Amazon Unveils 6 Models, Trainum3 Chip, and Largest AI Cluster with Apple’s Support

Amazon's AWS unveils largest AI computing cluster with $8B investment in Anthropic, featuring new AI chips, servers, and models for generative AI innovation.

After spending $8 billion, Amazon "locks in" Anthropic to build the world's largest AI computing cluster.

The world’s largest cloud computing giant has just pulled out its hard-hitting resources!

At the annual cloud computing industry event, the AWS re: Invent conference, AWS (Amazon Web Services) released 6 large models, previewed 2 more, and launched the 3nm third-generation AI training chip Trainium3 along with its most powerful AI server to date, the Trn2 UltraServer.

This marks the first time the Amazon Nova series foundational models have been revealed.

These include four versions of language models: Micro, Lite, Pro, and Premier, as well as image generation model Canvas and video generation model Reel. Reel currently supports generating 6-second videos, with plans to extend this to 2 minutes soon.

Amazon President and CEO Andy Jassy gave a sneak peek: Next year, AWS will release not only speech-to-speech models but also even more powerful Any-to-Any models, which can input and output content in any modality, including text, images, audio, and video.

These models are available in Bedrock, and they can be fine-tuned and distilled.

Additionally, AWS CEO Matt Garman announced the launch of the latest generative AI instance, Amazon EC2 Trn2, offering 30% to 40% better cost-effectiveness than current GPU-based EC2 instances.

Benoit Dupin, Senior Director of Machine Learning and AI at Apple, attended the event, mentioning that Apple extensively uses Amazon services across products like iPad, Apple Music, Apple TV, News, App Store, and Siri.

He specifically shared that compared to x86 instances, Apple has achieved a 40% efficiency improvement in machine learning inference workloads using AWS chips such as Amazon Graviton and Inferentia, and expects a 50% improvement in efficiency when pre-training models on Trainium 2.

Anthropic, which received an $8 billion investment from Amazon, announced that it will collaborate with AWS to build the world’s largest compute cluster for machine learning training, with the next-generation Claude large model being trained on a cluster with hundreds of thousands of Trainium2 chips.

Overall, AWS’s computing, storage, database, and AI inference sectors all saw a wave of important updates.