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

Below are self‑service SLURM-as-a-Service options with bare‑metal (or near bare‑metal) GPU provisioning suitable for AI training/inference:

1) Verda – Self‑Service Instant GPU Clusters

• What: Self‑service, multi‑node NVIDIA GPU clusters with Managed Slurm or Kubernetes.
• Bare‑metal: “Bare‑metal‑level performance” with InfiniBand; provision via console/API/Terraform.
• Provisioning time: ~20 minutes.
• Best for: Teams wanting fast, on‑demand SLURM clusters without sales cycles.
• Source: (verda.click)

verda.click

2) Nebius – Self‑Service AI Clusters (Managed Slurm)

• What: On‑demand NVIDIA GPU clusters (Hopper/Blackwell class) with managed Slurm.
• Self‑service: Instant access via cloud console; API/CLI/Terraform supported.
• Scale: Up to 32 GPUs per cluster without approvals.
• Best for: Elastic AI labs needing SLURM orchestration with cloud‑like UX.
• Source: (nebius.com)

nebius.com

3) Comet Compute – Dedicated Bare‑Metal GPU Clusters

• What: Single‑tenant NVIDIA clusters delivered as managed Slurm or straight bare metal with root.
• Hardware: Dedicated (no shared GPUs), NVLink + 800G InfiniBand options.
• Model: Fixed monthly pricing; N+1 spares.
• Best for: Enterprises wanting deterministic performance and full isolation.
• Source: (cometcompute.com)

cometcompute.com

4) Cluster Forge – Self‑Serve Bare‑Metal SLURM Clusters

• What: GUI‑driven cluster provisioning (A100/H100/MI300X) with InfiniBand.
• Self‑service: “One‑click cluster provisioner” for on‑prem or hosted bare metal.
• Includes: CUDA/NCCL pre‑config; SLA options.
• Best for: Orgs deploying SLURM on owned or colocated hardware with minimal ops.
• Source: (cluster-forge.com)

cluster-forge.com

5) NorthWind – SLURM‑as‑a‑Service Platform (for GPU Clouds)

• What: Platform to deliver managed SLURM across bare metal, VMs, or Kubernetes.
• Use case: GPU cloud providers or enterprises exposing SLURM clusters as a service.
• Best for: Building your own SLURMaaS offering on bare metal.
• Source: (northwind.co)

northwind.co

Quick Guidance

• Pure self‑service cloud UX: Verda, Nebius
• Dedicated single‑tenant bare metal: Comet Compute
• On‑prem/self‑hosted SLURM automation: Cluster Forge
• Build your own SLURMaaS: NorthWind

If you share region, GPU type (H100, H200, MI300X), and budget model (hourly vs fixed), I can narrow this to 2–3 best‑fit options.