New Report: AI Compute in Germany
Philip Fox, 14 October 2025
Germany is falling behind on computing power for advanced AI. Its share of global AI compute is currently less than 2% and likely to decrease even further in the next few years. Many countries have announced a more ambitious buildout, including the USA, China, the UAE, South Korea, Saudi-Arabia, the United Kingdom, France and India. Since AI adoption will predictably increase across the economy and public services, the lack of computing power puts Germany’s future competitiveness and geopolitical sovereignty at risk.
These are the key findings of a report, published today, by Dr. Philip Fox (KIRA), Prof. Monika Schnitzer (LMU Munich) and Daniel Privitera (KIRA).
The authors describe three possible compute strategies for Germany until the end of 2028:
Low ambition (850k H100-equivalents): focus on AI adoption in key areas
Medium ambition (3.4m H100-equivalents): focus on AI adoption across all economic sectors and public administration
High ambition (6m H100-equivalents): sufficient capacity to train and inference frontier models
The Federal Government will deliver on its goal of making Germany an “AI nation” only if it takes initiative now and supports the urgently needed expansion of AI infrastructure.
The report makes six recommendations:
Expand AI training and inference capacity according to strategic priorities
Create political enablers, e.g. centralized steering and trade diplomacy to guarantee GPU access
Secure cheap, reliable energy
Accelerate data center planning and permitting
Ramp up data center security
Strengthen the domestic AI ecosystem
Germany is lagging behind on AI, but it’s not too late for a step change. Boosting domestic compute capacity is a crucial (but far from the only step) in that direction. Germany should act now, while it’s still among the few countries in the world with enough economic and political capacity to change course.