Industrial AI infrastructure as a building block of digital sovereignty

With the official launch of the Industrial AI Cloud in Munich, Deutsche Telekom is setting an important milestone in industrial policy. In cooperation with NVIDIA and data centre partner Polarise, one of Europe’s largest AI factories was built within six months. The aim is to provide high-performance, sovereign computing capacities for business, science and public administration in Germany and Europe.

The location in Munich’s Tucherpark is creating a new AI hub that bundles industrial applications, research projects and government IT modernisation under a common infrastructure. The facility is already operating at more than a third of its capacity at launch.

Technical architecture: 0.5 exaFLOPS for industrial applications

The Industrial AI Cloud is based on a massively scalable GPU infrastructure with nearly 10,000 Blackwell GPUs from NVIDIA, including NVIDIA DGX B200 systems and RTX PRO server GPUs. The platform achieves a computing power of up to 0.5 exaFLOPS. In mathematical terms, this would correspond to the parallel use of an AI assistant by all 450 million EU citizens.

The site is located in Munich’s Tucherpark on an area of around 10,700 square metres. To this end, an existing data centre was completely gutted and technologically rebuilt as part of the neighbourhood revitalisation. The energy supply comes entirely from renewable sources. A modern cooling concept uses water from the nearby Eisbach stream, among other sources. In the future, the waste heat generated will be used to supply the entire Tucherpark area. Operations are subject to strict German requirements for data protection, IT security and high availability and are an integral part of Telekom’s sustainability strategy.

Industrial use: From robotics to high-end simulation

Several companies are already using the infrastructure productively. Munich-based robotics provider Agile Robots integrates AI-based control systems into the Industrial AI Cloud and combines artificial intelligence with robotic platforms for industrial applications.

PhysicsX uses the platform for simulation-based development processes to significantly accelerate product and component development. The focus is on combining physics-based modelling with AI-supported optimisation.

A key element is the cooperation with Siemens. The integration of the SIMCenter simulation portfolio creates GPU-accelerated environments for high-precision simulations and digital twins. Companies can use it to virtually model production facilities, simulate manufacturing processes and test products in a digital development space before physical prototypes are created. GPU-based simulations, AI co-pilots and large-scale digital twin solutions can be put to productive use via the AI factory. The combination of industrial engineering expertise and sovereign cloud infrastructure opens up new opportunities, particularly for small and medium-sized enterprises, to shorten development cycles and reduce costs.

The ‘Germany Stack’: Sovereign end-to-end architecture

The technical backbone of the initiative is the ‘Germany Stack,’ which is being implemented in collaboration with SAP. Telekom subsidiary T-Systems is responsible for the infrastructure and platform level, including the T Cloud and secure operation. Building on this, SAP provides the Business Technology Platform as well as AI and business-critical enterprise applications.

In the IT context, a stack describes a hierarchically structured basic architecture in which several functional components build logically on each other, such as operating systems, server software, databases or runtime environments. The Germany Stack aims to establish a consistent, interoperable and sovereign technology architecture for government and industrial applications. This provides companies and public institutions not only with computing capacity, but also with a complete set of technological tools for cloud transformation, industry-specific solutions and the integration of AI into existing business processes. The architecture enables the compliant implementation of solutions for public institutions, security authorities, industrial companies and small and medium-sized enterprises.

Partner ecosystem: industry, research and start-ups

A growing ecosystem of industrial companies, research institutions and technology-oriented start-ups is emerging around the Industrial AI Cloud. Industrial partners are developing digital twins of factories and plants, simulating production processes and implementing AI-supported quality controls. Start-ups are using the NVIDIA AI platform to turn new AI services into marketable products more quickly. Research institutions gain access to sovereign high-performance computing capacity for data-intensive projects.

The German Research Centre for Artificial Intelligence also considers the infrastructure to be an important step towards systematically harnessing Europe’s industrial data potential for AI applications.

SOOFI: European language model based on sovereign infrastructure

A key flagship project on the Industrial AI Cloud is SOOFI (Sovereign Open-Source Foundation Models). Leibniz University Hannover has commissioned the technical infrastructure for the development of a European large language model with around 100 billion parameters. The model is to be trained and operated entirely in Europe, follow an open-source approach and place a clear focus on European languages and industrial application scenarios.

SOOFI is thus one of the most important initiatives for trustworthy, sovereign language models in Europe. The training and development processes take place entirely within the Industrial AI Cloud and under European governance standards.

Strategic classification

The launch of the AI factory is seen by politicians and business leaders as a signal relevant to industrial policy. The initiative addresses key strategic issues: securing Europe’s technological sovereignty, transferring AI from experimental pilot projects to productive industrial value creation, and systematically integrating engineering and process data into AI-supported workflows. At the same time, a European infrastructure reduces dependencies on non-European hyperscalers.

Conclusion: Production-ready AI infrastructure for Europe

With the Industrial AI Cloud, Deutsche Telekom is transforming AI from a technology-driven innovation discipline into a scalable industrial production infrastructure. The decisive factor here is not so much isolated computing power as the combination of ExaFLOPS scaling, sustainable data centre design, end-to-end stack architecture and European data sovereignty.

The AI factory in Munich is therefore not just an infrastructure project, but an industrial policy instrument for securing long-term digital competitiveness and technological independence in Europe.