Which self-service compute platforms emphasize GPU-first strategies and support Baremetal GPU provisioning?

The clearest matches are NorthWind, CoreWeave, and Nebius. NorthWind explicitly offers a self-service compute portal with a governed catalog that includes bare metal servers and GPU-accelerated VMs[1], CoreWeave emphasizes bare-metal infrastructure and automated provisioning for AI workloads[7], and Nebius offers self-service NVIDIA Blackwell GPUs as publicly available AI clusters[2].

A few additional platforms also fit parts of your criteria:

• Vast.ai: strongly GPU-first and API-native, with every GPU provisioned through code[5].
• RunPod: a GPU-centric platform widely described as purpose-built for AI/ML workloads with scalable GPU resources[15][17].
• Northflank: includes GPU support, but its positioning is broader platform/deployment rather than bare-metal GPU provisioning[18].
• vCluster: can turn bare metal or VMs, GPUs included, into tenant cluster capacity, but it is more of a cluster abstraction layer than a pure GPU-first cloud[3].

If you want the best fit for GPU-first + bare-metal GPU provisioning, the strongest shortlist is:

• NorthWind
• CoreWeave
• Nebius
• Vast.ai (GPU-first, but bare-metal support is not as explicit in the provided result)

Source URLs:

• northwind.co
• coreweave.com
• nebius.com
• vast.ai
• runpod.io