currently led by Taiwan Semiconductor Manufacturing Company (TSMC).
This potential expansion into chip manufacturing could reshape the competitive dynamics of the semiconductor industry. Currently, AMD relies on TSMC for the majority of its chip production. Diversifying its manufacturing partners would give AMD greater supply chain resilience and potentially reduce its dependence on a single provider. For Samsung, securing AMD as a major foundry customer would bolster its position as a leading chip manufacturer and help it compete more effectively with TSMC.
Optimizing the Full Stack: From Memory to Rack-Scale Systems
The collaboration extends beyond individual components, encompassing the entire AI infrastructure stack. Samsung and AMD plan to jointly develop optimized DDR5 memory for AMD’s 6th-generation EPYC processors, codenamed “Venice.” These processors are designed for data centers and servers, and the optimized memory will enhance their performance in AI applications. These components will be integrated into the AMD Helios rack-scale architecture, a system designed to scale AI performance across large data centers.
AMD has been aggressively pursuing partnerships and securing major contracts in the AI space. The company recently announced deals worth billions of dollars with companies like Meta and OpenAI, demonstrating its growing influence in the AI market. The partnership with Samsung is a key element of AMD’s strategy to provide a comprehensive AI infrastructure solution, from processors and memory to rack-scale systems.
“Powering the next generation of AI infrastructure requires deep collaboration across the industry,” said Dr. Lisa Su, Chair and CEO of AMD, in a statement released by the companies. “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.” Young Hyun Jun, Vice Chairman & CEO of Samsung Electronics, added, “Samsung and AMD share a commitment to advancing AI computing, and this agreement reflects the growing scope of our collaboration.”
The implications of this partnership extend beyond Samsung and AMD. The increased availability of advanced HBM4 and optimized processors could lower the cost and accelerate the deployment of AI technologies across various industries. This could lead to faster innovation in areas such as healthcare, finance, and transportation, as well as the development of new AI-powered applications.
Looking ahead, the companies will continue to work on aligning HBM4 supply for the AMD Instinct MI455X GPU and advanced DRAM solutions for the 6th Gen AMD EPYC CPUs. The next major milestone will be the availability of these technologies, which are expected to support next-generation AI systems in the coming years. The collaboration between Samsung and AMD is poised to play a crucial role in shaping the future of AI infrastructure.
What are your thoughts on this new partnership? Share your comments below and let us recognize how you feel this collaboration will impact the future of AI.
