Project Rainier, a large-scale artificial intelligence compute cluster developed by Amazon Web Services (AWS), is now operational. The infrastructure features nearly half a million Trainium2 chips and was deployed less than a year after its initial announcement. AWS partnered with Anthropic, an AI safety and research company, which is already using the system to run workloads for its Claude AI model.
Anthropic’s use of Project Rainier provides more than five times the computing power it previously used for training AI models. AWS expects that by the end of 2025, Claude will be operating on over one million Trainium2 chips for both training and inference tasks.
The cluster spans multiple data centers in the United States and is considered one of AWS’s most ambitious projects to date. Ron Diamant, an AWS distinguished engineer and head architect of Trainium, said: “Project Rainier is one of AWS’s most ambitious undertakings to date. It’s a massive, one-of-its-kind infrastructure project that will usher in the next generation of artificial intelligence models.”
Project Rainier uses what AWS calls an “EC2 UltraCluster of Trainium2 UltraServers.” Each UltraServer combines four physical servers equipped with 16 Trainium2 chips each, interconnected via high-speed NeuronLinks designed to reduce latency and accelerate complex calculations across all 64 chips per unit. These UltraServers are then connected at scale through Elastic Fabric Adapter networking technology.
AWS builds its own hardware and software stack for this infrastructure, allowing it to optimize performance and reliability throughout every level—from chip design to server architecture.
In addition to technical advancements, AWS has focused on sustainability measures within Project Rainier. All electricity consumed by Amazon operations in 2023 was matched with renewable energy resources. The company continues investing in nuclear power, battery storage, and global renewable energy projects as part of its commitment to reach net-zero carbon emissions by 2040.
New data center components being rolled out are expected to cut mechanical energy consumption by up to 46% and reduce embodied carbon in concrete by 35%. Water efficiency has also been improved; according to recent findings from Lawrence Berkeley National Laboratory (LBNL), AWS data centers achieve a water usage efficiency (WUE) of 0.15 liters per kilowatt-hour—more than twice as efficient as the industry average standard of 0.375 liters per kilowatt-hour.
AWS states that Project Rainier represents not only a technical achievement but also sets a template for future deployments aimed at enabling breakthroughs in fields such as medicine and climate science.
