Amazon Web Services (AWS) and NVIDIA have extended their partnership to enhance infrastructure, software, and services for innovative generative artificial intelligence (AI) solutions.
The collaboration is geared towards equipping customers with the essential technologies for training foundational models and developing generative AI applications, as outlined in a press release on Tuesday (Nov. 28).
In a significant move, AWS will be the pioneering cloud provider to introduce NVIDIA GH200 Grace Hopper Superchips with cutting-edge multinode NVLink technology to the cloud environment. These chips will be accessible on Amazon Elastic Compute Cloud (Amazon EC2) instances, enabling mutual clients to scale up to thousands of GH200 Superchips. The GH200 NVL32 multinode platform integrates 32 Grace Hopper Superchips using NVIDIA NVLink and NVSwitch technologies within a single instance.
Furthermore, NVIDIA and AWS are set to collaborate on hosting NVIDIA DGX Cloud, a service for AI training, on AWS. This initiative will feature the first DGX Cloud incorporating GH200 NVL32, offering developers access to the largest shared memory within a single instance. DGX Cloud on AWS is poised to expedite the training of cutting-edge generative AI and large language models.
The partnership also encompasses Project Ceiba, where NVIDIA and AWS are jointly developing the world’s fastest GPU-powered AI supercomputer. This supercomputer will house 16,384 NVIDIA GH200 Superchips and will be utilized by NVIDIA for its in-house research and development endeavors in generative AI.
AWS is introducing three novel Amazon EC2 instances powered by NVIDIA GPUs: P5e instances designed for large-scale generative AI and high-performance computing (HPC) tasks, and G6 and G6e instances catering to a diverse array of applications such as AI fine-tuning, inference, graphics, and video workloads. G6e instances are particularly well-suited for the creation of 3D workflows and digital twin applications utilizing NVIDIA Omniverse.
The collaboration extends to software development, with NVIDIA NeMo Retriever microservice offering tools for crafting highly precise chatbots and summarization tools. Additionally, NVIDIA BioNeMo streamlines and expedites the model training process for drug discovery within pharmaceutical companies.
In a notable development from September, NVIDIA has emerged as the leading provider of computer chips utilized in AI operations. The company’s market value surged past the trillion-dollar milestone earlier this year, driven by the soaring demand for its chips crucial in training generative AI models.