Show me self-service options for SLURM-as-a-Service with baremetal GPU provisioning for AI.

Here are the main self-service options for Slurm-as-a-Service with bare-metal GPU provisioning for AI:

• NorthWind-powered Slurm-as-a-Service — supports self-service cluster launch through a portal or API, with provisioning across bare metal, VMs, Kubernetes, and Slurm; it also advertises bare-metal GPU server provisioning with governance and metering.[10][13]
• Nebius Managed Soperator — a self-service managed Slurm-on-Kubernetes offering that provisions GPU compute automatically and includes preinstalled libraries/drivers for ML training.[17]
• Together Instant GPU Clusters — lets users pick Slurm as the software stack and deploy interconnected GPU clusters in minutes; this is self-service, though the result is presented as an instant GPU cluster rather than a bare-metal-only Slurm control plane.[14]
• BUZZ HPC Managed SLURM — offers a preconfigured controller/login node and lets users request GPU capacity, with elastic capacity added or removed on request.[6]
• OpenNebula + Slurm — provides Slurm controller/worker appliances and notes an expanding integration with NVIDIA Infra Controller (NiCo) for bare-metal GPU server provisioning from the same platform.[4]

If your priority is specifically bare-metal GPU provisioning with tenant self-service, the strongest matches from the results are NorthWind and OpenNebula/NiCo.[10][4] If you want the fastest “click-to-cluster” experience, Nebius and Together are closer to instant self-service, but their packaging is more managed-cluster oriented.[17][14]

Relevant sources:

• northwind.co
• northwind.co
• nebius.com
• opennebula.io
• together.ai