A Speedboat Partnership for Germany

Polarise serves as the strategic infrastructure partner behind Deutsche Telekom’s Industrial AI Cloud in Munich, delivering one of Germany’s most powerful and sovereign AI infrastructures on an accelerated timeline. This case study outlines how Polarise enabled Deutsche Telekom by turning an existing site into a high-density AI Factory that will host up to 10,000 NVIDIA Blackwell GPUs and provide 0.5 EFLOPS of computing capacity for European industry, starting early 2026.

Customer overview

Deutsche Telekom is one of Europe’s leading telecommunications providers and a critical digital infrastructure player for enterprises and public institutions. To strengthen AI sovereignty for Germany and Europe, Deutsche Telekom is building the Industrial AI Cloud in Munich that allows enterprises to train and run advanced AI models with sensitive and business-critical data under European rules.

The Industrial AI Cloud targets use cases ranging from industrial digital twins and robotics to large language models and sector-specific AI applications for manufacturing, logistics, and public sector workloads.

Challenge

By 2024/25, around one fifth of companies had brought sensitive and mission-critical data back from hyperscale clouds into local environments, significantly increasing demand for sovereign, secure, and high-performance AI capacity under European governance. Deutsche Telekom needed to rapidly create a large-scale AI-cloud in Germany to serve enterprises, mid-sized companies, and startups, without waiting years for new EU-backed Gigafactories to materialize.

The key challenges included:

• Delivering extreme compute density for thousands of NVIDIA DGX B200 and NVIDIA RTX PRO systems while maintaining energy efficiency and resilience.
• Converting an existing site into an AI Factory with minimal time-to-market instead of undertaking lengthy greenfield construction while complying with highest industry standards and security measurements.
• Delivering on an extremely tight schedule to serve first customers by beginning of 2026.

Solution

Polarise was able to quickly access the already successfully scouted site in Tucherpark in Munich, as it has already been in its portfolio. Initially, a data center designed entirely for Polarise’s own usage was envisaged, but it quickly became clear that the location could ideally meet the requirements for the German Industrial AI Cloud. Drawing on prior experience delivering sovereign AI Factories across Europe, including a similar facility in Oslo, Polarise applied an AI-centric design and heat reusage concept optimized for the selected high-density GPU portfolio of the Industrial AI Cloud and supervised the entire fitout. Further, Polarise is responsible for ensuring the smooth operation of the entire system down to rack level.

The solution combines:

• Renovation and transformation of a former banking data center into a purpose-built AI Factory, rather than building a new facility from scratch.
• Deployment of Polarise’s unique “AI Pods”, modular units that function as small data centers within the data center, enabling lightning-fast rollout of dense AI capacity with predictable power and cooling envelopes.
• Design of a scalable, energy-efficient infrastructure that makes use of the water of the nearby Eisbach river for sustainable cooling and reuses waste heat from the AI servers.
• Preparation for integration into district heating concepts, following the model used by Polarise in Oslo, in full compliance with the revitalisation strategy for Munich’s Tucherpark.

Technical implementation

Polarise’s Munich AI Factory is being built to host the Industrial AI Cloud’s more than 1,000 NVIDIA DGX B200 systems and NVIDIA RTX PRO servers, with up to 10,000 NVIDIA Blackwell GPUs and 0.5 EFLOPS of computing power. The site provides around 20 petabytes of storage and is connected via four 400 Gbit/s fiber links, delivering high-throughput, low-latency connectivity into Deutsche Telekom’s network and T-Cloud for seamless performance and integration into Deutsche Telekom’s sovereign cloud portfolio.

Across several thousand square meters of server space, Polarise’s infrastructure design ensures full energy efficiency, adherence to high security and quality standards, and the ability to scale further as demand grows.

Collaboration model

The collaboration is structured so that Polarise focuses on the physical and technical AI infrastructure and its operation, while Deutsche Telekom, NVIDIA, and SAP build the broader cloud and application ecosystem on top.

• Polarise designs, builds, and operates the AI-centric data center infrastructure, including high-density power, cooling, and AI Pods within the AI Factory.
• Deutsche Telekom uses this facility as the physical foundation for the Industrial AI Cloud, offering enterprises access to GPU capacity, migration services, security solutions, and process integration.
• NVIDIA provides the GPU platforms and AI technologies, while SAP delivers the SAP Business Technology Platform and applications as part of a sovereign “Deutschland-Stack” for public and security-related organizations.

Outcomes

The partnership is delivering tangible outcomes years ahead of comparable projects in other parts of Germany and Europe, positioning Munich as a leading AI infrastructure hub. The AI Factory will become operational in early 2026, with Deutsche Telekom set to bring the first high-performance AI chips for the Industrial AI Cloud online as early as the beginning of the year.

Key results include:

• Around 50 percent increase in available AI compute capacity in Germany once the facility is in full service.
• A sovereign, European-governed infrastructure for enterprises, SMEs, startups, and public sector entities to build and run AI solutions with sensitive data.
• One of the most powerful AI infrastructures in Germany, delivered on an accelerated schedule through smart and environmental-friendly reuse of existing infrastructure and cooling capacities by the nearby river and modular AI Pods rather than greenfield builds.

Impact on AI sovereignty and industry

The Munich AI Factory and Industrial AI Cloud directly support Germany’s and Europe’s strategic goal of technological and data sovereignty in AI. By keeping inference and training capacity within Germany, enterprises can develop digital twins, industrial simulations, robotics applications, and large language models while maintaining control over proprietary data.

This infrastructure lays the foundation for industrial productivity gains and job creation in sectors such as manufacturing, logistics, and aviation, while enabling partners like Perplexity to deliver local AI inference services for German users and businesses. Polarise’s model of rapidly deployable, sustainable AI Factories throughout Germany creates a repeatable blueprint for scaling sovereign AI capacity across Europe in record time.