Amazon has revealed plans to inject up to $50 billion into developing advanced artificial intelligence (AI) and high-performance computing infrastructure tailored for U.S. government clients via its Amazon Web Services (AWS) division.
Construction will start in 2026, introducing nearly 1.3 gigawatts of new data centre capacity specifically designed to serve federal agencies with cutting-edge technology.
This expansion will allow government departments to utilise a wide range of AI tools, including Amazon’s own SageMaker platform for custom AI model training, Amazon Bedrock for deploying foundation models and AI agents, custom Trainium AI processors, as well as access to Anthropic’s Claude models and Nvidia GPUs.
The initiative aims to boost the ability of public sector organisations to build custom AI solutions, improve their data processing efficiency, and bolster workforce productivity.
The announcement comes amid a surge of investments from other tech giants such as Anthropic, Meta, Oracle, OpenAI, and SoftBank, who are also committing vast sums to expand AI infrastructure in the U.S. Notably, the Stargate venture by Oracle, OpenAI, and SoftBank aims to dedicate around $500 billion over four years towards similar AI capabilities.
“This investment removes the technology barriers that have held government back and further positions America to lead in the AI era,” Matt Garman, CEO of AWS, stated.
The data centres’ power capacity—1.3 gigawatts—highlights the immense computational capability being deployed, supportive of highly complex AI operations comparable in scale to powering numerous thousands of homes.
Amazon recently adjusted its capital expenditure forecast for 2025 to $125 billion, up from an earlier $118 billion estimate, reflecting the company’s intensified focus on expanding cloud and AI infrastructure worldwide.