Sezzle FinTech Tracker January 2024 Banner

AWS and NVIDIA Expand Collaboration to Deliver Generative AI Infrastructure

Amazon Web Services (AWS) and NVIDIA have expanded their collaboration to deliver advanced infrastructure, software and services for generative artificial intelligence (AI) innovations.

The partnership aims to provide customers with the necessary technologies for training foundation models and building generative AI applications, the companies said in a Tuesday (Nov. 28) press release.

As part of the collaboration, AWS will be the first cloud provider to bring NVIDIA GH200 Grace Hopper Superchips with new multinode NVLink technology to the cloud, according to the release. These chips will be available on Amazon Elastic Compute Cloud (Amazon EC2) instances, allowing joint customers to scale to thousands of GH200 Superchips. The GH200 NVL32 multinode platform connects 32 Grace Hopper Superchips with NVIDIA NVLink and NVSwitch technologies into one instance.

Additionally, NVIDIA and AWS will collaborate to host NVIDIA DGX Cloud, an AI-training-as-a-service, on AWS, the release said. This will be the first DGX Cloud featuring GH200 NVL32, providing developers with the largest shared memory in a single instance. DGX Cloud on AWS will accelerate the training of cutting-edge generative AI and large language models.

The collaboration also includes Project Ceiba, where NVIDIA and AWS are designing the world’s fastest GPU-powered AI supercomputer, per the release. This supercomputer will feature 16,384 NVIDIA GH200 Superchips and will be used by NVIDIA for its own research and development in generative AI.

AWS will introduce three new Amazon EC2 instances powered by NVIDIA GPUs: P5e instances for large-scale generative AI and high-performance computing (HPC) workloads, and G6 and G6e instances for a wide range of applications including AI fine-tuning, inference, graphics and video workloads, according to the release. G6e instances are particularly suitable for developing 3D workflows and digital twin applications using NVIDIA Omniverse.

The collaboration also extends to software development, the release said. NVIDIA NeMo Retriever microservice offers tools to create highly accurate chatbots and summarization tools, while NVIDIA BioNeMo simplifies and accelerates pharmaceutical companies’ training of models for drug discovery.

It was reported in September that NVIDIA has become the go-to company for computer chips used in AI processes. The company’s market value hit the trillion-dollar mark earlier this year thanks to high demand for its chips that are used to train generative AI models.