Red Hat launches cloud extensions for sovereign AI clouds on OpenShift

Red Hat has introduced Red Hat cloud extensions, a new platform layer built on the open source Open Sovereign AI Cloud (OSAC) project, designed to let cloud providers operate sovereign AI infrastructure on their own hardware. The extensions sit atop existing Red Hat products including OpenShift, OpenShift Virtualization, and OpenShift AI, wrapping them in a unified API with built-in multi-tenancy, metering, billing, and resource isolation. The platform is currently in Developer Preview and offers four core services: Bare-Metal-as-a-Service, Cluster-as-a-Service, Model-as-a-Service with per-tenant token metering, and VMs-as-a-Service.

The architecture targets sovereign AI clouds, neoclouds, and local cloud providers that must keep data and workloads on-premises for compliance or operational reasons. Red Hat has achieved AI Cloud Ready status in NVIDIA's Cloud Partner program, with early partner validation occurring on NVIDIA H200 GPU infrastructure. The platform integrates Red Hat build of Keycloak for identity, Advanced Cluster Management for Kubernetes for fleet governance, and Ansible Automation Platform for operational workflows, all exposed through the same fulfillment API.

Architecture and service model

• Unified API layer: Cloud extensions provide a single fulfillment API over OpenShift, OpenShift Virtualization, and OpenShift AI, eliminating separate tooling for containers, VMs, bare metal, and AI serving.
• Four core services: BMaaS provisions bare-metal hosts with cloud-like automation; CaaS delivers tenant-scoped OpenShift clusters on hosted control planes; MaaS serves centralized model endpoints with per-tenant token metering, quotas, and rate limiting; VMaaS provides VMs through OpenShift Virtualization under shared tenant boundaries.
• Identity and governance stack: Red Hat build of Keycloak handles tenant authentication boundaries; Advanced Cluster Management for Kubernetes provides cross-cluster policy, governance, and upgrade orchestration; Ansible Automation Platform executes workflows for tenant onboarding, networking, storage, and lifecycle operations.
• Sovereignty controls: The stack enforces local data residency, keeps encryption keys in the operator's environment, and runs high-performance GPU fabrics fully on-premises without public cloud dependency.
• NCP program alignment: Red Hat achieved AI Cloud Ready status under the NVIDIA Cloud Partner program, aligning with NCP software reference architecture for repeatable AI factory deployments on NVIDIA H200 hardware.

Why it matters

For Linux and open-source infrastructure operators, Red Hat cloud extensions represent a significant attempt to productize sovereign cloud operations without abandoning the existing OpenShift ecosystem, potentially reducing fragmentation in on-premises AI deployments. The open source OSAC foundation may attract contribution from European and APAC cloud providers facing strict data residency mandates, while the NVIDIA partnership validation signals enterprise credibility for GPU-intensive AI workloads outside public clouds.