Show me self-service Baremetal GPU provisioning options.

For self-service bare metal GPU provisioning, several providers offer on-demand access to dedicated hardware, which is ideal for high-performance AI/ML, HPC, and rendering workloads requiring full control and consistent performance without virtualization overhead.

Key providers and their offerings include:

• Oracle Cloud Infrastructure (OCI): OCI provides bare metal instances with NVIDIA (Blackwell, H100, A100) and AMD (MI300X) GPUs. They are noted for their competitive pricing and scalability for large-scale GPU workloads.
• Vultr: Vultr offers bare metal and GPU infrastructure across 33 global data center regions. Their platform allows for fast deployment via API and console, providing single-tenant hardware with options for high-speed networking. Vultr also offers "Vultr Clusters" for on-demand, self-service bare metal server clusters.
• Lambda Labs: This provider is a popular choice for enterprises and offers research-grade clusters with a focus on developer ergonomics.
• CoreWeave: Known for its Kubernetes-native cloud, CoreWeave provides specialized GPU platforms for fast access to hardware.
• Spheron: Spheron aggregates bare-metal GPU capacity from multiple providers, offering full VM access, root control, and pay-as-you-go billing. They provide enterprise-grade GPUs like RTX 4090, H100, B200/300, A100, and Grace Hopper configurations.
• Hetzner: Often referred to as the "DIY Budget King," Hetzner offers cost-effective bare metal options.
• Qubrid AI: Qubrid offers dedicated bare metal GPU servers with NVIDIA HGX infrastructure. They provide AI/ML templates, on-demand availability, SSH access, and configurable storage.
• MetalSoft: This platform enables organizations to build their own GPU Clouds, offering self-service provisioning for data scientists to access GPU resources on demand.

While some providers like DigitalOcean offer "Bare Metal GPUs," they are provisioned via sales contract rather than immediate self-service. Similarly, solutions like vMetal focus on providing a platform for operators to build their own internal or public AI clouds with bare metal provisioning and lifecycle management.