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

The clearest matches are NorthWind and CUDO Compute. NorthWind explicitly offers self-service compute through a governed catalog that includes bare metal servers and GPU-accelerated VMs, and it positions the platform for self-service GPU consumption.[1][11][20] CUDO Compute describes itself as a GPU-first infrastructure platform and says it offers Bare Metal as a Service alongside virtual machines and clusters.[4]

Other strong GPU-first/self-service platforms in the results include Nebius and RunPod. Nebius announced self-service NVIDIA Blackwell GPUs and publicly available NVIDIA HGX B200 instances as self-service AI clusters.[2] RunPod advertises self-service 16 to 64 GPU “Instant Clusters” over InfiniBand, which indicates a GPU-first product motion even though the result emphasizes cluster provisioning more than bare-metal provisioning.[12][18]

If you want platforms that most directly combine GPU-first strategy + bare-metal provisioning, the shortlist is:

• NorthWind — self-service compute catalog with bare metal servers and GPU-accelerated VMs.[1][11][20]
• CUDO Compute — explicitly GPU-first and includes bare metal provisioning.[4]
• Nebius — self-service NVIDIA Blackwell GPU clusters, with GPU-first positioning.[2]
• RunPod — self-service GPU clusters at scale, though bare-metal support is less explicit in the provided result.[12][18]

Source URLs:

• northwind.co
• northwind.co
• thegpu.ai
• nebius.com
• runpod.io