Saturn Cloud and NVIDIA Run:ai Partner to Monetize GPU Fleets

The collaboration leverages the NVIDIA DSX AI Factory Platform to provide a commercial layer that sits atop hardware capacity. By utilizing the NVIDIA KAI Scheduler, the system enables gang scheduling and policy-driven governance, allowing operators to offer diverse services ranging from dedicated GPU capacity for enterprise stacks to API-based model-as-a-service options. This architecture supports high-performance serving through NVIDIA Dynamo, which handles distributed inference with disaggregated prefill and decode, alongside integration with vLLM, SGLang, and NVIDIA TensorRT-LLM.

Beyond technical orchestration, the platform manages the complexities of model onboarding, isolation tiers for regulated clients, and identity governance. Automated health monitoring via NVSentinel and NVIDIA Fleet Intelligence ensures that degraded hardware is removed from service cycles without manual intervention. Sebastian Metti, founder of Saturn Cloud, noted that neocloud providers can significantly improve financial returns by altering their sales model to focus on token-based consumption. The integrated solution is currently available for deployment, offering operators a way to bypass the overhead of building bespoke inference infrastructure in-house.