SUSE and NVIDIA launch sovereign AI platform for private enterprise use

Defining AI sovereignty

SUSE recently announced a series of products and partnerships with NVIDIA designed to give companies more authority over their private AI deployments. The primary goal is to provide the infrastructure needed for autonomous AI agents to function inside a corporate environment. By focusing on sovereignty, the companies aim to help organizations keep their data and intellectual property within their own private systems.

Dr. Thomas Di Giacomo, Chief Technology and Product Officer at SUSE, noted that the move responds to a growing demand for control over AI systems and the hardware that powers them. He stated that as next-generation silicon and autonomous agents reshape how businesses operate, AI sovereignty becomes non-negotiable. This means companies want to ensure their AI tools do not leak information to outside parties.

The update includes SUSE becoming a launch partner for NVIDIA's OpenShell Reference Architecture. This integration allows organizations to run AI agents on their own infrastructure while applying specific access policies. These policies are designed to restrict the movement of proprietary data outside of the private system. This approach is particularly relevant for companies operating in highly regulated or sensitive environments.

The hardware and software stack

The new architecture combines several specific components from NVIDIA to manage different parts of the AI workflow. It uses NVIDIA Vera Rubin for model inference, which is the process of generating outputs from a trained model. For the actual execution of AI agents, the system utilizes NVIDIA Vera CPUs. Monitoring and policy enforcement are handled by NVIDIA BlueField-4 DPUs, which are specialized processors for data movement and security.

To manage how these agents run, SUSE will use Kubernetes sandboxing and API gateways alongside the OpenShell integration. Kubernetes is a system for automating the deployment and management of containerized applications. Sandboxing provides an isolated environment for these applications to run without affecting the rest of the system. These tools work together to govern the behavior and security of the AI agents.

Additionally, SUSE introduced SLES for NVIDIA, a specific version of its Linux platform built to support NVIDIA's latest chips. This version includes support for the Vera Rubin hardware. It provides a single support route that covers the operating system, the kernel, GPU drivers, and the Kubernetes runtime. This unified support model is intended to make it easier for IT teams to troubleshoot production problems without coordinating between multiple vendors.

Validation and partner ecosystem

SUSE is also establishing a two-tier software certification structure for its AI Factory products. This program is meant to confirm that third-party software works correctly with the platform and to clarify who is responsible for support when issues occur. The two levels are 'Ready for' and 'Certified for.' The first level is based on a partner's self-validation and support commitments, while the second involves jointly validated configurations with automated deployment.

Six partners have already joined this program, including companies like Klique by ClearML and Traefik Labs. Their products cover a variety of needs, such as AI control planes, inference optimization, and governance tools. They also provide software for digital twin decision support and backup services for Kubernetes-based AI workloads. This ecosystem is designed to help customers move beyond simple trials into full production environments.

Rhys Oxenham, General Manager of AI at SUSE, explained that software validation is often a barrier for customers trying to scale. He noted that enterprises need a trusted ecosystem rather than just baseline infrastructure. By enriching the portfolio with pre-validated capabilities, SUSE aims to provide a deeply integrated platform. This allows customers to move seamlessly from experimentation to sovereign, enterprise-grade production.

Real-world use and quantum goals

The Secretariat of Finance of Guanajuato in Mexico is already using SUSE AI Factory with NVIDIA to test sovereign AI services. This proof of concept focuses on validating AI-led services and internal workflows for public administration. The goal is to keep all data and infrastructure under local control while exploring how AI can improve government operations. José Luis Huerta Santos, who leads Planning and IT Architecture there, stated that the open source foundation allows for rapid evaluation of AI workloads.

Beyond standard AI, SUSE is also working with Fujitsu on an AI-driven hybrid quantum platform. This stack combines SUSE AI Factory with NVIDIA CUDA-Q software on Fujitsu servers. It is designed to bring hybrid quantum computing together with NVIDIA AI Enterprise. The arrangement connects quantum simulation workloads with production enterprise applications that run on existing server infrastructure.

Fujitsu stated that this partnership provides the secure infrastructure foundation needed for enterprises to expand their technology portfolios. By running this stack on Fujitsu servers, customers can connect quantum simulations directly to production applications. This allows organizations to supercharge their existing hardware investments while exploring the potential of quantum computing. These developments highlight a broader trend toward creating secure, sovereign foundations for diverse computing needs.

Summary of the new platform

The collaboration between SUSE and NVIDIA addresses the core tension between the desire for powerful AI and the need for data privacy. By providing a sovereign AI platform, these companies are offering a way for businesses to utilize autonomous agents without sacrificing control over their intellectual property. The integration of specific hardware like Vera Rubin and BlueField-4 DPUs ensures that the infrastructure is optimized for these specific tasks.

For IT teams, the introduction of SLES for NVIDIA simplifies the management of complex AI environments. Having a single support route for the operating system and drivers reduces the friction of troubleshooting. This is a significant step toward making AI more accessible to companies that cannot rely on public cloud services for their most sensitive data. It provides a clear path for moving from experimental AI to reliable, production-ready systems.

Looking ahead, the expansion into hybrid quantum computing and the establishment of a rigorous certification program suggest a long-term strategy. SUSE and NVIDIA are building a comprehensive ecosystem that supports various levels of AI maturity. From public administration in Mexico to high-end quantum simulations, the focus remains on providing secure, sovereign foundations for the next generation of enterprise computing.