SUSE and NVIDIA launch sovereign AI platform for enterprise agents

Securing autonomous AI agents

SUSE has introduced new capabilities to help enterprises adopt Private Enterprise AI while maintaining operational control. The company is integrating the NVIDIA OpenShell Reference Architecture into its SUSE AI Factory with NVIDIA platform. This move is designed to help organizations manage autonomous AI agents, which are software systems capable of performing tasks independently.

A primary goal of this integration is providing zero-trust agent security. This means that the system constantly verifies the identity and permissions of every agent. By using supervisor proxies, the platform inspects agent traffic and authorizes access based on specific company policies. This setup helps prevent the unauthorized transfer of proprietary datasets or intellectual property outside of a company's private infrastructure.

The platform also addresses the financial side of AI deployment by offering a predictable cost model. Instead of paying third-party fees for every token processed by a public API, companies can use private model serving. This approach utilizes NVIDIA NIM microservices to run models on internal hardware, allowing for capacity-based infrastructure investments instead of variable usage costs.

Hardware and software alignment

To power these AI workloads, SUSE is launching SLES for NVIDIA. This is a version of their Linux Enterprise Server specifically engineered for NVIDIA's latest hardware. It provides immediate support for next-generation silicon, including the NVIDIA Vera Rubin infrastructure. This allows companies to adopt new hardware quickly without waiting for standard software updates.

One of the main benefits of this specialized software is the reduction of friction during production deployment. It bridges the gap between a developer's testing environment and the final production footprint. Because it is a trusted enterprise foundation, it ensures that code does not need to be rewritten or moved to a different platform when it moves from a lab to a live environment.

SUSE also provides single-source technical support for this entire stack. This includes the operating system, the Linux kernel, the GPU driver stack, and the Kubernetes runtime. By providing one point of contact for support, the company aims to eliminate the technical hurdles that often occur when different hardware and software components from different vendors fail to work together perfectly.

Sovereign AI in the public sector

The concept of sovereign AI is becoming increasingly important for government entities. Research suggests that by 2027, over one-third of nations may find themselves restricted to AI systems tied to specific regions and their unique data. To address this, SUSE is working with various partners to ensure that governments can maintain full data sovereignty and centralized management over their AI systems.

Mexico’s Guanajuato Secretariat of Finance is currently using the SUSE AI Factory with NVIDIA to test these capabilities. They are exploring a cognitive government model to streamline internal workflows and provide smart public services to citizens. This pilot program evaluates how high-performance GPU infrastructure can bridge the gap between initial software testing and full-scale production.

By using an open source foundation, the Secretariat of Finance can evaluate various AI workloads while keeping citizen data secure. The project demonstrates how sovereign infrastructure can be used to manage hybrid cloud and on-premises environments. This allows the government to maintain compliance and control over its operational costs while still adopting modern AI technologies.

Expanding the AI ecosystem

SUSE is also introducing a two-tiered ISV certification program to help partners integrate with their AI Factory. This framework ensures that third-party software can operate seamlessly alongside SUSE and NVIDIA technologies. The program includes a 'Ready for' certification for basic compatibility and a 'Certified for' certification for jointly engineered configurations with automated deployment.

Several partners have already joined the ecosystem to provide specific tools for AI management. For example, Klique provides an AI control plane for resource management, while Tensormesh manages cache for long-context workloads. Traefik Labs offers a gateway to mask sensitive data and block unsafe prompts, while AlchemiStudio provides a platform for real-time cost and budget control.

The ecosystem also includes specialized tools for disaster recovery and niche applications. CloudCasa by Catalogic provides backup for AI workloads running on Kubernetes, protecting things like vLLM chatbots. Additionally, Wiser Technology offers a platform for real-time decision support using digital twins from vehicle fleets. These partnerships aim to move customers from simple experimentation into production-ready, sovereign AI environments.

Hybrid quantum intelligence

In a move to bring advanced computing to the enterprise, SUSE and Fujitsu are partnering on a hybrid quantum platform. This initiative combines the SUSE AI Factory with NVIDIA and NVIDIA CUDA-Q software on Fujitsu servers. It is designed to move hybrid quantum intelligence out of research labs and into standard enterprise data centers.

This partnership marks a milestone as it delivers a production-ready stack that combines NVIDIA AI Enterprise with enterprise-level hybrid quantum computing on sovereign infrastructure. It allows customers to use sovereign infrastructure to run quantum simulations. These simulations can then be connected directly to production applications, potentially accelerating complex problem-solving tasks.

According to Udo Wuertz, CTO of FSAS Technologies EMEA, this stack allows customers to supercharge their existing hardware investments. By running the stack on Fujitsu servers, companies can utilize the new quantum computer and Digital Annealer. This provides a secure foundation for enterprises that want to explore quantum capabilities without sacrificing the stability of their current infrastructure.