Amazon Web Services (AWS) and NVIDIA have announced an expanded strategic collaboration aimed at delivering infrastructure, software, and services to support generative artificial intelligence (AI) innovations. The collaboration combes NVIDIA’s latest multi-node systems with advanced GPUs, CPUs, and AI software with AWS’s virtualization and security capabilities, Elastic Fabric Adapter (EFA) interconnect, and UltraCluster scalability.
These technologies are tailored for training foundational AI models and creating generative AI applications. This collaboration builds upon a longstanding relationship that has played a crucial role in advancing generative AI technologies. The joint effort aims to empower businesses across various industries to harness the potential of generative AI.
Key highlights of the collaboration include:
1.Introduction of NVIDIA GH200 Grace Hopper Superchips on AWS: AWS will become the first cloud provider to offer NVIDIA GH200 Grace Hopper Superchips with new multi-node NVLink technology in the cloud. These chips will be integrated into Amazon Elastic Compute Cloud (Amazon EC2) instances, benefiting from Amazon’s high-speed networking (EFA) and virtualization capabilities. This setup will allow customers to scale their AI workloads by connecting multiple GH200 Superchips, enabling more powerful AI model training.
2.NVIDIA DGX Cloud on AWS: NVIDIA’s AI-training-as-a-service, DGX Cloud, will be hosted on AWS, featuring the GH200 NVL32 architecture. This move will provide developers with access to the largest shared memory available in a single instance. DGX Cloud on AWS is expected to accelerate the training of advanced generative AI models, including large language models with over one trillion parameters.
- Collaboration on Project Ceiba. AWS and NVIDIA will collaborate on Project Ceiba, designed to create the world’s fastest GPU-powered AI supercomputer. This supercomputer will incorporate GH200 NVL32 technology and Amazon EFA interconnect. It will be used by NVIDIA for its research and development activities, particularly in the field of generative AI.
- New Amazon EC2 Instances: AWS will introduce three new Amazon EC2 instance types. These instances will be powered by NVIDIA GH200, H200, L40S, and L4 GPUs, catering to various AI, high-performance computing (HPC), design, and simulation workloads. The enhanced capabilities of these instances will enable customers to handle complex AI workloads and improve training performance.
- NVIDIA Software on AWS: NVIDIA will provide software on AWS to support generative AI development, including the NeMo LLM framework, NeMo Retriever microservice, and BioNeMo. These tools will enhance custom model development, semantic retrieval, and drug discovery in the generative AI field.
