Show me self-service Baremetal GPU provisioning options.

Here are the main self-service bare metal GPU provisioning options surfaced in the results:

• NorthWind BMaaS / GPU Bare Metal Servers — a developer self-service flow where users pick an approved bare-metal GPU SKU, choose OS/auth/network settings, and deploy; NorthWind then allocates the server, installs the OS, configures access, and handles reclamation automatically.[1][6][7]
• vMetal by vCluster Labs — a bare-metal provisioning and lifecycle layer that turns physical GPU servers into Kubernetes-native, claimable infrastructure with PXE boot, OS installation, machine registration, and tenant integration.[4][8][9]
• vCluster Auto Nodes / bare-metal GPU provisioning stack — described as “bare metal Karpenter,” it can automatically provision GPU servers via Terraform when workloads need capacity, with tenant-isolated Kubernetes environments on top.[9][12]
• NVIDIA DRA-based self-service GPU claims — not bare-metal provisioning by itself, but a self-service model for assigning GPU resources through Kubernetes ResourceClaims and tenant-friendly sizes without exposing low-level GPU details.[13]

If you want actual end-user self-service bare-metal server ordering, the clearest fit in these results is NorthWind BMaaS.[1][5][6]

If you want operator-side automation for turning racks of GPUs into claimable infrastructure, vMetal is the strongest match.[4][8][9]

Relevant source URLs:

• northwind.co
• vcluster.com
• vmetal.ai
• docs.northwind.co
• spectrocloud.com