AWS Commits US$50bn to New Federal AI Data Centres in the US

Amazon Web Services (AWS) has announced plans to invest up to US$50bn in new AI and high-performance computing infrastructure designed specifically for US federal agencies. 

The investment will expand capacity across AWS Top Secret, AWS Secret and AWS GovCloud (US) Regions and introduce data centres engineered for classified AI workloads for the first time.

The project is scheduled to break ground in 2026 and will give federal agencies access to AWS AI services across all classification levels, from unclassified to top secret environments.

Expanding capacity in classified regions

AWS says nearly 1.3GW of compute capacity will be added across three government cloud regions. The data centres will support AI training, model deployment and large-scale analytics using technologies that mirror AWS’ commercial AI stack.

Agencies will gain access to Amazon SageMaker for training and customisation, Amazon Bedrock for deploying foundation models and Amazon Nova, Anthropic Claude and open weights models. The new facilities will also use AWS Trainium chips alongside Nvidia infrastructure to provide the throughput required for AI and HPC workloads.

Matt Garman, AWS CEO, says: “Our investment in purpose-built government AI and cloud infrastructure will fundamentally transform how federal agencies leverage supercomputing. We’re giving agencies expanded access to advanced AI capabilities that will enable them to accelerate critical missions from cybersecurity to drug discovery.”

AWS states that these capabilities will be offered to existing and future US government customers across all three regions. The company currently serves more than 11,000 government agencies and the expansion is intended to support projects that require secure hosting and high density compute.

A significant expansion of AWS’ government footprint

The US$50bn investment follows a period of substantial infrastructure growth from Amazon. 

Throughout 2025, the company committed approximately US$100bn to AWS data centre development, including US$20bn in Pennsylvania and US$10bn in North Carolina. Some of these projects include facilities located next to energy assets such as the Susquehanna nuclear plant, reflecting the power demands of large AI clusters.

The new AI infrastructure for federal agencies aligns with the Trump administration’s AI Action Plan published in July 2025. The plan outlines more than 90 federal policy actions and emphasises the need to build domestic AI infrastructure to support national missions, scientific research and public sector innovation.

AWS maintains its lead in the cloud market with a c, ahead of Microsoft Azure and Google Cloud. The company says continued investment in AI data centres is required to meet growing demand from both public and private sector customers.

Infrastructure built for government security requirements

AWS has operated government-focused regions for more than a decade. In 2011 it launched AWS GovCloud (US-West), the first cloud environment designed to meet federal compliance rules. AWS Top Secret-East followed in 2014 as the first air-gapped commercial cloud accredited for classified workloads.

The company expanded this footprint with the AWS Secret Region in 2017 and then added GovCloud (US-East), Top Secret-West and Secret-West between 2018 and 2025. The upcoming US$50bn investment will build on this architecture with new high-density data centres tailored for AI workloads rather than general compute.

AWS positions the development as a way to remove barriers that have slowed the adoption of AI across federal missions.

Matt says: “This investment removes the technology barriers that have held government back and further positions America to lead in the AI era.”

The first construction phase is expected to begin in 2026, with additional details on the regional rollout to be provided as the project advances.