AWS has just debuted Trainium2 - a chip tailored toward training AI models - and providing access to Nvidia's upcoming H200 Tensor Core GPU.To further support customers, AWS is hosting a dedicated compute cluster and they can now start testing out the new Graviton4 general-purpose chips.
Amazon Web Services has announced new chips customers can use to build and deploy AI applications, as well as providing access to Nvidia's latest GPU offerings. In essence, Amazon seeks to create a cloud market that offers cost-effective options, featuring both their own Amazon-branded products and leading-edge products from other vendors.This may help them competition with Microsoft, who recently revealed their AI chip, the Maia 100, as well as offering access to the Nvidia H200 GPUs on their Azure cloud.At the Reinvent conference, Amazon announced they will allow customers to use Nvidia's H200 AI graphics processing units. They also showed the Trainium2 AI chip and the power-saving Graviton4 processor.The H200 is the same chip OpenAI used to develop their ChatGPT chatbot last year, and given a subsequent surge in demand for these chips, renting them out from cloud providers has become increasingly sought after. Amazon's Trainium2 chips are tailored for creating models such as AI chatbots, and companies such as Databricks and Anthropic (an OpenAI challenger) plan to use these chips for their projects. Graviton4 processors, which draw less power than Intel or AMD chips, will increase output thirty percent over existing Graviton3 models, making them a fast and cost-effective option for companies wanting to keep their AWS cloud bills low.Already, 50,000 customers are using Graviton chips. As for the partnership with Nvidia, AWS will give research teams and customers access to more than 16,000 Nvidia Grace Hopper Superchips, with Nvidia GPUs and Arm-based general-purpose processors.Graviton4 virtual-machine instances are available for testing now, and will become commercially available in the coming months. No release date has been announced for virtual-machines with either Trainium2 or H200 chips.
Be sure to check out CNBC PRO for these stories:A "cup and handle" pattern is forming in the S&P 500. Here's how to closely monitor it for potential breakoutBank of America believes the S&P 500 will expand to 5,000 by next year and expects an ideal environment for stock selectionMorgan Stanley is optimistic about an artificial intelligence trend and has identified 6 stocks that capitalize on itWall Street's favored stocks of Warren Buffett are featured on CNBC PRO
Comentarios