The full NVIDIA Run:ai platform is now open source. NVIDIA Run:ai is a component of NVIDIA DSX OS, the open-source, modular software for building, operating, and scaling AI factory infrastructure within the NVIDIA DSX AI Factory Platform. Its Kubernetes-native AI workload and GPU orchestration help organizations maximize infrastructure utilization and workload throughput. AI cloud operators can now deploy, operate, and integrate the platform without a commercial NVIDIA Run:ai software license.
Together, NVIDIA Run:ai and Saturn Cloud give AI cloud operators a streamlined path from GPU infrastructure to differentiated, customer-facing services. NVIDIA Run:ai orchestrates AI workloads and governs how GPU resources are allocated and shared. Saturn Cloud adds automated deployment and the service-delivery layer, including branded access, shared and dedicated service tiers, model management, usage metering, and billing. Together, they help operators improve GPU utilization while delivering scalable AI services under their own brand.
The full NVIDIA Run:ai platform, now open source
NVIDIA has open sourced the integrated NVIDIA Run:ai platform, including its control plane, user interfaces, APIs, workload-management, scheduling, and resource-governance capabilities. KAI Scheduler, open source under Apache 2.0 since April 2025, remains the platform’s scheduling core. NVIDIA Run:ai also works with other open NVIDIA DSX OS components, including NVIDIA Grove and NVIDIA Dynamo.
Operators can deploy the complete NVIDIA Run:ai platform and integrate it with the broader NVIDIA ecosystem, including NVIDIA Dynamo for distributed inference serving. Together, these technologies provide an open, modular foundation for orchestrating shared GPU infrastructure and serving distributed inference, with no commercial NVIDIA Run:ai software license required.
How Saturn Cloud leverages the open NVIDIA Run:ai stack
NVIDIA Run:ai addresses the complex resource-management challenges involved in sharing physical infrastructure across AI workloads and teams. At the platform level, projects, departments, quotas, and policies govern how capacity is assigned and shared. NVIDIA KAI Scheduler handles workload placement through capabilities such as gang scheduling, hierarchical queues, fair-share scheduling, preemption, and fractional-GPU-aware placement. NVIDIA Grove describes the structure, dependencies, scaling behavior, and topology requirements of multi-component workloads, enabling schedulers to make topology-aware placement decisions.
Complementing NVIDIA Run:ai, NVIDIA Dynamo manages multi-node inference serving with disaggregated prefill and decode across backend engines including vLLM, SGLang, and NVIDIA TensorRT-LLM. NVIDIA NVSentinel and NVIDIA Fleet Intelligence provide fleet-health monitoring, fault detection, and remediation capabilities.
Where Saturn Cloud fits
Saturn Cloud uses NVIDIA Run:ai as the workload and GPU orchestration foundation and adds everything an operator needs to sell what the fleet produces: multi-tenant isolation on shared hardware, per-token metering and billing, model management, API access, and enterprise identity and governance. Integrated with the NVIDIA DSX AI Factory Platform, Saturn Cloud deploys the full open-source NVIDIA Run:ai stack in one click, so operators can turn raw compute into a branded, multi-tenant AI factory without building the surrounding platform in-house.
Underneath, this is one integrated offering. NVIDIA Run:ai orchestrates AI workloads and GPU resources, NVIDIA Dynamo provides distributed inference serving, and Saturn Cloud runs the commercial layer on top. From there, operators can sell the same fleet in multiple ways: dedicated GPU capacity by the hour, per-token model-as-a-service for customers who need a ready-to-call endpoint, and managed fine-tuning for teams adapting their own models, each metered and billed through the same system.
Capabilities of the integrated platform
| Capability | What it does |
|---|---|
| One-click NVIDIA Run:ai deployment | Stands up the full open-source NVIDIA Run:ai platform on an operator’s own cluster, enterprise-ready, without assembling it by hand. |
| Multi-tenant isolation and security | Saturn Cloud provides shared and dedicated isolation tiers, with enterprise identity (SSO), role-based access control (RBAC), and compliance governance for enterprise buyers. |
| Per-token metering and monetization | Tracks consumption down to individual requests and output tokens, so operators can bill for model-as-a-service endpoints. |
| Automated model onboarding | Handles the deployment and lifecycle of open-weight models across shared GPU fleets. |
Getting started
The NVIDIA Run:ai-integrated Saturn Cloud platform is available now for neoclouds, telcos, and enterprise AI factory operators.
To explore the open-source NVIDIA Run:ai platform, visit the NVIDIA Run:ai repository in the NVIDIA DSX AI Factory organization on GitHub. To deploy NVIDIA Run:ai with Saturn Cloud on your own infrastructure, contact us.
