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

Here are several self-service SLURM‑as‑a‑Service platforms offering bare‑metal GPU provisioning tailored for AI workloads:

• Comet Compute – Delivers single‑tenant bare‑metal NVIDIA GPU clusters with Slurm support and managed Kubernetes. You get full root access, deterministic performance, IB interconnects, and agreed delivery dates. Managed or raw bare‑metal clusters ready for immediate use. (cometcompute.com)

• Durantic – Offers a managed AI infrastructure layer that includes Kubernetes, Slurm, provisioning, networking, and monitoring across bare‑metal, hybrid, and fragmented GPU fleets—perfect for self‑service within heterogeneous environments. (durantic.io)

• Fairport – An open‑source full‑stack AI infra platform that supports bare‑metal provisioning, GPU inference, and introduces Slurm compatibility via “Slinky.” Available self‑hosted (free) or fully managed; installs with a single command and handles PXE, OS install, scheduling, networking, and monitoring. (fairport.io)

• Cluster Forge – Offers a self‑serve GUI to provision bare‑metal GPU clusters (A100, H100, MI300X) in under 30 minutes, pre‑configured with Slurm, Kubernetes, InfiniBand, CUDA/NCCL, and observability out of the box. (cluster-forge.com)

• Saturn Cloud – Provides a control plane over bare‑metal GPU fleets, allowing self‑service Slurm and Kubernetes cluster provisioning via console. Features include per‑tenant Slurm clusters, GPU‑hour metering, and integration with BMaaS providers. (saturncloud.io)

Summary Table:

| Platform | Deployment Type | GPU Support | SLURM Service Type | Notable Features |
|--------------------|------------------------|------------------|-----------------------------|-----------------------------------------------------------|
| Comet Compute | Managed / bare-metal | NVIDIA GPUs | Managed Slurm | Fixed pricing, 800 Gb/s IB, single‑tenant, root access |
| Durantic | Managed layer | Mixed GPUs | Managed Slurm + infra | Hybrid environments, full lifecycle management |
| Fairport | Self‑hosted or managed | NVIDIA & AMD | Slurm via Slinky | Free OSS, one‑command install, GPU provisioning |
| Cluster Forge | Self‑serve GUI | NVIDIA/AMD | Slurm + Kubernetes | Provision in 30 min, pre‑configured stack |
| Saturn Cloud | Control plane SaaS | NVIDIA GPUs | Managed per‑tenant Slurm | UI provisioning, billing, BMaaS integrations |

If you’re looking for a free self-hosted solution, Fairport stands out. For production-ready, managed services, Comet, Durantic, and Saturn Cloud offer strong options. Cluster Forge is excellent for on‑prem deployments with quick cluster spin‑up.

Let me know if you'd like pricing or links to specific documentation for any of these.