Amazon launches $11 billion Rainier project to power Anthropic’s AI models

Amazon has launched its $11 billion Rainier project, which will utilize over one million Trainium 2 chips to power Anthropic’s Claude model by the end of the year. The 1,200-acre rural Indiana complex, near Lake Michigan, is dedicated to training Anthropic’s AI models exclusively on Amazon chips. 

The tech company stated that the Rainier data-center cluster currently operates on approximately 500,000 Tranium2 chips, and Anthropic has pledged to double its investment in Amazon’s AI infrastructure. Meanwhile, Amazon has committed roughly $8 billion, and Google has injected about $3 billion into Anthropic. Amazon Web Services (AWS) refers to Project Rainier as the world’s most powerful AI training compute cluster. 

Cryptopolitan recently reported that Anthropic is not out there performing; it is calculating, methodical, and scaling its infrastructure to the max. Its Claude models are split across Nvidia GPUs, Google’s TPUs, and Amazon’s Trainium chips, and each chip performs a specific task. Research is conducted from one location, training takes place in another, and inference is made elsewhere.  

Garman says the future is happening now

AWS CEO Matt Garman said the Rainier compute cluster is not a future concept but a project that is already online. However, skeptics are reportedly wondering whether there is sufficient funding, energy, and community support to bring these ambitious AI data centers to fruition.   

The executive claimed that Amazon is delivering, thanks to its decades of experience in handling large-scale logistics data. It has also formed deep and close ties with local and state officials through its massive fulfillment centers, HQ2 project, AWS data centers, and logistics hubs.  

Mike Krieger, the chief product officer at Anthropic, said the deals only look good on paper but will only materialize if they are racked, loaded, and usable by customers. However, he acknowledged that Amazon is incredible at that. Josh Sallabedra, who leads the Indiana Rainier site and has built Amazon data centers for over 14 years, said he has never seen the tech firm move this fast. Garman described it as a transition from cornfields to data centers, almost overnight. 

Krieger claims high demand for his company’s models

Anthropic’s Mike Krieger emphasized high demand for his company’s AI models, adding that the multi-chip strategy has enabled the company to serve as much as it has this year. However, he noted that the company will need all the processing power it can get. 

Meanwhile, Garman claimed he was well aware of the multi-chip and multi-cloud efforts, but he clarified that Amazon has no plans for anything drastic, like bidding for Anthropic. He added that both companies are satisfied with the partnership as it is.

Anthropic has also stated that it does not count exclusively on Amazon as it navigates its future path. The company announced last week that it is partnering with Alphabet to gain access to $1 million worth of Google’s customized TPUs (Tensor Processing Units). The AI firm disclosed that the deal is worth tens of billions of dollars, although no specifics were shared. 

Anthropic has emphasized that it is moving at a breakneck speed and burning stacks of cash to keep up with other AI giants. Its Claude chatbot already powers over 300,000 businesses, a 300x increase over the past two years. Large enterprise clients, each generating over $100,000 in annual revenue, have surged almost 7x in one year. Meanwhile, the company’s Claude Code generated $500 million in annualized revenue within its first eight weeks. 

Prasad Kalyanaraman, the Vice President of Infrastructure Services at AWS, believes it is critical to build the proper set of capabilities that model providers want. He added that controlling the stack from the infrastructure’s lower layers is equally critical.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.