Key takeaways
Amazon announced Monday it will invest up to $50 billion to expand artificial intelligence and high-performance computing capabilities specifically for US government customers, marking one of the largest cloud infrastructure commitments targeted at the public sector.
The investment, set to break ground in 2026, will add nearly 1.3 gigawatts of new AI and supercomputing capacity across Amazon Web Services' Top Secret, Secret, and GovCloud regions through the construction of data centers equipped with advanced computing and networking technologies.
Federal agencies will receive expanded access to AWS's comprehensive suite of AI services, including Amazon SageMaker AI for model training and customization, Amazon Bedrock for model and agent deployment, Amazon Nova, Anthropic's Claude models, and leading open-weight foundation models.
The infrastructure will incorporate both AWS Trainium AI chips and NVIDIA AI hardware.
AWS CEO Matt Garman stated the initiative aims to remove technological barriers facing government operations.
"Our investment in purpose-built government AI and cloud infrastructure will fundamentally transform how federal agencies leverage supercomputing," Garman said in a statement.
"We're giving agencies expanded access to advanced AI capabilities that will enable them to accelerate critical missions from cybersecurity to drug discovery. This investment removes the technology barriers that have held government back and further positions America to lead in the AI era."
Supporting national AI strategy
The announcement aligns with the Trump administration's AI Action Plan released in July 2025, which outlined over 90 federal policy actions to accelerate AI innovation, build American AI infrastructure, and lead in international diplomacy and security.
The plan emphasizes removing regulatory barriers and building infrastructure to support AI at scale while safeguarding national security.
Amazon's investment supports the plan's focus on establishing secure, US-based AI and cloud infrastructure for advanced computing initiatives deployed across federal agencies.
The White House has prioritized AI development as central to maintaining American economic competitiveness and national security advantages.
Building on the government cloud foundation
AWS has provided cloud infrastructure to government agencies since 2011, when it launched AWS GovCloud (US-West) as the first cloud provider to build infrastructure specifically for government security and compliance requirements.
In 2014, AWS introduced the first air-gapped commercial cloud accredited to support classified workloads, and in 2017 became the first cloud provider accredited across all US government data classifications: Unclassified, Secret, and Top Secret.
The company currently supports more than 11,000 government agencies with cloud services across various classification levels. The new investment represents a significant expansion of that existing infrastructure footprint.
Tech industry AI spending surge
The announcement comes amid an industry-wide race to build sufficient capacity to power AI services.
Amazon increased its capital expenditure forecast in October, projecting it will spend $125 billion in 2025, up from an earlier estimate of $118 billion. The company indicated it expects to spend even more during 2026.
Other tech giants have similarly announced massive infrastructure investments. Meta projects spending between $70 billion and $72 billion in 2025, while Microsoft expects to exceed the $88.2 billion it spent during fiscal 2025.
Google anticipates spending $91 billion to $93 billion in 2025 and plans to increase that amount the following year.
The federal government has seen growing interest from AI companies in recent months.
OpenAI launched a version of ChatGPT designed exclusively for federal agencies in January and announced a deal in August providing government agencies access to ChatGPT Enterprise for $1 annually.
Anthropic also expanded its government partnerships in August, offering federal agencies access to its Claude AI models.
Read more: