Rafay-powered Bare Metal GPUs as a Service (BMaaS)

Rafay-powered Bare Metal GPUs as a Service (BMaaS) gives customers direct access to the host operating system on dedicated GPU servers, for teams running large distributed training jobs or highly opinionated software stacks that need control a virtualized environment cannot offer.

The platform delivers each server with secure access, networked with the customer's other rented servers and the right storage mounted, running with zero virtualization overhead for maximum performance on AI training, inference, and HPC.

• Automated Bring-Up: Rafay matches the request to inventory, boots and discovers the server, installs the OS, and configures drivers and services.
• Networking & Tenant Isolation: North-south through the Tenant Access Network, east-west over InfiniBand, with EVPN and Pkey isolation via NVIDIA UFM so one tenant's traffic never reaches another's.
• Storage, Access & Firewall: Per-tenant storage mounted to the server, SSH via password or key, a public IP from the tenant pool, and automated security groups and NAT rules.

.webp)

Bare Metal-as-a-Service in the Rafay Platform

Check out the end-user experience in this quick click-through demonstration.

Simplify and Automate GPU Operations

Rafay automates how GPU servers are deployed, managed, and scaled across multiple tenants and environments.

Automated Bare Metal Bring-Up

Match servers from inventory, reserve IPs, boot and discover, install the OS, and configure drivers and services in one workflow.

Network Automation & Tenant Isolation

Per-tenant VRFs, north-south tenant access, and optional firewall rules, programmed automatically as tenants onboard.

High-Speed Interconnects

InfiniBand and Ethernet with RDMA for low-latency distributed training and HPC.

Latest GPU Hardware

NVIDIA H100, A100, L40S and more, up to 8 GPUs per server with NVLink and NVSwitch.

“We are able to deliver new, innovative products and services to the global market faster and manage them cost-effectively with Rafay.”
— Joe Vaughan, Chief Technology Officer, MoneyGram

Drive Efficiency and Business Value

Stand up a governed, multi-tenant bare metal service in weeks, not the months a hand-built platform requires

Move beyond commodity rentals with premium SKUs, performance SLAs, analytics, and integrated billing

Deliver air-gapped, compliant GPU environments tailored for regulated industries and national AI projects

Reallocate expensive GPUs to the next tenant in minutes, not weeks, with per-second billing to maximize utilization

Hybrid Cloud Meets Kubernetes

Learn how to Streamline Kubernetes Ops in Hybrid Clouds with AWS & Rafay.