Amazon plans to invest up to $50 billion to expand AI computing power for the U.S. government, adding 1.3GW of capacity

Amazon announced on Monday that the company will invest up to $50 billion to expand its artificial intelligence and supercomputing capabilities for U.S. government customers. Media reports describe this as one of the largest cloud infrastructure investments aimed at the public sector to date.

Amazon stated that the project is expected to begin construction in 2026. The project will add nearly 1.3 gigawatts of new AI and high-performance computing capacity in regions such as AWS Top Secret, AWS Secret, and AWS GovCloud by building new data centers, which will be equipped with advanced computing and networking systems.

1 gigawatt of computing capacity is roughly equivalent to the average power supply for about 750,000 American households.

Matt Garman, CEO of Amazon Web Services (AWS), stated:

"The AI and cloud infrastructure we are building for the government will fundamentally change how federal agencies use supercomputing. We are enabling agencies to access advanced AI capabilities more broadly, accelerating critical missions from cybersecurity to drug development.

This investment removes long-standing technological barriers that have hindered the government."

AWS is already a key cloud service provider for the U.S. government, currently serving over 11,000 government agencies. Amazon's stock price rose 1.89% during trading on Monday, closing at $224.87.

Technology companies, including OpenAI, Alphabet, and Microsoft, are investing billions of dollars to build AI infrastructure to meet the computing capacity demands required to support various services.

Amazon's plan aims to provide federal agencies with more convenient and comprehensive AWS AI services, including Amazon SageMaker for model training and customization, Amazon Bedrock for deploying AI models and agents, as well as foundational models like Amazon Nova and Anthropic Claude.

Amazon claims that this investment will enable government agencies to accelerate discovery and decision-making across various government tasks. By combining simulation and modeling data with AI, tasks that would typically take agencies weeks or even months to complete can now be accomplished in hours through automated experimentation guidance and real-time feedback loops. Research teams can process decades of global security data under real-time conditions and analyze hundreds of variables, transforming complex pattern analysis into immediately actionable insights while significantly reducing data volume.

Additionally, advanced computing can integrate data that was previously scattered across supply chains, infrastructure, and environmental monitoring into a complete picture. Work in defense and intelligence systems that would typically require weeks of manual analysis can now process satellite imagery, sensor data, and historical patterns on an unprecedented scale, automatically identifying threats and generating response plans