NVIDIA Run:ai Is Now Open Source as Part of NVIDIA DSX AI Factory Platform

NVIDIA Makes Run:ai Open Source to Standardize GPU Orchestration

NVIDIA has released its entire Run:ai platform as open source software, integrating it as a foundational component of the NVIDIA DSX AI Factory Platform. The move provides developers, customers, and ecosystem partners with direct access to production-grade tools for AI workload and GPU orchestration on Kubernetes. By establishing an open standard, NVIDIA aims to accelerate how organizations manage shared AI infrastructure, improving GPU utilization and workload throughput across the industry.

Modular Components and Adoption

The open source release is structured for flexible adoption, allowing teams to deploy the entire platform or integrate its modular components independently. The source code is now available in the NVIDIA DSX AI Factory GitHub organization, with a Helm chart for installation also accessible on NVIDIA NGC as a Community Product. This release strategically retires proprietary implementations in favor of open, native capabilities, such as using CUDA 13.4 for GPU fractioning.

• KAI Scheduler: A CNCF Sandbox project for GPU-aware Kubernetes scheduling, handling gang scheduling, fair-share quota allocation, and topology awareness.
• GPU Fractioning Operator: An operator that enables CUDA-native GPU sharing on Kubernetes, allowing multiple workloads to use fractions of a GPU with hard memory isolation.
• KAI Resource Management: An organizational layer for mapping resources to teams, managing quotas across a hierarchy, and defining node pools for heterogeneous hardware.
• Workload Map: A component that provides a consistent contract for managing workloads across the platform.

Ecosystem and Market Impact

This decision directly enables NVIDIA DSX ecosystem partners to build and ship differentiated AI factory solutions more quickly. Companies including Crusoe, Mirantis, Nebius, Rafay, and Saturn Cloud are already integrating the open source platform into their managed Kubernetes, AI cloud, and MLOps offerings. By providing the core orchestration layer, NVIDIA encourages partners to innovate on top of its stack rather than building competing solutions from scratch. While the open source version is available without enterprise support, NVIDIA will continue to offer commercially supported versions through NVIDIA AI Enterprise and NVIDIA Mission Control.

By open-sourcing Run:ai, NVIDIA is effectively setting the architectural standard for GPU orchestration on Kubernetes within its DSX ecosystem. This move encourages partners to build on top of NVIDIA's foundation rather than alongside it, reinforcing the company's software moat around its dominant hardware position.