Amazon revamps Indiana AI campus to launch massive Trainium server cluster
Amazon is redesigning part of its Indiana data-center complex to link multiple sites into a larger AI supercluster, deploying thousands of Trainium-based servers for its next frontier model.
Amazon is overhauling a portion of its extensive AI data-center campus in rural Indiana, consolidating multiple facilities into a unified "AGI SuperCluster" that will host thousands of new Trainium-based servers. Planning papers detail an "emergent request" to install more than 6,000 such servers and to revamp networking, storage and fiber-optic systems so the sites operate as a single, high-performance unit. The accelerated rollout is intended to deliver enough compute capacity for the AGI organization’s next frontier model before the upcoming re:Invent conference.
While the upgrade will not affect the existing Project Rainier hardware built for Anthropic, older Trainium 2 units will be replaced with newer Trainium 3 machines. The redesign reflects Amazon’s strategy, voiced by CEO Andy Jassy, to extract greater value from its AI infrastructure investments as demand outpaces supply. Despite recent job cuts in the AGI division and the winding down of the Nova model series, the company’s AI ambitions remain strong, backed by a raised 2026 capital-expenditure outlook of $220 billion.
Why it matters
The upgrade boosts Amazon's AI computing power, influencing the pace of next-generation model development and the broader AI market.
In this story