NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure

NVIDIA has announced the donation of NVIDIA OpenShell, an open source runtime designed to secure and govern autonomous AI agents, to the Cloud Native Computing Foundation (CNCF). As a CNCF Sandbox project, OpenShell will be collaboratively developed by the community, including security professionals, to enhance its safeguards and governance capabilities for agent environments. The move aims to address the growing need for secure infrastructure as AI agents become more prevalent in cloud-native applications.

To complement OpenShell, NVIDIA is contributing to Agent Substrate, a Google-led CNCF Sandbox project focused on operating long-running agent workloads at scale on Kubernetes. The integration with NVIDIA Dynamo for model inference and KAI Scheduler for resource allocation will enable unified agent execution, inference serving, and cluster management. This collaboration seeks to optimize resource usage by suspending and resuming agent environments, reducing idle compute costs while maintaining performance.

NVIDIA also introduced new components to its DSX AI Factory Platform, including NVIDIA Kubernetes Engine (NVKE), NVIDIA Kubernetes Recipes (NVKR), and NVIDIA Snapshot. NVKE and NVKR provide open source tools for deploying and managing GPU-powered Kubernetes clusters, with support from multiple ecosystem partners. These tools automate cluster lifecycle management and offer version-pinned images to ensure stability and compatibility across AI infrastructure platforms.

NVIDIA Snapshot addresses the challenge of rapid inference scaling by restoring checkpointed workloads in seconds without model reloads or warmup delays. Early evaluations by design partners Crusoe and Saturn Cloud demonstrated significant improvements, with Crusoe reducing workload readiness time from eight minutes to 31 seconds and Saturn Cloud cutting vLLM startup times from up to 8.5 minutes to 34 seconds on NVIDIA HGX B300 systems.