Red Hat, NVIDIA expand agentic AI collaboration
The companies are co-developing the OpenShell Reference Architecture and plan to fold it into Red Hat AI as a default platform feature.
Red Hat said on 22 October 2026 that it is expanding its collaboration with NVIDIA, aiming to deliver secured enterprise-grade agentic AI across open hybrid clouds through Red Hat AI Factory with NVIDIA.
As part of the work, the two companies are developing the OpenShell Reference Architecture, which Red Hat describes as a way to bring safety and control to enterprise agentic AI deployments. The architecture is based on OpenShell, an open-source agent sandbox runtime to which both Red Hat and NVIDIA are major contributors.
According to digitaltoday.co.kr, the design combines Model-as-a-Service access control, an AI Gateway, high-performance inference via vLLM and llm-d, NVIDIA Dynamo, NVIDIA NIM microservices and GPU orchestration technology. Red Hat and NVIDIA have also developed joint reference designs based on Red Hat AI Factory with NVIDIA.
The companies are working to integrate the OpenShell Reference Architecture as a default platform feature in Red Hat AI and Red Hat AI Factory with NVIDIA.
Red Hat added that it will support next-generation architectures including NVIDIA Vera Rubin on launch day, and that Red Hat Enterprise Linux for NVIDIA supports NVIDIA Vera CPUs.
Quick answers
What is the OpenShell Reference Architecture?
It is an architecture Red Hat and NVIDIA are developing on top of OpenShell, an open-source agent sandbox runtime, to add safety and control to enterprise agentic AI deployments.
Will the OpenShell Reference Architecture be part of Red Hat AI?
Red Hat says the two companies are working to integrate it as a default platform feature in Red Hat AI and Red Hat AI Factory with NVIDIA.
Which NVIDIA architectures will Red Hat support?
Red Hat said it will support next-generation architectures including NVIDIA Vera Rubin on launch day, while Red Hat Enterprise Linux for NVIDIA supports NVIDIA Vera CPUs.