Companies will be able to book the computing power of the "AI factory" and use it as required as early as the first quarter of 2026.

The construction of the factory will increase AI computing power in Germany by around 50 percent. In combination with Deutsche Telekom's network and the T Cloud, an ecosystem is being created for German and European companies that offers AI sovereignty (the ability to use AI independently, securely and according to one's own rules).

Industrial AI Cloud: six months from idea to reality

"Mechanical engineering and industry have made this country strong," says Tim Höttges, CEO of Deutsche Telekom. "But we are also being challenged here. AI is a huge opportunity. It will help to improve our products and strengthen our European strengths. In just six months, we will turn an idea into real AI computing power, made for Germany."

"Germany's strength in engineering and industry is legendary and is now being further enhanced by AI," adds Jensen Huang, founder and CEO of NVIDIA. "With the world's first Industrial AI Cloud and one of the largest deployments of GPUs in Germany, we are bringing NVIDIA AI and robotics here to usher in a new era of industrial transformation."

Industrial AI Cloud: application scenarios and productivity gains

The Industrial AI Cloud meets the high and diverse requirements of the industry. Numerous partners and customers have already expressed their interest.

For example, developers can create 3D digital twins of real factories by integrating NVIDIA Omniverse libraries for virtual planning, simulation, testing and improvement of factory design. A 3D twin is an exact digital replica of a real plant or machine. This increases efficiency and enables decisive productivity increases - for example for cars or airplanes in digital wind tunnels (simulations for flow and aerodynamic testing) or for virtual crash tests.

Another area of application is the development of robots through simulation-based learning and validation. One of the first users is Agile Robots: the robots are used in the production processes of car manufacturers and electronics producers. They carry out mechanical assembly and polishing work with maximum precision and adapt their movements in real time. The H10-W robot is even used to install server cabinets at the Munich site.

Large Language Models (LLMs) can also be developed and trained at a high level. LLMs are AI models that can analyze large amounts of text and generate speech or provide information based on this. Deutsche Telekom partner Perplexity will use the Industrial AI Cloud to provide secure AI inference locally for German users and companies. Inference means here: The AI provides answers or decisions based on the trained models in real time. True AI sovereignty means that the inference layer generates answers and intelligence within Germany.

Support, integration and security

Deutsche Telekom offers customers comprehensive services relating to the migration to the Industrial AI Cloud - i.e. the conversion and integration of existing systems to the new AI infrastructure. This includes the connection, introduction of AI solutions, security solutions and process integration.

SAP and Deutsche Telekom will also create a secure, sovereign and powerful digital infrastructure for public institutions and security applications. Deutsche Telekom will provide the physical infrastructure, while SAP will supply the SAP Business Technology Platform and applications - including modern AI technologies. In future, all digital solutions for these target groups will be developed on the joint "Germany stack" in order to guarantee the highest standards of data protection, security and reliability. A stack here refers to the combination of software, applications and infrastructure that work together.

Data, facts and figures about the Industrial AI Cloud

An existing data center is being completely renovated in order to build the Industrial AI Cloud together with partner Polarise. This is planned:

• More than 1,000 NVIDIA DGX™ B200 systems and NVIDIA RTX PRO™ servers
• Up to 10,000 NVIDIA Blackwell GPUs
• Computing power: 0.5 EFLOPS (Exa Flops, i.e. one billion billion calculations per second)
• Storage capacity: around 20 petabytes (1 petabyte = 1,000 terabytes)
• Connection: four 400 GB fiber optic connections
• Server park: several thousand square meters, fully energy-efficient and in accordance with the highest security and quality standards
• Infrastructure: 75 kilometers of fiber optic cable, partly laid by Agile Robots

The AI factory is scheduled to go into operation in the first quarter of 2026.

The expansion of the Munich data center is taking place independently of EU projects to build several AI gigafactories in Europe. The cooperation between Telekom and NVIDIA, together with other partners, is referred to as a "speedboat" - a fast innovation hub for companies that want to develop and use artificial intelligence today.

The cooperation shows how industry, telecommunications and technology providers can jointly create AI sovereignty "Made in Germany" and at the same time open up new opportunities for German companies, jobs and growth.

Source: Deutsche Telekom press conference