NVIDIA introduces OpenShell reference architecture for agents
17 companies named across 6 categories, one of 619 articles referencing Nvidia. Previous coverage: NVIDIA expands Isaac robotics platform with new humanoid (Oct 2026).
Companies mentioned
Best suited for
- Seniority
- Architect
- Job function
- Cybersecurity / Security Architect
- Persona
- Platform / Infrastructure Architect
- Buyer role
- Architect / Technical Evaluator
- Decision stage
- Which to buy
- Adoption curve
- Early Adopters
- Technology maturity
- Accelerated Attention
- Industry
- Information Technology / AI Infrastructure / AI Platforms, MLOps & Operations / AI Governance Platforms
Our classification, not the publisher's statement. Best suited for, not only for.
NVIDIA has introduced the OpenShell Reference Architecture at GTC Berlin, a framework for building, containing, observing and enforcing AI agents on enterprise infrastructure and edge systems. Its stated purpose is to keep agents operating within defined boundaries, with monitoring and control in place.
The framework adds enterprise governance, isolation and security to agent deployments. It builds on NVIDIA OpenShell secure runtime software and extends it into full-stack, industrial-scale infrastructure for agent development and operations. Its stated goal is safe deployment of agents in production.
The architecture combines NVIDIA Vera Rubin infrastructure for running enterprise workloads, including closed and open models built with NVIDIA Nemotron, with NVIDIA Vera CPUs that operate agent harnesses. NVIDIA BlueField DPUs run the NVIDIA Sentry unified security software framework, which provides continuous out-of-band monitoring of agent behavior and in-silicon policy enforcement.
NVIDIA states that Vera Rubin NVL72 systems deliver 30x higher throughput per megawatt for agentic workloads and that Vera CPUs deliver twice the performance of x86 processors with greater energy efficiency. Sentry on BlueField-4 quarantines and stops an agent that attempts to move outside its software boundary within milliseconds, and it boosts agent capacity by up to 20%.
The architecture runs on Canonical, Red Hat and SUSE Linux and Kubernetes infrastructure, as well as Microsoft Windows through Microsoft Execution Containers. It can run on Amazon Web Services, Google Cloud and Microsoft Azure, and it extends to edge systems through NVIDIA RTX Spark, NVIDIA DGX Spark and NVIDIA DGX Station. “AI capability and safety must advance together,” said Jensen Huang, founder and CEO of NVIDIA.
Partners including Anthropic, AWS, Cisco, CrowdStrike, Google Cloud, IBM, Microsoft, OpenAI, Palo Alto Networks, SAP, ServiceNow and Siemens are working with the architecture, and Dell Technologies, HPE and Lenovo are building infrastructure with it. The architecture is now available. Forward-looking statements in the release concerning adoption, performance and availability are subject to risks and uncertainties.
Press release, provided by Globe Newswire on behalf of Nvidia. Read the original.