In Munich, Deutsche Telekom is teaming up with Nvidia to establish a high-capacity AI Factory aimed at bolstering Germany’s and Europe’s digital sovereignty. Designed for large-scale training and inference, this infrastructure will provide secure hosting for high-value services such as SAP solutions. This ambitious initiative is part of a long-term industrial project backed by a sovereign and interoperable ecosystem.
Through its €1 billion partnership with Nvidia, Deutsche Telekom is advancing its position as a key European supplier, constructing an AI Factory capable of meeting the skyrocketing demand for AI resources while ensuring the integration of a European technological foundation. This announcement comes at a time of geopolitical fragmentation and heightened competition for computing resources, as major powers ramp up their capabilities to avoid dependence on external infrastructures.
At the heart of this project is the installation of over 1,000 Nvidia DGX B200 and RTX PRO Servers in a revamped data center in Munich, equipped with up to 10,000 interconnected Blackwell GPUs linked via 75 kilometers of fiber optic cables. The targeted processing power is set to reach 0.5 exaflops, coupled with a storage capacity of 20 petabytes. This will be one of the largest industrial AI platforms in Europe, tailored to meet the needs of businesses, government agencies, and sovereign providers. The facility, operated in partnership with Polarise, is expected to be operational by Q1 2026.
A Sovereign Base for Training and Inference
This initiative must be viewed in the context of Europe’s structural deficit in high-performance computing. While the U.S. boasts at least five exascale-class AI farms and China advances strategic computing architectures, Europe struggles to establish its operational autonomy. The German initiative positions itself as a pioneering effort, akin to a fast and agile vessel navigating alongside the European AI Gigafactories project. It could signal a more pragmatic reallocation of European resources toward industrial players ready to act quickly within a defined compliance framework.
The project extends beyond enhancing computing capacity; it explicitly includes training large language models (LLMs), deploying massively parallelized inference models, and optimizing industrial processes through 3D simulation. Integration of Nvidia’s Omniverse libraries will facilitate the design of digital twins for factories, products, or supply chains. Such applications pave the way for almost negligible-cost simulations, such as virtual crash tests or trials in digital wind tunnels, marking a significant operational breakthrough that reconciles productivity with sovereignty.
Furthering this infrastructure, Deutsche Telekom plans to offer comprehensive support for its clients: secure connections to the T Cloud, model transfers, application integration, regulatory compliance, and business deployments. The operator aims to create an industrialized environment where AI computing functions are tightly integrated with the client’s core business processes. In doing so, this initiative addresses the challenges identified by IT managers and business architects: resource pooling, scaling, and compliance by design.
A Germany Stack with SAP
The initiative is anchored by a foundational technology stack known as the “Deutschland-Stack.” This sovereign stack combines Deutsche Telekom’s physical infrastructures with solutions from the SAP Business Technology Platform, integrating embedded AI, business connectors, and application security. Together, they aim to offer a local alternative to mainstream cloud platforms, particularly for regulated sectors or those handling sensitive data.
This stack will host a comprehensive range of digital applications for both businesses and public institutions in Germany. The synergy between T Cloud, the AI Factory, and the SAP BTP assures a 100% European value chain that is interoperable and auditable. This foundation is particularly relevant as Germany, despite its dense industrial fabric and competitive challenges, seeks to rapidly modernize its production capabilities. The platform emerges not merely as infrastructure but as a strategic lever to secure the reinvention of industrial value chains, merging technological sovereignty, innovative agility, and software mastery.
An Existing Ecosystem Ready to Deploy AI Applications
Several companies have already expressed interest in leveraging the AI Factory. Among them is Agile Robots, whose systems will not only be used to physically install the racks in the Munich data center but also to train robots specializing in polishing and assembly tasks in industrial settings. This connection between production tools and the training platform embodies an integrated vision of the AI value chain, from simulation to physical execution.
Deutsche Telekom’s proposed model emphasizes openness to third parties, from startups to large enterprises, wishing to train, host, or deploy their models within a sovereign environment. This ecosystemic approach enhances the platform’s attractiveness, especially for specialized software vendors, AI integrators, and manufacturers developing their own autonomous agents. Here, the platform logic intertwines with a demand for traceability, algorithmic transparency, and European interoperability.



