NVIDIA announced on October 13 that it is making the full Run:ai platform available as open source software as part of the NVIDIA DSX AI Factory Platform. The company positions the release as an open foundation for developers, cloud providers and infrastructure operators building systems to coordinate AI workloads.

Run:ai focuses on sharing and orchestrating GPU resources across workloads that can include data processing, model training and inference. NVIDIA identifies four modular components that organizations can use independently or together. KAI Scheduler provides GPU-aware Kubernetes scheduling. GPU Fractioning Operator supports CUDA-native GPU sharing. KAI Resource Management addresses organizational quotas and hierarchy. Workload Map provides a common workload-management contract.

The announcement also identifies ecosystem partners working on integrations, including Crusoe, Mirantis, Nebius, OpenNebula Systems, Rafay, Saturn Cloud, Spectro Cloud and vCluster. NVIDIA directs developers to the platform repository and component projects to inspect the implementation.

For buyers and engineers comparing AI infrastructure tools, the useful development is access to code and modular building blocks rather than a new benchmark claim. Whether a particular environment benefits from the release will depend on its scheduling requirements, operational constraints and the maturity of each integration. NVIDIA's technical announcement is linked below.