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

NVIDIA announced at KubeCon + CloudNativeCon that it is contributing NVIDIA OpenShell, an open source runtime for securing and governing autonomous agents, to the Cloud Native Computing Foundation (CNCF), the Linux Foundation-affiliated nonprofit that hosts the conference. OpenShell enters the CNCF as a Sandbox project, which means it will be governed by the organizations building it into their products and strengthened by security professionals in the wider community. Jim Zemlin, CEO of the Linux Foundation, framed the move as giving developers a shared foundation for agent security as cloud-native applications increasingly rely on agents.
OpenShell is designed to control what AI agents are allowed to do. It applies deterministic controls to agent activity so teams can enforce policies as agents use tools, access inference services and execute workloads. Its extensible architecture lets organizations plug in their own compute, inference providers, credential storage and monitoring tools, so governance can be applied consistently across deployment environments while connecting to existing enterprise systems. According to NVIDIA, hundreds of organizations already use OpenShell, and the project has drawn tens of thousands of contributions in less than a year since launch. This suggests the software has meaningful production exposure, which matters for a security-sensitive runtime.
NVIDIA is also contributing to Agent Substrate, a Google-originated framework for running long-running agent workloads at scale on Kubernetes that has been accepted into the CNCF Sandbox. NVIDIA's work ensures Agent Substrate interoperates with NVIDIA Dynamo for model inference and KAI Scheduler for allocating shared cluster resources, tying together agent execution, inference serving and scheduling. The design preserves agent environments between tasks while releasing compute when agents are idle; by suspending and resuming environments across a shared pool, teams can scale agent workloads without holding dedicated compute for every session. NVIDIA and Google are additionally collaborating on an integration between OpenShell and Agent Substrate aimed at securing and governing agents across clouds and on-premises systems.
On the infrastructure side, NVIDIA announced three new open source components of the NVIDIA DSX AI Factory Platform. NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR), both now in public preview, provide building blocks for delivering managed Kubernetes services on GPU infrastructure. They automate cluster lifecycle management and ship signed, version-pinned images. NVIDIA says Mirantis, Netris, OpenNebula Systems, Rafay, Spectro Cloud and vCluster, along with Firmus and SK Telecom, have aligned with NVIDIA on integration plans to bring NVKE and NVKR into their offerings.
The third component, NVIDIA Snapshot, targets the delay of bringing inference capacity online. Loading models and warming up workloads can slow scaling and recovery; Snapshot is a modular component that restores checkpointed replicas in seconds without model reload or warmup, supporting faster startup, recovery and elasticity for inference and agentic workloads on Kubernetes. Design partners Crusoe and Saturn Cloud evaluated it on their own workloads using NVIDIA HGX B300 systems. Crusoe, an upstream contributor to Snapshot, tested GLM-5.3 on Crusoe Cloud and cut the time for the workload to become ready from roughly eight minutes to 31 seconds. Saturn Cloud, a token factory platform, benchmarked Snapshot for scale-from-zero inference and reduced vLLM startup times from between two and 8.5 minutes to six and 34 seconds.
Taken together, the announcements push more of the agent-operations stack into vendor-neutral governance while keeping NVIDIA's inference and scheduling software central to how it runs. In practice, the OpenShell donation and the Agent Substrate work give platform teams open building blocks for the two hard problems of agent deployment: enforcing what agents may do, and keeping compute costs down while sessions sit idle. A likely trade-off is maturity: Sandbox status at the CNCF denotes early-stage projects, and the Snapshot figures come from two design partners on specific hardware and model configurations, so teams should expect to validate gains on their own stacks. The Kubernetes components extend NVIDIA's push to position Kubernetes as the control plane for GPU-heavy AI factories, a market where cloud providers and competing accelerator vendors also offer managed tooling.
Why it matters: Teams running agents and GPU inference on Kubernetes get open, community-governed options for policy enforcement, agent scaling and faster replica startup, rather than proprietary lock-in. Developers evaluating these tools can start with the NVKE and NVKR public previews and the Snapshot benchmarks, though adoption will depend on how quickly CNCF governance and ecosystem integrations mature, which is an inference rather than a stated outcome.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Kubernetes, the open source system that manages how applications run across thousands of AI factories, has become the foundation for AI, and the demands on it are only growing. To help meet them, NVIDIA is deepening its commitment to open source at KubeCon + CloudNativeCon, the flagship conference for the Kubernetes and cloud-native communities. NVIDIA […]