Amazon Builds Giant AI Supercomputer for Anthropic
Amazon Web Services (AWS) is building one of the largest AI supercomputers in the world—called Project Rainier—designed specifically to boost the AI capabilities of its close partner, Anthropic.
This massive system will use hundreds of thousands of custom AI chips and is expected to come online later this year, making it a powerful tool in the ongoing AI arms race.
What Is Project Rainier?
Project Rainier is a supercomputing cluster being developed by AWS across multiple locations in the U.S.
It will be powered by AWS’s in-house Trainium2 chips, which are designed especially for training large AI models.
Key Highlights:
Massive scale: Each UltraServer in Project Rainier houses 64 AWS-designed Trainium2 chips, and when thousands of these servers are linked across multiple U.S. data centers, they form a massive EC2 UltraCluster—delivering up to 5× more computing power than Anthropic’s previous setup.
Cutting‑edge networking: Built with high-speed interconnects—AWS’s NeuronLink and Elastic Fabric Adapter (EFA) v3—to ensure ultra‑low latency across sprawling facilities.
Custom-designed infrastructure: Trainium 2, developed by Annapurna Labs in AWS’s Austin chip lab, provides optimized performance and cost efficiency compared to conventional Nvidia GPUs—estimated 30–40% better price‑performance on generative AI tasks.
Next-gen readiness: AWS is already working on Trainium 3 (expected late 2025), designed to double compute capabilities and boost energy efficiency—Project Rainier is fully prepared to scale as new silicon arrives.
Massive Scale, Massive Power
- One of the largest sites is being built in Indiana, with 30 data centers, each covering about 200,000 square feet.
- The Indiana facility alone is expected to use 2.2 gigawatts of electricity—enough to power over a million homes.
- The project will also use millions of gallons of water for cooling.
- All this infrastructure is aimed at giving Anthropic an edge as it develops more advanced versions of its Claude AI models.
- Amazon is investing over $11 billion in the Indiana site alone.
- The project has also received up to $8 billion in tax incentives from local and state governments.
- Anthropic, which Amazon has heavily invested in, is the primary user of this system.
What’s Next?
- Once it goes live, Project Rainier will help Anthropic train the next generation of its Claude AI models.
- Amazon might eventually offer access to this system to other customers through its cloud platform (AWS).
- With Trainium3 chips coming soon, the power of this cluster will only grow.
News Gist
Amazon is building Project Rainier, a massive AI supercomputer for Anthropic, using hundreds of thousands of Trainium2 chips across U.S. data centers.
Set to launch in 2025, it will deliver 5× more compute power for training next-gen Claude AI models.
Amazon Builds Giant AI Supercomputer for Anthropic
Warmwind OS: The Smart AI Operating System That Works for You
Tencent Unveils Hunyuan A13B: Smarter, Faster AI Model
Apple Considers OpenAI or Anthropic to Power Next-Gen Siri
Amazon Launches New DeepFleet AI Model
