Amazon launches biggest AI data center in Indiana, forges AI future free of Nvidia

Amazon has launched its biggest and most strategic artificial intelligence data facility, an 11-billion-dollar complex in Indiana, which is already training large-scale AI models entirely on its own custom chips, bypassing the industry’s dominant supplier, Nvidia.

The center, called Project Rainer, is an unprecedented bet of Amazon to capture the whole AI stack, from the silicon to the data center power.

The center is situated on a 1200 acre property that was a piece of farmland just a year ago. The center is devoted to Anthropic, an AI-based startup behind the Claude chatbot.

Amazon Web Services (AWS) CEO Matt Garman highlighted that the project is not a future concept but a current operational reality.

“This is running and training their models today,” Gurman told CNBC, outlining the company’s ability to move from “cornfields to data centers, almost overnight.”

The most noticeable aspect of the Indiana data center is its full dependence on the in-house Trainium 2 chips of Amazon.

AWS boasts of 500,000 installed (and another 500,000 on the way) and claims it is the “largest” known non-Nvidia compute deployment in the world.

This is a strategic move to eliminate the reliance of Amazon on the market-leading GPUs offered by Nvidia and enables more profound and effective cooperation with the important partners such as Anthropic.

The next generation “Trainium 3” chips are designed with specific considerations of the needs of frontier AI models.

Garman explained that the new chip will soon be deployed at the Indiana site, citing, “It gives better performance, it gives better latency characteristics, it gets better power consumption.”