Germany Needs a Dual-Track Strategy for AI Sovereignty OpenAI GPT 4o Image
cepStudy

Artificial Intelligence

Germany Needs a Dual-Track Strategy for AI Sovereignty

Published on 11.17.2026

Dr. Anselm Küsters, LL.M.
Dr. Anselm Küsters, LL.M.
  • To counteract deepening technological dependencies, the cep is calling for targeted investment in industrial AI and long-term research into Germany’s own high-performance AI models.
  • The next AI revolution will take place in factories. Germany possesses key prerequisites for physical AI, such as domain expertise and unique manufacturing and maintenance data.
  • A key aspect of Germany’s AI problem is lacking utilisation of existing solutions – ranging from base models and inference and training code to benchmark datasets.

To safeguard its technological and economic capacity to act, Germany must fundamentally change its approach to Artificial Intelligence. A new study by the Centre for European Policy (cep) recommends a two-pronged strategy: targeted investment in industrial AI applications and the long-term development of its own cutting-edge technological capabilities.

cepStudy

The study analyses the interplay between energy, chips, infrastructure, models and applications, conceptualizes them as an "AI burger" held together by the regulatory framework. For AI middle powers, the important thing is not to produce every ingredient domestically, but to ensure that one is not dependent on individual suppliers for those that are strategically important: “Germany is far too dependent on other countries for key AI components, particularly high-performance computing infrastructure and leading AI base models,” says AI expert Anselm Küsters, who carried out the study on behalf of the Friedrich Naumann Foundation. Germany has 500 times less AI computing capacity than the US. The model calculations in the study also suggest that the country’s comparatively high electricity prices and slow connection to the grid are among its underestimated disadvantages as a business location. “At the same time, the German economy possesses strategic strengths that have not yet been sufficiently utilised in the international AI competition.”

The study identifies significant opportunities for combining application-specific AI with industrial production processes. Germany has high-quality data sets and technical expertise that could give it decisive competitive advantage in developing specialised AI systems and so-called physical AI. According to Küsters, however, focusing exclusively on specialised industrial applications – which are currently receiving a lot of political attention – would be short-sighted: “Germany and Europe must be in a position to simultaneously develop, understand, and critically evaluate powerful AI models themselves.” The progress made by Chinese AI companies and South Korea’s successes in AI competitions demonstrate that technological catch-up is possible even with limited resources. “We therefore need reliable, long-term research structures rather than short-term funding programmes and complex approval procedures. It is only through constant experimentation and attempts to keep pace with technological developments that the necessary expertise, ideas, and research communities can flourish,” Küsters argues.

Another key finding of the study is that technological autonomy does not solely arise from the promotion and development of European AI solutions. “A European AI solution that nobody uses does not create sovereignty. What matters is not whether alternatives exist, but whether it is realistic to switch to them,” says Küsters. High switching costs, dependence on dominant cloud and software platforms, a partial lack of technical expertise among small and medium-sized enterprises, and high regulatory uncertainty still make it far too difficult for companies and public institutions to switch to European alternatives. Therefore, Küsters calls for greater support for the practical application of existing European AI technologies, particularly those from the open-source sector.

Download PDF

Germany Needs a Dual-Track Strategy for AI Sovereignty (publ. 11.17.2026) PDF 3 MB Download
Germany Needs a Dual-Track Strategy for AI Sovereignty