NVIDIA Run:ai Goes Fully Open Source
NVIDIA has open-sourced the complete Run:ai GPU orchestration platform for Kubernetes, including its control plane, cluster services, installer, CLI, and MCP server. The code is available under Apache 2.0 in NVIDIA’s DSX AI Factory GitHub organization.
This turns Run:ai from a proprietary GPU-management product into a serious infrastructure building block for AI clouds and enterprise clusters.
- –The release goes beyond KAI Scheduler, exposing the full multi-cluster control plane and Kubernetes operators.
- –GPU-aware scheduling, fractional allocation, quotas, preemption, and observability target the biggest operational bottleneck in AI: underused compute.
- –A free community product lowers adoption barriers, while NVIDIA can still monetize enterprise support and managed offerings.[CTech](https://www.calcalistech.com/ctechnews/article/jqdtr2997)
- –The NGC catalog remains a practical distribution channel for self-hosted deployments.[NVIDIA NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/runai/collections/run-ai-self-hosted/-)
DISCOVERED
1h ago
2026-10-13
PUBLISHED
2h ago
2026-10-13
RELEVANCE
AUTHOR
AIQuanting