• Amazon launches Graviton4 and Trainium2, two new AI chips for its cloud computing service
• Graviton4 is based on Arm architecture and provides up to 30% better compute performance, 50% more cores, and 75% more memory bandwidth than current generation Graviton3 processors
• Trainium2 is designed to deliver up to 4x faster training than first-generation Trainium chips and will be able to be deployed in EC2 UltraClusters of up to 100,000 chips

Detailed News

Amazon has announced the launch of two new artificial intelligence (AI) chips for its cloud computing service: Graviton4 and Trainium2. The chips are designed to provide customers with more powerful and energy-efficient AI computing capabilities.

Graviton4 is the fourth generation of Amazon's custom Arm-based server chips. Offers 75 percent more memory bandwidth, 50 more cores, and up to 30 percent better computing performance as compared to the prevailing Graviton 3 processors. This gives it great flexibility as a processing engine in applications such as machine learning, high-performance computing, or big data analytics.

Trainium2 is the second generation of Amazon's custom AI training chips. It is designed to deliver up to 4x faster training than the first-generation Trainium chips. This makes it ideal for training large language models (LLMs) and other complex AI models.

Amazon is also expanding its collaboration with NVIDIA to deliver the most advanced infrastructure, software, and services to power customers' generative AI innovations. This includes new Amazon Elastic Compute Cloud (EC2) instances that combine NVIDIA GPUs with Graviton4 processors and AWS software tools for AI development and deployment.

The launch of Graviton4 and Trainium2 is a major milestone for Amazon's cloud computing business. These new chips will give Amazon an upper hand over its competitors such as Microsoft and Google which have equally made huge investments in AI chips.