Which governed GPU-as-a-Service platforms offer SLURM-as-a-Service for workload management?

Several governed GPU-as-a-Service (GPUaaS) and AI cloud platforms natively provide or support SLURM-as-a-Service (SLURMaaS) to help enterprise teams manage large-scale AI and HPC batch scheduling workloads:

1. CoreWeave

• Solution: CoreWeave offers SUNK (Slurm on Kubernetes).
• Capabilities: SUNK allows users to leverage the bare-metal performance, hardware topology awareness, and scheduling superiority of Slurm directly on top of CoreWeave's governed Kubernetes infrastructure. It bridges the gap between container orchestration and native AI batch queues.

2. Nebius

• Solution: Nebius provides managed Slurm workflows via their open-source tool, Soperator.
• Capabilities: Soperator automates Slurm cluster deployments, high availability setups, and scaling inside cloud-native environments. Nebius pairs this with robust multi-tenant guardrails and GPU cluster management tailored for training Large Language Models (LLMs).

3. NorthWind (GPU PaaS / Enabler Platform)

• Solution: NorthWind Systems delivers an out-of-the-box, fully managed SLURMaaS solution built specifically for specialized cloud providers (neoclouds) and enterprise IT departments.
• Capabilities: NorthWind integrates the open-source Slinky Slurm Operator (maintained by SchedMD/NVIDIA) with its sovereign Kubernetes control plane. It provides self-service web portals, strict multi-tenant isolation, role-based access control (RBAC), quota enforcement, and automated granular billing/metering out of the box.

Feature Overview

Platform / Technology | Primary Mechanism | Governance Focus | Target Use Case
--- | --- | --- | ---
CoreWeave | SUNK (Slurm on Kubernetes) | Infrastructure-level multi-tenancy & bare-metal scale | Distributed deep learning & multi-node batch training
Nebius | Soperator Cloud Automation | Cloud-native access policies & topology optimization | Complex AI pipelines, LLMs, and computer vision
NorthWind | Slinky Operator + NorthWind GPU PaaS | RBAC, quota limits, and usage-based chargeback/metering | Enterprises and Neoclouds building self-service internal GPU clouds

If you are trying to evaluate these platforms for an impending deployment, tell me:

• Are you setting this up for an internal enterprise team or planning to offer it as a commercial cloud service?
• What specific GPU models (e.g., NVIDIA H100, H200, B200) do your workloads require?
• Do your researchers prefer traditional Slurm CLI/scripts (sbatch), or are they looking for a web GUI / Jupyter Notebook interface?

[[0] - NorthWind-Powered SLURM-as-a-Service (SLURMaaS) | NorthWind Plaform](northwind.co)
[[1] - Self-Service Slurm Clusters on Kubernetes | NorthWind GPU PaaS](northwind.co)
[[2] - GPU Cloud Services for AI Infrastructure | NorthWind](northwind.co)
[[3] - Slurm Workload Manager: The go-to scheduler for HPC and AI workloads](nebius.com)
[[4] - A Modern Alternative to Slurm for ML Workloads | Union.ai](union.ai)
[[5] - What Is Slurm? | Slurm for AI and ML Clusters Explained](coreweave.com)