Amazon Web Services (AWS) has activated Project Rainier—its largest-ever AI data center development—transforming 1,200 acres of farmland in New Carlisle, Indiana, into one of the world’s most powerful AI compute hubs. In less than a year, AWS constructed seven massive data center buildings now filled with nearly 500,000 Trainium2 chips, all dedicated to training and running Anthropic’s Claude AI model. Two additional campuses are under construction, with the full site expected to include 30 buildings drawing up to 2.2 gigawatts (GW) of electricity, enough to power more than 1.5 million homes.
The $11 billion project—the largest capital investment in Indiana’s history—was fast-tracked through extensive collaboration among AWS, Anthropic, and local and state authorities. AWS received long-term sales tax exemptions on eligible capital investments through the Indiana Economic Development Corporation, along with $4 billion in property and technology tax abatements over 35 years. The project created nearly 9,000 construction jobs and will support 1,000 permanent positions, including 600 paying above the county average. AWS also committed $7 million for highway upgrades and $114 million for water and sewer infrastructure improvements.
AWS executives describe Rainier as the largest cluster of non-NVIDIA chips in the world, built entirely with AWS-designed silicon—Trainium2 today and Trainium3 soon. Each “UltraServer” integrates 64 Trainium2 chips linked by NeuronLink and Elastic Fabric Adapter interconnects, forming a vast EC2 UltraCluster capable of training foundation models at hyperscale. Anthropic is already using the site to train and deploy Claude models, with capacity expected to exceed one million Trainium2 chips by the end of 2025.
• 7 operational data centers today, expanding to 30 buildings across 1,200 acres
• 500,000 Trainium2 chips active; 1 million expected by year-end
• 2.2 GW total power draw; 1,000 permanent jobs
• $11B investment—the largest in Indiana’s history
• Entirely AWS silicon—no NVIDIA GPUs
• Trainium3 rollout planned for late 2025
• 100% renewable energy matching, on-site water treatment reducing use by 23%
AWS has been designing custom silicon since acquiring Annapurna Labs in 2015, introducing Inferentia in 2019, Trainium1 in 2021, and now Trainium2, which CEO Andy Jassy said delivers 30–40% better price-performance than GPU-based systems. AWS emphasized that vertical integration—spanning chips, software, and data center design—allows faster deployment, lower cost, and optimized energy use.
Power for the Indiana complex is supplied by Indiana Michigan Power, drawing from nuclear, coal, and renewable sources. The project alone will double the utility’s peak demand by 2030, pushing regional electricity prices higher, while Amazon expands its local wind and solar generation projects totaling 635 MW. Cooling is primarily air-based, requiring water for less than 2% of the year. AWS is studying aquifer impacts and revising site designs to reduce effects on wetlands.
🌐 Analysis:
Project Rainier positions AWS as a central player in the post-GPU era of AI infrastructure, where custom silicon, regional diversification, and speed-to-deployment define competitiveness. The Indiana campus demonstrates AWS’s ability to move faster than hyperscale rivals: Microsoft’s 2-GW Hyperion site in Louisiana and Google’s TPU-based builds are still under development. By integrating chip design, construction, and power strategy, AWS is defining a template for next-generation AI campuses—optimized for cost, performance, and sustainability.
🌐 We’re tracking the latest developments in AI infrastructure. Follow our ongoing coverage at: https://convergedigest.com/category/ai-infrastructure/
