U.S. Federal Supercomputing on AWS
U.S. Federal Supercomputing on AWS
An investment of up to $50 billion in American AI leadership, the largest federal technology commitment in AWS history.
U.S. Federal Supercomputing on AWS advances American AI leadership by delivering integrated AI and HPC capability across AWS Government Regions.
AI and HPC Supercomputing Pods on AWS are available in AWS GovCloud (US), AWS Secret Cloud, AWS Top Secret Cloud.
Supercomputing built for the mission
AI and HPC Supercomputing Pods on AWS are integrated, elastic, secure, and agile CPU, GPU, and accelerator clusters optimized to accelerate R&D and mission outcomes for workloads of all sizes. These pods take advantage of dense network placement and compute diversity within a single Availability Zone. Pods grow from a single-core proof of concept to production scale on demand. Pay-as-you-go economics eliminate the lengthy technology refresh cycles typical of on-premises deployments.
Native integration with AWS services extends the functionality of AI and HPC Supercomputing Pods on AWS. Amazon Bedrock delivers multimodal AI inference, and Amazon SageMaker AI provides model training and inference. Amazon S3 and Amazon FSx for Lustre provide storage. AWS Parallel Computing Service (PCS), Amazon Elastic Kubernetes Service, and AWS Batch orchestrate compute.
AI and HPC Supercomputing Pods on AWS operationalize America’s AI Action Plan – available today to elevate mission outcomes.
Supercomputing built for American AI leadership
Technology access no longer gates the pace of innovation. Dedicated gigawatt-scale capacity runs on a supercomputing-grade fabric where AI and HPC share one data and security boundary. The latest accelerators and frontier models are available, and compliance is inherited at key classification levels. On-demand consumption lets government and industry build and deploy mission capabilities at the pace the mission commands, drawing on the full portfolio of AWS services.
Converged AI and HPC performance
Run AI and HPC on shared data, networking, and compliance boundaries, covering over 90 percent of supercomputing use cases. Elastic Fabric Adapter provides low-latency, petabit-scale networking.
Built into AWS Government Regions
Deploy in existing AWS Government Regions with direct access to accredited AWS services.
Leverage the latest generations of silicon
Access new CPUs and GPU generations within months of release. Adapt compute and managed services, including Amazon Bedrock and Amazon SageMaker AI, to mission requirements.
Capacity that grows with the mission
Burst compute capacity on demand. Scale storage and databases independently. Augment on-premises supercomputing with cloud capacity and AI tooling. Validate applications, move to production, and scale on the same infrastructure.
Who benefits from AI and HPC Supercomputing Pods on AWS
This investment reaches every part of the federal mission community. Agencies, the industrial base, technology partners and research institutions each have a path to AI and HPC Supercomputing Pods on AWS.
Train AI models, run HPC simulations, and deploy agents at key classification levels to compress innovation and production timelines from years to months. Agencies move from simulation to inference to decision in one continuous workflow.
Collaborate to develop and scale AI and HPC-powered mission solutions on a common platform at the speed that operations and the contract demand. One platform runs high-fidelity HPC simulation alongside AI model training and inference. Validated design, test and analysis move toward production at mission speed.
ISVs and SaaS technology partners build and deliver application software and solutions that support the supercomputing ecosystem with inherited security and compliance controls. Partners integrate AI and agentic services to bring innovative offerings tailored to mission and domain needs, in every security partition.
University Affiliated Research Centers (UARCs), Federally Funded Research and Development Centers (FFRDCs), and academic institutions are a core part of the supercomputing community. These entities gain access to supercomputing-class AI and HPC capabilities without the capital cost and refresh cycles of on-premises systems for the term of their engagement. Through Cooperative Research and Development Agreements (CRADAs) and other partnership vehicles, researchers collaborate with agencies on a common, compliant AI and HPC platform.
AI and HPC convergence changes how missions innovate
The convergence of AI, HPC and software is here today and accelerating what’s possible by using AI to augment HPC numerical simulation methods. Mission owners can orchestrate expert AI models, agents, and natural language interfaces to explore complex problems through intelligent orchestration and conversation. They specify challenges and receive AI-driven recommendations validated by high-fidelity simulation and modeling. Missions that once waited weeks between simulation, analysis, and the next experiment now iterate continuously on the same data, same fabric, and same compliance boundary. This compresses mission timelines from years to months and from weeks to days. It solves the most complex computational challenges with unprecedented speed and precision.
Through Amazon Bedrock AgentCore and Amazon Quick, mission owners orchestrate expert AI models and agents. Frontier models from Amazon Nova, Anthropic Claude, Meta Llama, NVIDIA Nemotron, OpenAI GPT, xAI Grok, and leading open-weight models seamlessly integrate with HPC workloads. Agencies can choose the right model for each component of the mission.
Amazon Bedrock provides fully managed inference, serving foundation models through a single API in your security partition, with no servers to provision, patch, or scale. Mission teams focus on results, not infrastructure.
AI now guides experiments, simulations generate training data, and agents orchestrate the workflow end to end. Amazon Bedrock AgentCore runs multi-step agentic workflows across tools, data, and simulation. Built-in guardrails enforce access, data boundaries, and audit logging. Retain visibility and control of agent decisions and actions through centralized governance policies.
Managed HPC workflow orchestration services schedule and distribute workloads across the right mix of CPU, GPU, and accelerator resources. Researchers reduce time to results instead of managing clusters. Services such as AWS Parallel Computing Service, AWS Batch, and Amazon Elastic Kubernetes Service handle job placement, scaling, and resource optimization across the full range of mission workloads.
Develop, train, and fine-tune models using Amazon SageMaker AI and Amazon Bedrock, tailored for the mission use case and specific operational requirements.
Compliant AI and HPC Supercomputing Pods on AWS accelerate innovation
AWS Government Regions are accredited across key U.S. government data classifications, including Controlled Unclassified Information, Secret, and Top Secret. They have achieved certifications and authorizations for FedRAMP Class D, DoD CSP SRG Impact Levels 4-6, ICD 503, ICD 705, ITAR, and CMMC among other relevant compliance frameworks. Missions can promote workloads and data from unclassified research through CUI development to Secret and Top Secret production.
Move faster with inherited compliance
Infrastructure and services for AI and HPC Supercomputing Pods on AWS are offered at key accreditation levels. Multi-partition operations enable consistent, repeatable architectures and deployment patterns, and a seamless user experience that spans mission needs. Model, simulate, and train at one level, then deploy across cross-domain pathways to higher classifications, all while maintaining full data custody and sovereignty.
Deploy where the mission demands it
AI and HPC Supercomputing Pods on AWS are available through a range of cloud-based, hybrid and on-premises options including AWS AI Factories. AWS-operated AI and HPC infrastructure, with managed AI and HPC orchestration services run in the cloud or inside your own facilities — with both options offering full stack capability and data sovereignty.
AI and HPC Supercomputing Pods on AWS use cases
AI and HPC Supercomputing Pods on AWS put integrated compute, storage, models, and agents to work on the missions that matter most. Five patterns show where the capability goes to work.
Position America's security capabilities for the AI era with AI and HPC infrastructure across key classification levels. Defense and intelligence workflows that once required weeks of manual analysis can now automatically detect threats and generate response plans. They process satellite imagery, sensor data, and historical patterns at unprecedented scale.
The convergence of AI with HPC modeling and simulation creates a new paradigm for scientific discovery. Researchers can process massive datasets across hundreds of variables in real time, using AI to guide investigations and optimize parameters automatically.
With converged AI and HPC, scientists can pair high-resolution simulation with AI models that draw on decades of observational data in real time. Forecasts sharpen, warnings come sooner, and atmospheric research iterates at unprecedented speed.
Federal researchers can advance critical challenges across autonomous systems development, cybersecurity, energy innovation and healthcare. Integrating HPC simulation and modeling data with AI, agencies accelerate experimental cycles and turn complex analysis into actionable insights within secure, compliant environments.
Transform how federal agencies leverage supercomputing for critical operations. AWS enables your teams to unify fragmented supply chain, infrastructure, and environmental data into a comprehensive operational picture.
Explore resources
Every mission can start today
Customers do not all start in the same place. Some have mature HPC and AI teams, others are running their first proof of concept. Start small – test a model or run a simulation – then scale as requirements grow. Programs like AWS Training and Certification and AWS Workforce Development help teams build the skills needed to retool and transform operations.
The first step is a conversation.
Start with your mission use case
- Engage your AWS account team to schedule a working-backwards planning session for your priority mission workloads.
- Gain access to AI and HPC Supercomputing Pods on AWS.
- Conduct a proof of concept aligned to your deployment timeline.
- Complete a TCO analysis for your production workloads.
- Ask about potential credits programs to get started.