Product Highlights - AMD Instinct™ MI350 Series GPUs are designed for Generative AI and high-performance computing (HPC) acceleration in data centers [1] - The GPUs are built on the 4th Gen AMD CDNA™ architecture [1] - The series aims to deliver efficiency and performance for training AI models, high-speed inference, and complex HPC workloads [1] Target Applications - The GPUs are suitable for training massive AI models [1] - They are also applicable for high-speed inference [1] - Complex HPC workloads such as scientific simulations, data processing, and computational modeling can also benefit from these GPUs [1] Legal and Trademark Information - ©2025 Advanced Micro Devices, Inc [1] - AMD, the AMD Arrow Logo are trademarks of Advanced Micro Devices, Inc in the United States and other jurisdictions [1]

Core Insights - Micron Technology has announced the integration of its HBM3E 36GB 12-high memory into AMD's Instinct™ MI350 Series GPUs, emphasizing the importance of power efficiency and performance in AI data center applications [4][5] - This collaboration marks a significant milestone for Micron in the high-bandwidth memory (HBM) industry, showcasing its strong customer relationships and execution capabilities [4][5] Product Features - The HBM3E 36GB 12-high solution provides outstanding bandwidth and lower power consumption, supporting AI models with up to 520 billion parameters on a single GPU [5] - The AMD Instinct MI350 Series platforms can achieve up to 8 TB/s bandwidth and a peak theoretical performance of 161 PFLOPS at FP4 precision, with a total of 2.3TB of HBM3E memory in a full platform configuration [5] - This integrated architecture enhances throughput for large language model training, inference, and scientific simulations, allowing data centers to scale efficiently while maximizing compute performance per watt [5] Strategic Collaboration - Micron and AMD's close working relationship optimizes the compatibility of Micron's HBM3E product with the MI350 Series GPUs, providing improved total cost of ownership (TCO) benefits for demanding AI systems [6] - The collaboration aims to advance low-power, high-bandwidth memory solutions that facilitate the training of larger AI models and the handling of complex high-performance computing (HPC) workloads [7]