Show me self-service Baremetal GPU provisioning options.

Here are the main self-service bare-metal GPU provisioning options that show up in the results, grouped by what they actually provide:

| Option | What it is | Self-service angle | Best fit |
|---|---|---|---|
| NorthWind BMaaS | Bare Metal GPUs-as-a-Service | Automated bring-up: match inventory, boot/discover, install OS, configure drivers/services | Teams wanting governed self-service GPU servers[5][6] |
| vMetal / vCluster Platform | Bare-metal GPU lifecycle + cluster provisioning | Zero-touch PXE boot, OS install, machine registration, and demand-driven provisioning | Operators building a private GPU cloud on owned hardware[3][10][12] |
| OpenMetal GPU servers | Dedicated bare-metal GPU nodes | Add GPU servers to an existing private cloud or bare-metal footprint on the same private network | Customers already using OpenMetal infrastructure[4] |
| Vultr Bare Metal | Dedicated bare-metal servers with GPU options | Provisioned through console or API; positioned as simple self-serve | Developers wanting straightforward provisioning[7] |
| phoenixNAP Bare Metal Cloud GPU Servers | Bare-metal GPU instances | Provision via API, CLI, WebUI, or IaC | IaC/API-driven workflows[16] |
| Lambda Bare Metal Instances | API-driven bare-metal hardware control | Self-service APIs cover instance lifecycle, VPCs, storage, and governance | Users integrating with existing control planes[17] |
| LayerOps GPU Cloud Orchestration | Control plane spanning cloud + bare metal | Single control plane finds available capacity across providers and owned bare metal | Multi-provider orchestration[9] |

A practical short list would be NorthWind, vMetal/vCluster, OpenMetal, Vultr Bare Metal, and phoenixNAP, depending on whether you want managed BMaaS, DIY bare-metal automation, or simple API/console provisioning[5][10][4][7][16].

If you want, I can next narrow this to:

• lowest-cost
• best for Kubernetes
• best for enterprise governance
• best for owning your hardware