SUSE also made announcements at NVIDIA GTC Berlin on October 21: a new Linux edition called SLES for NVIDIA and integration of NVIDIA’s OpenShell Reference Architecture into SUSE AI Factory with NVIDIA. SUSE frames both around “sovereign” control, meaning enterprises run their AI agents and models on infrastructure they control.
SLES for NVIDIA
SLES for NVIDIA is built on SUSE Linux Enterprise Server 16. SUSE says it is meant to run AI workloads on NVIDIA’s newest hardware from day one, including the Vera Rubin infrastructure, while keeping the stability of an enterprise base. SUSE says it is designed to scale to rack-scale systems and to keep pace with NVIDIA’s release cycles. Support comes from a single vendor and covers the OS, Linux kernel, GPU driver stack and Kubernetes runtime, all validated together. General availability is planned for November 2026, and SUSE is demoing SLES for NVIDIA 16.1 at its GTC booth.
OpenShell in SUSE AI Factory
SUSE is a launch partner for the NVIDIA OpenShell Reference Architecture. That design pairs NVIDIA Vera Rubin for model inference with Vera CPUs running OpenShell for agent execution and BlueField-4 DPUs running NVIDIA Sentry for monitoring that is independent of the host. SUSE combines it with Kubernetes sandboxing and API gateways so that agent traffic passes through policy-checking proxies, agent code runs inside the corporate network, and models are served privately through NVIDIA NIM rather than paid per token through public APIs.
A public-sector pilot
SUSE says the Secretariat of Finance in Mexico’s state of Guanajuato is testing SUSE AI Factory with NVIDIA in a “cognitive government” proof of concept, alongside its existing SLES and SUSE Multi-Linux Manager setup.
Red Hat made a similar move the same day, adding NVIDIA Vera CPU support to RHEL for NVIDIA. SUSE also announced a hybrid quantum computing partnership with Fujitsu.
Sources: SUSE press release (GlobeNewswire), SUSE blog.