Building Self-Service, Sovereign AI Infrastructure for Defence
BWI GmbH is the German Bundeswehr’s primary IT and digitalization partner, responsible for the infrastructure behind its Software Defined Defence strategy. That means building infrastructure that can support modern applications across central data centers, edge locations, and even disconnected field environments.
As part of that mission, BWI gives Bundeswehr units and research partners access to modern AI capabilities, including language models, computer vision, and geospatial analysis.
Through its experimental platform, xPlatforms (BwXLab), BWI provides GPU-powered infrastructure for these AI applications. The challenge was making that infrastructure easy to access and manage while keeping it secure, sovereign, and flexible enough to work across different environments.
The Challenge
Making GPU Infrastructure Available on Demand
BWI had the GPU infrastructure needed to support growing AI use cases. The challenge was making that infrastructure available to users quickly, securely, and wherever they needed it.
Traditional provisioning processes relied on individual requests, tickets, and manual setup. However, for teams working with data analytics, model training, and AI inference, this made it difficult to get the resources they needed without long wait times.
At the same time, GPUs are expensive and cannot simply be approved, purchased, and deployed whenever demand increases. Different Bundeswehr units and research partners needed to share the available hardware, while keeping their workloads and data securely separated.
Lastly, BWI needed to maintain control of its infrastructure. The platform had to run without relying on foreign cloud providers and support environments beyond the data center, including disconnected field locations.
Solution
Kubermatic AI
Kubermatic and SVA worked closely together to find the right solution for BWI. SVA brought its long-standing relationship with BWI and deep understanding of its requirements, while Kubermatic brought the Kubernetes and cloud-native expertise. Together, they built the foundation for BWI's xPlatforms environment on
Kubermatic AI.
Instead of setting up separate environments for every team, BWI can now run different AI workloads on a shared pool of GPUs. Teams remain securely isolated, while BWI gets more value from the hardware it already has. The platform already supports language model inference, data science, computer vision, image generation, and GPU-powered notebooks, all running side by side on the same infrastructure.
One of the biggest benefits for BWI was speed. What once required procurement, hardware setup, and lengthy infrastructure processes can now be done through self-service in minutes. And because the platform runs on BWI's own infrastructure, it does not depend on foreign cloud providers. It can also support disconnected environments, including GPU hardware already deployed at a field site.
SVA has a long-standing relationship with BWI and a deep understanding of its environment and requirements. SVA continues to support BWI as new requirements and use cases emerge, while Kubermatic provides the underlying Kubernetes and cloud-native platform expertise. Together, SVA and Kubermatic give BWI faster access to AI, better use of its GPU resources, and the control needed to run AI securely across connected and disconnected environments.
