Amazon commits up to $50B to expand AI and supercomputing for US govt customers
AWS said the multi‑year plan will add nearly 1.3GW of high‑performance capacity across its classified and GovCloud regions, with construction set to begin in 2026.
Amazon has announced a commitment of up to $50 billion to expand artificial intelligence (AI) and supercomputing infrastructure for U.S. government agencies, in what has been described as one of the largest public‑sector cloud infrastructure commitments to date.
The programme will be delivered through Amazon Web Services (AWS) and is set to start construction in 2026.
AWS will build and deploy purpose‑built AI and high‑performance computing (HPC) capacity across its classified and sovereign U.S. regions—AWS Top Secret, AWS Secret, and AWS GovCloud (US).
The plan targets nearly 1.3 gigawatts of additional compute, delivered via new data centres using advanced compute and networking technologies.
Scope and services
Under the initiative, federal customers are promised expanded access to AWS’s AI stack, including Amazon SageMaker for custom model training, Amazon Bedrock for model and agent deployment, Amazon Nova, and access to third‑party and open‑weight foundation models such as Anthropic Claude.
The build‑out will incorporate AWS Trainium chips alongside NVIDIA‑based infrastructure.
AWS has already served more than 11,000 U.S. government entities; the company said the new capacity will be available to existing and future customers across classification levels.
Matt Garman, CEO, AWS, said the investment will remove technology barriers and “fundamentally transform” how agencies use supercomputing for critical missions—from cybersecurity to scientific discovery.
The investment comes as hyperscalers step up spending to meet soaring demand for AI workloads in the public sector.
For agencies, purpose‑built and accredited capacity across Secret and Top Secret environments could mean faster access to modern tooling, potential cost efficiencies, and reduced lead times for mission programmes that depend on HPC and AI.
The competitive landscape
Amazon’s announcement has arrived as rivals have accelerated their own public‑sector AI efforts.
In September 2025, Microsoft and the U.S. General Services Administration unveiled a government‑wide agreement intended to speed adoption of secure AI tools and cloud services, including Microsoft 365 Copilot, with the promise of significant cost savings for agencies.
Separately, the U.S. Department of Energy, alongside Argonne National Laboratory, NVIDIA and Oracle, announced plans for the DOE’s largest AI supercomputer to accelerate open science—underpinning a broader build‑out of national AI infrastructure.
Oracle and Microsoft have also expanded their government multicloud connectivity, adding support for interconnect in U.S. government cloud regions to help agencies run mixed Azure/Oracle workloads at FedRAMP High.