Deutsche Telekom is injecting €250 million into its Munich-based AI data center to integrate Nvidia Blackwell B300 GPUs, aiming to scale capacity to 20 megawatts. This expansion targets industrial-grade workloads, moving beyond consumer AI hype to focus on tangible infrastructure for robotics and simulation.

Industrial Focus Over Consumer Hype

Launched in February, the facility already houses nearly 10,000 processors serving manufacturers who need to train robots, simulate trucks and factories, and run artificial intelligence models using proprietary industrial data. Unlike the consumer AI boom, this infrastructure is designed for proprietary industrial data, leveraging Germany’s strong manufacturing base to compete in a sector where Europe has previously lagged behind in previous waves of artificial intelligence and consumer online platforms.

Power Constraints and Chip Selection Strategy

The new capacity of 20 megawatts is double that of a conventional data center but remains significantly below the 100-plus megawatts consumed by the largest super-sized facilities. Deutsche Telekom CEO Tim Höttges emphasized that Germany should "not just be a user of artificial intelligence, but should create value and prosperity itself." This strategic positioning comes on top of the roughly €1 billion already committed by the company and its partners for Germany’s first industrial AI factory.

The Economics of Uncertainty

Infrastructure builders face a tough balancing act in chip procurement. Deutsche Telekom has moved to add its rare, high-end Blackwell processors that were in high demand, while delaying the installation of Nvidia RTX chips, whose utilization was limited to around 40% in May despite initial expectations for industrial AI. This situation highlights a key challenge for Europe in the field of artificial intelligence: providers are forced to guess which of the rapidly evolving chips customers will ultimately prefer and how much they are willing to pay for them, given the unpredictable nature of industrial AI adoption.

Key Takeaways

  • Deutsche Telekom is investing €250M to add Nvidia Blackwell B300 GPUs to its Munich facility.
  • The upgrade aims to reach 20 megawatts of power capacity, double the typical data center load.
  • The facility supports industrial use cases like robotics training, digital twin simulations, and large-scale model processing.
  • Utilization rates for Nvidia RTX chips were low (~40%), prompting a strategic pivot to Blackwell.
  • This investment is part of a larger €1 billion commitment to Germany’s industrial AI factory.

The Bottom Line

Deutsche Telekom's pivot to Blackwell GPUs signals that Europe is finally prioritizing specialized industrial AI over generic cloud capacity, but the struggle to predict chip utility shows the infrastructure race is still fraught with financial risk.