NVIDIA Contributes OpenShell to CNCF and Launches Kubernetes Tools to Boost AI Infrastructure

Image: NVIDIA · Source

NVIDIA deepens its open source engagement by donating OpenShell, a runtime securing autonomous AI agents, to the Cloud Native Computing Foundation and unveiling Kubernetes-based tools to improve AI factory scalability and security.

NVIDIA has significantly expanded its commitment to open source with new contributions and tools aimed at enhancing AI infrastructure at scale. At KubeCon + CloudNativeCon, a premier event for Kubernetes and cloud-native developers, NVIDIA announced it is donating its OpenShell project to the Cloud Native Computing Foundation (CNCF), which is part of the Linux Foundation. OpenShell is an open source runtime designed to secure and govern autonomous AI agents, enabling deterministic control of their activities and integration with existing enterprise security systems.

OpenShell's donation to CNCF will allow the broader community to help evolve and enhance security measures for autonomous agents, which are increasingly critical in AI workloads. Linux Foundation CEO Jim Zemlin highlighted the importance of shared infrastructure to ensure agents operate securely in cloud-native applications.

In addition, NVIDIA is contributing to Agent Substrate, a Google-initiated CNCF Sandbox project that offers a framework for operating long-running AI agent workloads on Kubernetes efficiently. This includes pausing idle agents to reclaim compute resources, thus enabling scalable deployment of AI agents.

To further simplify managing GPU infrastructure for AI, NVIDIA introduced new components to its DSX AI Factory Platform: NVIDIA Kubernetes Engine (NVKE), NVIDIA Kubernetes Recipes (NVKR), and NVIDIA Snapshot. NVKE and NVKR provide open source solutions for creating and managing Kubernetes clusters with GPU support, currently available in public preview. These tools have garnered support from partners including Mirantis, SK Telecom, and others, promising broader ecosystem integration.

NVIDIA Snapshot addresses the latency traditionally involved in scaling inference workloads by restoring checkpointed replicas in seconds without requiring model reload or warmup. Early adopters, such as Crusoe and Saturn Cloud, have tested Snapshot successfully, achieving startup time reductions from minutes to seconds.

Together, these initiatives mark NVIDIA's push to cement Kubernetes as the foundational platform for scalable, secure AI operations, enabling organizations to build AI factories with greater efficiency and enhanced governance over autonomous agents. These projects underscore the growing intersection of AI and cloud-native technologies, aiming to support the next generation of AI infrastructure worldwide.