DE

AI Sovereignty
How Germany Can Close the AI Gap

A new study by the Naumann Foundation reveals that Germany has significant ground to cover regarding AI infrastructure – yet, at the same time, major opportunities lie ahead.
Netzwerktechniker installiert Kabel in einem modernen Rechenzentrum.

Netzwerktechniker installiert Kabel in einem modernen Rechenzentrum.

© Foto: RodDIYguy via Pixabay

A new study by the Friedrich Naumann Foundation for Freedom highlights the substantial gap Germany must bridge in terms of AI sovereignty and infrastructure. When measured against economic output, the United States possesses approximately 94 times more recorded AI computing capacity. In absolute terms, US capacity is more than 500 times greater than Germany’s. However, computing power alone is not enough to secure a sovereign position in the global AI race.

The study ‘Germany Needs a Dual-Track Strategy for AI Sovereignty’ – authored by Anselm Küsters of the Centre for European Policy (cep) for the Naumann Foundation’s Global Innovation Hub in Taiwan – examines Germany’s standing across the entire AI value chain, from energy and chips to data centres, models, applications, and regulatory frameworks.

Anselm Küsters proposes a dual-track strategy. First, Germany requires targeted, long-term funding to develop its own high-performance AI models – not to catch up with the US, but to ensure it retains the ability to evaluate and adapt systems developed elsewhere. Second, it needs to focus on the broad integration of existing AI into industrial practice, which is where Germany’s true strengths lie.

Limited computing capacity makes it difficult for Germany to develop advanced AI models independently or to apply AI with full sovereignty. High electricity costs, slow grid connections, and a reliance on foreign technology providers further exacerbate the problem.

Additionally, the study identifies significant legal uncertainty as a major barrier to the more intensive use of AI. At the same time, the study points to concrete opportunities: Germany can leverage its industrial experience and high-quality production data to become more technologically independent.

High electricity costs and slow grid connections are hindering expansion

Energy costs represent a significant competitive disadvantage for the country. A model calculation for the continuous operation of 25,000 AI accelerators – including the associated systems – illustrates the impact of this disparity: in Germany, annual electricity costs would amount to approximately 29–32 million euros, compared to about 12 million euros in Finland. However, the availability of sufficient power grid connections can be an even greater challenge than the price itself. If securing new grid connections takes longer than procuring the computing hardware and constructing the facilities, the entire expansion process is delayed.

Germany’s presence among world-class AI models is also limited. The Epoch AI database records roughly 20 significant models from Germany over the entire history of the field, compared to several hundred from the United States. It is not a lack of companies – the online directory EuropeanTechMap lists more than 50 German AI firms – but rather a failure to translate this foundation into the development of proprietary ‘foundational models’.

Furthermore, there is legal uncertainty. While effective regulation of AI and data protection can, in principle, be an advantage for AI development, the rules must be clearly defined. The associated compliance burden weighs heavily on small and medium-sized enterprises (SMEs), which are less likely to have dedicated legal departments or specialised teams capable of integrating new regulatory requirements into their operational processes.

Germany’s opportunities lie in its industrial base

Despite these weaknesses, the study identifies promising starting points. Germany possesses industrial experience, a skilled workforce, and high-quality data from production, machine operation, maintenance, quality control, and logistics. These data can help in developing AI for specific tasks where general-purpose language models alone are insufficient.

Further opportunities lie in software that enables AI systems to carry out multi-stage tasks, as well as in tools that coordinate various models and applications. In an industrial context, in particular, such systems must reliably integrate with machinery, corporate data, and existing workflows. Here, Germany can combine its technical expertise with in-depth knowledge of specific production processes.

Steps towards greater AI sovereignty

The analysis points to an agenda that strengthens domestic technological capabilities while simultaneously facilitating the practical deployment of European solutions:

1. Expand computing power and accelerate energy supply

Germany must pair the expansion of AI data centres with faster permitting processes and more reliable grid connection procedures. Flexible electricity contracts should reward operators for scheduling shiftable computing tasks during hours when power supply is abundant. Furthermore, Germany – in collaboration with European partners – should explore and test methods for interconnecting multiple data centers to handle large-scale training tasks, thereby optimising the use of existing capacity.

2. Provide long-term funding for cutting-edge research and foster competition

The study recommends a research programme featuring multi-year funding commitments and streamlined allocation procedures, with the involvement of the Federal Agency for Disruptive Innovation (SPRIND). This does not necessarily require massive sums; the study cites South Korea’s competition to develop indigenous foundational AI models – supported by a total government budget of approximately 350 million US dollars – as a case in point. It demonstrates that targeted support for talent and competing teams can yield ambitious results, even with significantly smaller budgets.

3. Promote open technologies and facilitate vendor switching

For large-scale AI contracts, public-sector clients should mandate open interfaces, data portability, and realistic options for switching providers. Competition authorities should investigate whether the bundling of cloud services, development tools, and AI offerings hinders market access for other providers.

Concurrently, there is a need to support transparent training processes and independent model testing. The study notes that the federal government spent at least 630 million euros on proprietary software licences in 2025. A portion of these funds could, in the future, be redirected towards open infrastructure and its professional maintenance. The goal is to develop sustainable, long-term alternatives and to increase the genuine freedom of choice for both public and private users.

4. Support small and medium-sized enterprises (SMEs) with implementation, skills development, and legally compliant operation

Germany must assist companies in transitioning from isolated experiments to the reliable, operational use of applications. This includes practical training for managers, employees, and works councils, as well as concrete support for restructuring workflows.

Meeting legal requirements should be made easier through clear guidelines, reusable templates, accessible testing facilities, and industry-specific resources. Furthermore, the study recommends investing in technical tools that can verify AI systems, monitor their usage, and ensure the traceability of data origins. Such measures can foster the security and trust necessary to facilitate the adoption of AI.

*Frederic Spohr heads the Taiwan and Korea offices of the Friedrich Naumann Foundation for Freedom (FNF) and leads its Global Innovation Hub in Taipei