Project Rainier
AWS AI compute cluster using Trainium2 chips for Anthropic
Project Rainier is a massive artificial intelligence computing cluster built by Amazon Web Services. It uses nearly half a million custom Trainium2 chips spread across multiple data centres in the United States. The system supports Anthropic in training and running its Claude AI models.
- Uses nearly 500,000 Trainium2 chips for AI training
- Primary tenant is Anthropic for Claude model development
- Spread across multiple US data centres including Indiana
- Water usage efficiency is 0.15 litres per kilowatt-hour
- Delivered less than one year after initial announcement
What it is
Project Rainier is a large-scale artificial intelligence computing infrastructure developed by Amazon Web Services. It functions as a unified cluster of custom-designed hardware rather than independent servers. The system relies on Trainium2 chips, which are specialised processors built for training complex AI models. AWS describes it as one of the largest AI compute clusters in operation. The project connects tens of thousands of servers to work on single problems simultaneously. It provides significantly more processing power than previous generations of AWS infrastructure. The primary purpose is to support the development and deployment of advanced AI systems.
How it came about
AWS first announced the project during its Re:Invent event in late 2024. The company collaborated closely with AI research firm Anthropic during the planning phase. Construction and deployment proceeded rapidly, with the cluster becoming operational less than a year after the initial announcement. By mid-2025, the system was fully active and supporting live workloads. Anthropic began using the infrastructure to build future versions of its Claude model. The rapid timeline marked a significant acceleration in AWS infrastructure delivery compared to prior projects.
The building and the halls
The infrastructure spans multiple data centre facilities across the United States. One confirmed location is a new campus in St. Joseph County, Indiana. This site involves an investment of eleven billion dollars and may include up to twenty-two buildings. The hardware architecture uses UltraServers, each combining four physical servers with sixteen chips each. These units connect via high-speed NeuronLinks, identified by blue cables. Yellow cables mark the Elastic Fabric Adapter networking that links servers across different buildings. This design minimises latency and allows the cluster to scale across separate physical sites.
Power and cooling
AWS states that all electricity consumed by its operations was matched with renewable energy resources in 2023 and 2024. The company is investing in nuclear power and battery storage to support future growth. Project Rainier facilities incorporate new components designed to reduce mechanical energy consumption by up to forty-six per cent. Water usage is minimised through engineering innovations. In Indiana, the data centres use outside air for cooling from October to March without any water. During warmer months, water is used for cooling only for a few hours each day. This approach results in a water usage efficiency well below the industry average.
What runs inside
Anthropic is the primary tenant using Project Rainier for its AI development. The cluster supports the training and inference of the Claude language model. By the end of 2025, Claude workloads are expected to run on more than one million Trainium2 chips. The system provides five times the compute power used for previous model training cycles. Increased computational capacity allows for the creation of smarter and more accurate AI models. The infrastructure serves as a template for deploying raw computational power at a scale previously unseen in the industry.
The place
St. Joseph County in Indiana is a key site for the project. The region was selected for the construction of new data centre buildings to support the cluster. The local campus represents a major infrastructure investment for the area. The design of these facilities prioritises energy efficiency and sustainability upgrades. The choice of location allows for the use of free-air cooling during colder months. This reduces the environmental impact of the facility compared to traditional data centre designs that rely heavily on water-based cooling systems year-round.
Environment and impact
The project aims to maintain AWS commitment to net-zero carbon emissions by 2040. New construction methods reduce embodied carbon in concrete by thirty-five per cent. Water stewardship is a central focus, with the company claiming industry-leading efficiency metrics. The Lawrence Berkeley National Laboratory reported AWS water usage efficiency at 0.15 litres per kilowatt-hour. This figure is less than half the industry standard of 0.375 litres. The rapid deployment of such large-scale computing resources raises questions about energy demand, though AWS offsets this through renewable energy purchases and efficiency gains.
Network and links
The cluster uses third-generation Elastic Fabric Adapter networking technology. This system operates at petabit scale with low latency. It connects UltraServers within individual data centres and links them across different facilities. NeuronLinks provide high-bandwidth connections inside the UltraServer units. This two-tier network architecture ensures fast communication between chips while maintaining flexibility for scaling. The network allows tens of thousands of servers to act as a single cohesive unit for complex calculations.
Questions people ask about Project Rainier
What is Project Rainier used for?
It is an AI compute cluster used by Anthropic to train and deploy its Claude language model. The system provides the massive processing power needed to develop frontier AI models with greater accuracy and capability.
Where is Project Rainier located?
The cluster is spread across multiple data centres in the United States. One confirmed site is a new campus in St. Joseph County, Indiana, which involves an eleven billion dollar investment and up to twenty-two buildings.
How does Project Rainier cool its servers?
It uses a combination of free-air cooling and limited water usage. In Indiana, facilities use outside air for cooling from October to March without water. In warmer months, water is used for only a few hours daily, improving efficiency.
Who owns and operates Project Rainier?
Amazon Web Services owns and operates the infrastructure. It was developed in collaboration with Anthropic, which is the primary customer using the compute capacity for its artificial intelligence research and product development.
Good reading and sources
There is no English Wikipedia article on Project Rainier yet, so the facts on this page come from the sources listed above: aboutamazon.com/news/aws/aws-project-rainier-ai-trainium-chips-compute, datacenterdynamics.com/en/news/amazon-details-project-rainier-compute-. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryDatacenter wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.