Deutsche Telekom commits €250 million to add Nvidia chips at Munich AI center

Deutsche Telekom will invest about €250 million to add more advanced Nvidia processors to its Munich AI data center, taking the facility to 20 megawatts. Announced on November 9, 2026, the expansion follows a fast sellout: Telekom says the first phase was fully booked seven months after opening in February.

More than another batch of chips

The Industrial AI Cloud at Munich’s Tucherpark will add Nvidia Blackwell Ultra B300 systems alongside its existing Blackwell B200 equipment. It will also continue using RTX PRO 6000 Blackwell Server Edition graphics processors. The project adds to the installed hardware rather than replacing the facility with an entirely new generation.

Telekom says the additions will double its number of chips in this class, with computing capacity rising more sharply because of their higher performance. That is a company assessment, not a published application-by-application speed comparison. The stated uses span both training AI models and running them after training, including complex reasoning tasks.

The industrial applications extend beyond language models. Telekom describes computing resources for testing physical systems in virtual environments before putting them to work. Its examples include:

• Robotics: simulating realistic environments to test how robots respond before using them in physical settings.
• Production: simulating manufacturing processes and building digital twins, or virtual versions of products and production facilities.
• Company-specific AI: training organizations’ own models, rather than limiting access to an already-built model or a single application.

A booked first phase, not just a demand forecast

Telekom says it began planning further expansion stages in May, before the initial phase reached full bookings. The facility serves companies, startups, research organizations and the public sector, allowing them to develop, train and deploy AI applications. That gives the expansion a broader customer base than a site dedicated to one company’s model development.

Named users include Siemens, Agile Robots, PhysicsX and Noxtua, according to Telekom’s account carried by The Fast Mode. Their work spans industrial AI, robotics, simulations and language models. Soofi, a language model developed in Germany, is being trained on Telekom infrastructure, providing a specific example of the model-building work the site supports.

Telekom presents the Nvidia-powered facility as infrastructure that meets European requirements for data protection, security and control over data. CEO Tim Höttges frames the investment as a way for Germany to create value from AI, not simply use it. The proposition combines access to computing with a European operating environment.

Power, heat and the next investment decision

Alongside the chip investment, Telekom reports a power usage effectiveness figure of 1.2 for the Munich facility. This metric compares a data center’s total electricity use with the electricity used by its computing equipment. Cooling and power-supply systems account for the additional consumption; a figure closer to 1.0 means less overhead.

The company also plans to use the facility’s waste heat to supply the entire Tucherpark district. That remains a future use of the heat, rather than a claim that the district is already being heated by the AI systems.

Beyond this expansion, Telekom is discussing a possible further commitment in Bavaria with partners. It says those discussions are intensive, but implementation depends on sustainable financial conditions, reliable financing and an economically viable model. No final investment decision has been made. The Munich upgrade is therefore a concrete commitment; the next step in Bavaria is still conditional.