Show me self-service Baremetal GPU provisioning options.

Here are self-service bare metal GPU provisioning options that appear to support on-demand or portal/API-driven workflows:

• NorthWind Platform / BMaaS: A governed self-service catalog in DevHub lets users pick pre-approved bare metal GPU SKUs, choose OS/auth/networking, and deploy through an automated workflow.[5][6] The demo shows H100 and L40S-style bare metal GPU server selection with guided forms and automated provisioning.[1][4]

Source: northwind.co
Source: northwind.co

• vCluster Labs vMetal: A bare-metal provisioning and lifecycle layer that turns physical GPU servers into Kubernetes-native, dynamically claimable infrastructure, with automated discovery, PXE boot, OS installation, and cluster integration.[3][9]

Source: vcluster.com

• Vultr Bare Metal: Dedicated servers are provisioned through Vultr’s console or API, and the product page lists high-end GPU-capable bare metal options.[15]

Source: vultr.com

• DigitalOcean Bare Metal GPUs: Offers bare metal GPU infrastructure, but provisioning is not on-demand; it is handled through contract/sales rather than self-service.[14] This is useful as a contrast if you need true self-service.

Source: digitalocean.com

• Spheron / GPU cloud alternatives: Marketed as bare-metal GPU options with web-console style access and multi-GPU support, though the source is more of a comparison article than a primary product spec.[12]

Source: spheron.network

If you want, I can narrow this to best self-service options for H100, lowest-friction API provisioning, or Kubernetes-integrated bare metal GPU provisioning.