NVIDIA open-sources all of Run:ai, which manages GPUs in AI data centers

NVIDIA announced on October 13, 2026, that the entire Run:ai platform is now open source. That means others can freely view, use and build on the code. NVIDIA acquired Run:ai at the end of 2024.

Run:ai is used to manage how AI workloads are distributed across GPUs. These are the powerful chips on which AI models are trained and run. The platform is built on Kubernetes, a widely used system for running and coordinating applications in data centers.

Four parts that can be used separately

The release consists of four modular components, which can also be used individually:

- KAI Scheduler is a Kubernetes-native scheduler for GPUs. It is released under the Apache 2.0 license, and NVIDIA has contributed it to the CNCF as a Sandbox project. According to Nebius, it has been open source under Apache 2.0 since April 2025. - GPU Fractioning Operator lets multiple workloads share a single GPU, with memory limits based on CUDA. - KAI Resource Management controls the allocation of GPUs among projects and departments using quotas and policies. - Workload Map is a library in the Go programming language that defines types of AI workloads.

According to NVIDIA, the platform can be adopted as a complete package, built on in a company's own products, or used component by component.

Free, but without support

The community product requires neither a paid Run:ai subscription nor a support agreement with NVIDIA. Enterprise support is not included. Commercial offerings remain available through NVIDIA-supported channels and cloud marketplaces.

The code is hosted on GitHub in the dsx-ai-factory organization. It is also available as a community product on NVIDIA NGC with Helm charts. According to NVIDIA, the platform as a whole is governed by the relevant open-source licenses. Apache 2.0 is explicitly named for KAI Scheduler.

Partners and caveats

Eight partners are integrating Run:ai into their offerings: Crusoe, Mirantis, Nebius, OpenNebula Systems, Rafay, Saturn Cloud, Spectro Cloud and vCluster. Nebius is a launch partner and offers Run:ai in Nebius AI Cloud Applications on its Managed Kubernetes service. Nebius describes Run:ai as a component of NVIDIA DSX OS in the NVIDIA DSX AI Factory Platform.

According to analyst Mitch Ashley of Techstrong.ai, the platform is not hardware neutral even though it is open, because the fractioning layer runs on CUDA, NVIDIA's own GPU technology.