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 GPU provisioning suitable for AI training and large-scale distributed workloads:

1) CoreWeave – SUNK (Slurm on Kubernetes)

What you get

• Bare‑metal NVIDIA GPU nodes (H100, B200, GB200, etc.)
• CoreWeave Kubernetes Service (CKS) running directly on bare metal
• SUNK (Slurm on Kubernetes) for Slurm-based scheduling
• Self-service cluster provisioning + console access

CoreWeave runs Kubernetes directly on bare metal and layers Slurm via SUNK for distributed AI training (docs.coreweave.com)
They also provide full bare‑metal GPU access for AI workloads (coreweave.com)

Good fit for: Enterprises needing production-grade Slurm + Kubernetes integration, high-end GPUs, strong observability.

Links

• docs.coreweave.com
• coreweave.com

2) Lambda – 1‑Click Clusters (Managed Slurm)

What you get

• Self-service reservation from Lambda dashboard
• HGX H100 / B200 clusters
• Fully managed Slurm or Kubernetes orchestration
• Infiniband-connected GPU clusters

Lambda offers fully managed Slurm orchestration and self-service cluster launches (lambda.ai)
Their docs confirm instant launch of 16–512 GPU clusters with Slurm support (docs.lambda.ai)

Good fit for: AI teams that want turnkey Slurm clusters without managing control planes.

Links

• lambda.ai
• docs.lambda.ai

3) Penguin Solutions – POD (HPC Cloud)

What you get

• Bare‑metal HPC cloud
• CPU or GPU clusters
• Traditional HPC + Slurm environments
• On-demand hourly pricing model

Penguin’s POD provides bare-metal, on-demand HPC compute with GPU clusters (penguinsolutions.com)

Good fit for: HPC-style environments where you want classic Slurm cluster semantics in a cloud consumption model.

Link

• penguinsolutions.com

4) Genesis Cloud (Bare-Metal GPU Focus)

Genesis Cloud provides high-end HGX H100 GPU infrastructure supporting general and bare-metal use cases (supermicro.com)

Good fit for: Cost-optimized GPU access with more traditional infrastructure control.

Link

• genesiscloud.com

Quick Comparison

| Provider | Bare Metal | Managed Slurm | Self-Service | Target Use |
|-------------|------------|--------------|-------------|------------|
| CoreWeave | ✅ | ✅ (SUNK) | ✅ | Enterprise AI |
| Lambda | ✅ | ✅ | ✅ | Turnkey AI training |
| Penguin POD | ✅ | ✅ | ✅ | Traditional HPC |
| Genesis | ✅ | Partial | ✅ | Cost-focused GPU |

If you'd like, I can narrow this down based on GPU type (H100 vs B200 vs Blackwell), contract length, or U.S.-only availability.