The architecture is a deployment stack for building, containing, observing and enforcing AI agents across enterprise infrastructure and edge systems.
NVIDIA combines Vera Rubin infrastructure for model inference, Vera CPUs for agent harnesses and tool use, open-source OpenShell runtime software, and BlueField-4 DPUs running the NVIDIA Sentry security framework. NVIDIA says Sentry provides continuous out-of-band monitoring and in-silicon policy enforcement; if an agent moves outside its software boundary, it can be quarantined and stopped in milliseconds. The company also claims BlueField and Sentry can raise agent capacity by up to 20%.
The reference architecture is listed as running on Canonical, Red Hat and Microsoft Windows with Microsoft Execution Containers, and extending to RTX Spark, DGX Spark and DGX Station at the edge. NVIDIA names Anthropic, AWS, Google Cloud, Microsoft, OpenAI, SAP and ServiceNow among organizations working with it.
The announcement does not provide independent performance tests, deployment pricing, hardware availability details, or evidence that the listed integrations are generally available in production.