AWS to expand AI and supercomputing for the US government starting in 2026
AWS is preparing a major buildout of its secure cloud regions for federal use, with up to $50 billion in investment beginning in 2026. The plan expands Top Secret, Secret, and GovCloud capacity and adds nearly 1.3 gigawatts of power for advanced compute, backed by new data centers with updated chips and networking.
"Our investment in purpose-built government AI and cloud infrastructure will fundamentally transform how federal agencies use supercomputing," AWS CEO Matt Garman said. "We're giving agencies expanded access to advanced AI capabilities that will enable them to accelerate important missions from cybersecurity to drug discovery."
"The investment removes the technology barriers that have held government back and further positions America to lead in the AI era," he added.
What this means for your agency
- Access to AWS's full AI stack: Amazon SageMaker for model development, Amazon Bedrock for deploying models and agents, Anthropic Claude, Amazon Nova, AWS Trainium, and Nvidia AI hardware.
- More secure capacity across Top Secret, Secret, and AWS GovCloud (US) for data-intensive workloads and time-sensitive missions.
- Faster research and decision-making as larger datasets can be processed in real time.
Practical examples: teams working on global security can process decades of data on demand to spot patterns that once took months. Defense and intelligence groups can automate parts of analysis by running satellite imagery, sensor streams, and historical records through large-scale compute.
Scope and infrastructure
Amazon added 3.8 gigawatts of data-center capacity in the past year, and this plan adds another 1.3 gigawatts dedicated to advanced compute. For context, one gigawatt is roughly the average electricity use of about 750,000 US homes.
AWS's government cloud has hit key milestones over the past decade: GovCloud launched in 2011, the first air-gapped commercial cloud for classified workloads arrived in 2014, and in 2017 AWS became the first provider cleared to offer unclassified, secret, and top-secret regions.
Policy and market context
AWS says the investment supports federal efforts to build secure domestic AI infrastructure, aligning with White House AI initiatives. Market analysts note AWS remains the cloud leader but is pushing harder on AI infrastructure as Google and Oracle accelerate their offerings.
The broader tech sector continues to spend heavily on compute for training and running large models, including OpenAI, Alphabet, and Microsoft.
Indiana expansion adds capacity and jobs
Separately, Amazon plans about $15 billion in new Indiana data center campuses, on top of the $11 billion announced last year in St. Joseph County. The sites will add 2.4 gigawatts of capacity and use the same advanced infrastructure applied in Project Rainier, which Amazon describes as the world's largest AI supercomputer.
The Indiana buildout is expected to create more than 1,100 technical roles across operations, networking, engineering, and security, plus thousands of supply-chain jobs in construction, electrical work, and fiber installation.
How to prepare now
- Prioritize workloads: map mission needs to Top Secret, Secret, or GovCloud based on data classification and latency requirements.
- Data readiness: implement tagging, lineage, and retention policies; establish a model registry; enable end-to-end logging for auditability.
- Security and compliance: plan for ATO, Zero Trust patterns, key management, and cross-domain solutions; align controls with FedRAMP High where applicable.
- Capacity and cost planning: estimate training vs. inference needs; evaluate reserved capacity and scheduling for GPUs/Trainium; set cost guardrails.
- Model strategy: pick approaches (fine-tuning, retrieval-augmented generation, agents) and standardize on deployment via SageMaker and Bedrock.
- Workforce: upskill teams on AI engineering, data science, and secure MLOps. For role-based training, see AI courses by job.
Key takeaways
- $50B investment starting in 2026 to expand classified and GovCloud AI infrastructure.
- Nearly 1.3 GW in added advanced compute plus new chips and networking.
- Full access to AWS AI services and Nvidia hardware for model training and inference.
- Use cases span cybersecurity, drug discovery, global security analysis, and imagery processing.
- Additional Indiana buildout adds 2.4 GW and significant technical and supply-chain jobs.
Your membership also unlocks:
