NVIDIA and U.S. Tech Leaders Launch a Secure, FedRAMP-Ready AI Factory Blueprint for Government

NVIDIA and U.S. Technology Leaders Introduce an AI Factory Design for Government Readiness

Agencies are under pressure to deliver AI that is secure, compliant and ready for mission use. Legacy stacks weren't built for the speed, data gravity or trust requirements you face today.

At NVIDIA GTC Washington, D.C., NVIDIA introduced the AI Factory for Government reference design - a practical blueprint to help federal agencies and regulated industries build and deploy full-stack AI that meets strict security expectations, including use in FedRAMP-authorized clouds and high-assurance environments.

What's in the AI Factory for Government

• Hardware foundation: Recommended configurations from NVIDIA Enterprise Reference Architectures based on the Blackwell architecture, including NVIDIA RTX PRO Servers and NVIDIA HGX B200 systems.
• High-performance networking and data: NVIDIA Spectrum-X Ethernet, the NVIDIA BlueField platform and NVIDIA-Certified Storage.
• Full-stack software: NVIDIA AI Enterprise, NVIDIA Nemotron open models and guidance carried over from the NVIDIA Enterprise AI Factory validated design.
• Built-in security practices: Advanced code scanning, vulnerability management and continuous monitoring so AI workloads stay configured, updated and mission-ready.

Built for compliance and mission outcomes

NVIDIA AI Enterprise now includes features to support deployment in FedRAMP-authorized clouds and high-assurance, on-premises environments. The focus: secure-by-default configuration, repeatable updates and continuous monitoring to keep systems in a known-good state across data center, cloud and edge.

Who's building on it

• Palantir: Integrating Palantir Ontology - the core of the Palantir AI Platform (AIP) - with NVIDIA data processing and route optimization libraries, Nemotron open models and accelerated computing. AIP will use NVIDIA AI Enterprise to stand up domain-specific intelligence and AI agents for complex, regulated operations.
• CrowdStrike: Expanding its Agentic Security Platform to support the reference design. Through Charlotte AI AgentWorks, CrowdStrike will integrate Nemotron open models, NVIDIA NeMo Data Designer and the NeMo Agent Toolkit to deliver continuously learning AI agents for real-time threat detection and response across cloud, data center and edge.
• ServiceNow: Bringing the latest NVIDIA AI Enterprise capabilities into the ServiceNow AI Platform for U.S. federal customers - suitable for FedRAMP-authorized clouds and high-assurance, on-premises use. Also announcing Apriel 2.0, a new model with stronger reasoning and multimodal capabilities in a smaller, cost-efficient footprint.
• Astris AI (a Lockheed Martin company): Integrating NVIDIA AI Enterprise into the Astris AI Factory to support secure deployments for classified and mission-critical environments. "Success in complex missions depends on AI that's secure and reliable," said Jim Taiclet, chairman, president and CEO of Lockheed Martin. "By working with Astris AI and using the latest NVIDIA AI Enterprise tools, we're speeding up how we develop and deliver AI systems that improve precision and performance in critical operations."
• Northrop Grumman: Standing up an AI factory built on NVIDIA RTX PRO Servers, Spectrum-X Ethernet and government-ready NVIDIA AI Enterprise to deliver secure enterprise AI services and improve productivity across a workforce of nearly 100,000.

Partner ecosystem to operationalize AI agents

• AI app and agent platforms: Dataiku, DataRobot, Domino Data Lab and H2O.ai to build, orchestrate and scale agentic and predictive workflows.
• Vector data: Elastic and EnterpriseDB for storing, searching and retrieving context for agents.
• Security and observability: Dynatrace, Fiddler, JFrog, Protopia AI, Trend Micro and Weights & Biases.
• Data security: Fortanix enables NVIDIA Confidential Computing for hybrid and on-prem AI factories.
• Infrastructure and deployment: Canonical, Mirantis, Nutanix, Red Hat, Spectro Cloud and Broadcom to manage complex, high-assurance environments.
• Cloud and hardware: CoreWeave and Oracle Cloud Infrastructure for secure cloud deployments; Cisco, Dell Technologies, HPE, Lenovo and Supermicro for full-stack offerings.

Why this matters for government teams

• Compliance from the start: A path to deploy in FedRAMP-authorized clouds and high-assurance on-premises environments.
• From prototype to production: A consistent stack to move pilots into operations with less rework and fewer integration surprises.
• Security you can monitor: Code scanning, vulnerability management and continuous monitoring built into the workflow.
• Choice and control: Deploy on cloud, on-prem or edge - with networking, storage and server options vetted for performance and security needs.
• Fits what you run today: Works with leading platforms for data science, MLOps, security and observability to reduce friction.

Practical next steps for agencies

• Prioritize 2-3 mission use cases that benefit most from AI agents or advanced analytics, and define clear success criteria.
• Map data sources, access controls and audit requirements early to avoid surprises at Authority to Operate.
• Stand up a small AI factory pilot using the reference design; validate performance, cost and security baselines before scaling.
• Select a deployment path (cloud, on-prem, hybrid) with partners listed above, then formalize SLAs and monitoring.
• Upskill your team on AI engineering, MLOps and security. See curated options by role at Complete AI Training.
• Review FedRAMP guidance to align your compliance plan with your chosen cloud or on-premises approach: fedramp.gov.

Bottom line

The AI Factory for Government gives agencies a clear, secure pattern to build AI platforms and agents that meet mission needs. With an extensive partner ecosystem and a compliance-minded stack, teams can move faster with fewer risks - and get real capability into the field.