Amazon and Anthropic Complete AI Supercomputer with 500K Chips

What happened

Amazon Web Services (AWS) has completed Project Rainier, an AI supercomputer built in collaboration with Anthropic. The supercomputer features nearly 500,000 Trainium2 chips and is already being used by Anthropic to train its AI model, Claude. AWS plans to expand the system to over 1 million Trainium2 chips by the end of 2025.

Why it matters

• Project Rainier provides Anthropic with more than five times the compute power previously used for training its AI models.
• The large-scale AI infrastructure supports faster and more cost-effective AI training and inference.
• AWS’s vertical integration from chip design to software enables optimized performance and innovation in AI computing.

Key facts

1. 1. Project Rainier includes nearly 500,000 Trainium2 chips currently in operation.
2. 2. Anthropic expects to use over 1 million Trainium2 chips for Claude by the end of 2025.
3. 3. Project Rainier delivers five times the compute power compared to Anthropic’s previous GPU clusters.