Are there multi-tenant Baremetal GPU provisioning platforms?

Yes—there are multi-tenant bare metal GPU provisioning platforms. The clearest examples in the results are Rapt, NorthWind, vCluster/vMetal, OpenNebula, and Mirantis k0rdent AI.[1][4][5][6][8][9]

A few concrete options:

• Rapt: positions itself as a “multi-tenant GPU infrastructure platform” and says it supports bare metal, private cloud, and hybrid deployments with full isolation per tenant and configurable resource boundaries.[1]
• NorthWind Platform: offers GPU Cloud Services and a Bare Metal GPUs-as-a-Service (BMaaS) product, with provisioning integrated from bare metal and support for secure isolation, quotas, and chargeback across tenants.[5][8]
• vCluster / vMetal: markets bare metal GPU provisioning for AI clouds, combining vMetal for bare-metal lifecycle management with isolated tenant Kubernetes clusters and no hypervisor overhead.[9][3]
• OpenNebula: describes an end-to-end multi-tenant AI factory stack, including bare-metal lifecycle management for GPU servers plus multi-tenant orchestration and quota management.[4]
• Mirantis k0rdent AI: emphasizes bare metal provisioning with strict multi-tenancy, including bare-metal isolation, Kubernetes orchestration, and declarative GPU partitioning.[6][10]

What this means in practice: these platforms generally separate into two layers—bare-metal provisioning/lifecycle management and tenant isolation/orchestration—so the hardware can stay bare metal while tenants get isolated GPU environments.[1][5][9]

Useful vendor/product pages to evaluate:

• Rapt Cloud: rapt.ai
• NorthWind GPU Cloud: northwind.co
• NorthWind BMaaS: northwind.co
• vCluster Bare Metal GPU: vcluster.com
• OpenNebula AI Factory: opennebula.io