AMD Instinct™ MI350 series GPUs are designed for leadership AI and the acceleration of high-performance computing (HPC). Built on the 4th Gen AMD CDNA™ architecture, they deliver exceptional performance and efficiency for AI training, inference and complex HPC workloads.

• Built for next-gen AI and HPC

Powered by 4th Gen AMD CDNA architecture, MI350 series GPUs are designed to deliver leadership performance for both AI inferencing and high-performance computing workloads.

• Unmatched memory and bandwidth

Equipped with 288GB of HBM3E memory and 8TB/s memory bandwidth, MI350-series GPUs enable training and inference on large-scale AI models using fewer GPUs, improving efficiency and reducing cost.

• Expanded datatype support for accelerated AI

Supports advanced FP16, FP8 FP6, and FP4 data types to maximize computational throughput and power efficiency — ideal for generative AI, NLP and vision workloads.

• Drop-in compatible with existing platforms

MI350-series GPUs are air-cooled and seamlessly integrate with previous AMD Instinct MI300X/MI325X infrastructures, offering a cost-effective upgrade path without needing major hardware overhauls.

• AI-optimized software stack

Fully supported by AMD ROCm™ software, an open-source platform providing Day 0 support for major AI frameworks like PyTorch, TensorFlow, JAX, ONNX Runtime, Triton and more.

• Effortless deployment with Kubernetes

The AMD GPU Operator simplifies containerized deployment in Kubernetes clusters, enabling rapid configuration and scaling of AI applications.

• Security designed for enterprise AI

Built-in features like Secure Boot, Secure Update, SR-IOV and Infinity Fabric™ Link security ensure secure multi-tenant environments and protect AI models and data.

• Trusted by industry leaders

Major AI players including Microsoft and Meta rely on AMD Instinct GPUs to power large models such as Llama 405B and GPT, validating performance, scalability and reliability.

• Open ecosystem, with no vendor lock-In

Enables open Ethernet-based networking for scalable AI clusters and supports integrations across diverse cloud service providers and OEM platforms.

• Sustainable and efficient performance

Optimized for energy efficiency and reduced TCO, MI350 series GPUs are an ideal choice for data centers looking to scale AI while minimizing operational cost and environmental impact.

The AMD Instinct MI350 series GPUs and platform empower data centers with leadership performance. Whenever you want to make a movie, the first thing you need is a good script. MI350 series GPUs provide the computational power and efficiency to bring great AI and HPC visions to life, driving innovation at scale.

Learn More: ESC A8A-E12U | ASUS Servers