An AI factory will be built in Munich. Nvidia and Deutsche Telekom will increase Germany's computing power by half
The project has the nature of a sovereign platform for production companies and public institutionsoperating in German jurisdiction. In Munich, an existing data center will be rebuilt from scratch. There will be a cluster with a capacity of approximately 0.5 EFLOPS, with over 1,000 DGX B200 systems and RTX Pro servers, approximately 20 PB of mass storage and network facilities connected, among others, four 400 Gb connections and 75 km of optical fibers. The installation of server racks is to be supported by Agile Robots.
To illustrate these possibilities, let's imagine an “AI machine” capable of approximately half a trillion calculations per second, i.e. it's a bit like connecting over a thousand ultra-efficient computers. Additionally, they are equipped with memory equivalent to approx. 40,000. laptops with a network that would play tens of thousands of 4K videos at once.
The partner ecosystem for this project is as important as the infrastructure itself. The first users and cooperating companies include: SAP, Siemens, Agile Robots, Wandelbots, PhysicsX and Perplexity. The latter of the companies intends to use the Munich Cloud for sovereign data processing in Germanyso that the answers generated by the AI model are created and processed locally.
The platform will also support digital twins of factories in Omniverse, training robots in simulation and the development of large language models for industrial applications.
Check also: Poland will apply to the European Commission for the construction of an AI gigafactory. Billions of PLN on the table
The project is independent of EU plans
Although the project is being developed independently of the EU's plans to build AI gigafactories, it fits well with Brussels' ambitions to increase the availability of computing power in Europe. The European Commission has launched the InvestAI program with the goal of mobilizing EUR 200 billion, including a dedicated EUR 20 billion fund for 4-5 gigafactories. The first invitations have already received 76 offers from 16 EU countries.
The Munich AI factory is intended to be a fast, private venture – a complement to large, publicly supported projects.
Why is it important for Europe and the region? Data sovereignty is a priority. Local, controlled infrastructure allows companies and the public sector to build and run AI systems in accordance with European data protection regulations and the AI Act – without the need to use centers outside the EU. The example of Perplexity, which wants to process data in Europe, shows that such needs are very real.
In addition, there is also the industrialization of AI. Nvidia and Deutsche Telekom have a shared environment accelerate implementations in industry – from optimization of production lines to aerodynamics simulations and crash tests in virtual tunnels. And investors are still counting on the so-called competence spillover effect. Access to computing power in Munich will make it easier for companies from DACH and across Central and Eastern Europe to create resource-intensive applications, shorten model training times, and lower entry barriers for SMEs and start-ups that have so far been losing the fight for GPUs.
Read also: The Minister of Digital Affairs reveals how much the AI gigafactory in Poland will cost
Europe wants to catch up with the USA
The latest project is also a step in catching up with the US, although the scale is still incomparable to American megaprojects. Overseas, OpenAI, Oracle and SoftBank are developing the Stargate venture with a target volume of hundreds of billions of dollars and several gigawatts of powerand in parallel, OpenAI concluded one of the largest cloud contracts in history with Oracle, worth USD 300 billion. Against this background, the billion euro in Munich is a balanced, European move – smaller, but ready for use earlier and subject to the principles of sovereignty.
The Nvidia and telecom project certainly strengthens the region's competences in areas where Europe is traditionally strong: automation, materials engineering, automotive, aviation. Collaboration with players like SAP and Siemens suggests the AI factory could become the missing link between mature industry and newer generative models. With room for both pre-training of specialized models and secure inferencing of customer data. This is the foundation for a more resilient value chain, from design, through virtual testing and robotics, to predictive maintenance.
However, the limitations should not be ignored. Europe still relies on hardware and software from an American manufacturerand the scale of investment does not eliminate the difference from American centers with hundreds of thousands of the latest chips. Analysts emphasize that this is only the first step. To build a full AI value chain made in Europe, further investments in energy, cooling, competences and local models will be needed.
Author: Grzegorz Kubera, journalist of Business Insider Polska
