Amazon Web Services has announced a big plan to build a dedicated AI cloud for the U.S. government. The project, which could cost up to $50 billion, is designed to support federal agencies as they expand their use of artificial intelligence and high-performance computing.
The project will add roughly 1.3 gigawatts of new computing capacity to Amazon’s existing government cloud systems. Construction is expected to begin in 2026, with the build-out spread across AWS’s secure platforms, including AWS Top Secret, AWS Secret, and AWS GovCloud (US).
The plan is to build an AI and high-performance computing infrastructure tailored to the U.S. government’s specific security and operational needs. This system will give over 11,000 federal agencies access to Amazon’s latest AI technologies.
Once the system is fully online, Washington will have a direct line into AWS’s most capable AI tools. Among them are:
- Amazon SageMaker: For training and customizing massive AI models.
- Amazon Bedrock: For developing “agentic” AIs that can perform complex, multi-step tasks.
- Open-Weights Models: Such as Amazon Nova and Anthropal’s Claude, providing flexible foundation models for specialized applications.
The system will be powered by a mix of AWS’s Trainium AI chips and high-end accelerators from Nvidia. Amazon says the setup will allow federal agencies to build custom AI tools, process massive datasets, and ultimately boost productivity across the government workforce.
Transforming Defense, Intelligence, and Science
Amazon believes large language models and agentic AI will significantly change how federal agencies operate. According to the company, the new cloud could modernize defense and intelligence processes, support advanced scientific research, and consolidate currently disconnected data systems.
Potential applications span a wide range of critical fields:
- Autonomous systems and robotics
- Advanced cybersecurity
- Energy and healthcare research
- Digital simulations for training and forecasting
The $50 billion project supports the U.S. government’s AI Action Plan, focused on developing a sovereign and advanced AI and cloud infrastructure domestically.
AWS chief Matt Garman says this step just feels like the natural next move in the company’s work with government agencies.
“This investment in purpose-built government AI and cloud infrastructure will fundamentally transform how federal agencies leverage supercomputing,” Garman stated. “It will remove the technology barriers that have held government back and further position America to lead in the AI era.”
Amazon’s work in the government cloud space dates back to 2011, when it introduced AWS GovCloud (US-West), the first cloud platform built to meet federal security standards. The launch set the stage for several major expansions that followed:
2014: AWS Top Secret – East, an air-gapped cloud for the highest classification levels.
2017: AWS Secret region, enabling handling of both classified and unclassified data.
2018-2025: Further expansion with AWS GovCloud (US-East), AWS Top Secret – West, and AWS Secret – West.
In a way, this $50 billion plan feels like the end goal of everything AWS has been building toward. It’s a move meant to lock in Amazon as the government’s go-to partner for cloud and AI for years to come.
Maybe you would like other interesting articles?