In a significant move to advance cloud-based artificial intelligence, Amazon Web Services (AWS) and NVIDIA have deepened their strategic partnership, announcing new supercomputing offerings and AI services designed to revolutionize generative AI. This collaboration combines AWS’s cloud infrastructure with NVIDIA’s cutting-edge computing technology, marking a pivotal moment for developers and enterprises aiming to harness the power of AI at an unprecedented scale.

Central to this announcement is the introduction of NVIDIA GH200 Grace Hopper Superchips on AWS, making AWS the first cloud provider to offer this advanced AI supercomputer capability. The GH200 combines NVIDIA’s high-performance GPU with an efficient Arm-based CPU, offering a shared memory up to 20 TB, designed to meet the demands of large-scale AI and machine learning workloads. This setup leverages AWS’s Elastic Fabric Adapter and Nitro System for high-throughput, low-latency networking and enhanced security, aiming to provide supercomputer-class performance for generative AI applications.

Additionally, the collaboration brings the NVIDIA DGX Cloud to AWS, featuring the GH200 NVL32 NVLink technology for the first time. This service targets developers and enterprises, providing them with a powerful platform for training complex AI models, including those exceeding 1 trillion parameters. The partnership also unveils Project Ceiba, a venture to create the world’s fastest GPU-powered AI supercomputer, designed to accelerate NVIDIA’s AI research and development efforts.

To cater to a broad range of AI, high-performance computing, and graphics needs, AWS is set to launch new EC2 instances powered by NVIDIA GPUs. These include the P5e instances, equipped with NVIDIA H200 GPUs for generative AI tasks, and the G6 and G6e instances, featuring NVIDIA L4 and L40S GPUs, respectively, tailored for AI inference, graphics, and video processing. These instances are engineered to support demanding applications like digital twin simulations and photorealistic 3D modeling using NVIDIA Omniverse.

Further enhancing the AI development ecosystem, NVIDIA has introduced new software tools on AWS. The NVIDIA NeMo Retriever microservice, BioNeMo on Amazon SageMaker, and integration with NVIDIA Omniverse are designed to streamline AI workflows in areas such as chatbot creation, drug discovery, and digital twin development. These tools aim to simplify the development of custom AI models and enhance the performance and accuracy of AI applications.

This partnership between AWS and NVIDIA not only underscores the companies’ commitment to driving innovation in AI but also sets a new benchmark for cloud-based supercomputing resources, offering businesses and developers the tools they need to push the boundaries of what’s possible with generative AI.