"Is Germany well prepared for the next AI breakthrough? Unfortunately not.
"The Government knows AGI is coming" - this was the title of a recent interview in the New York Times with Ben Buchanan, the White House's top AI advisor during Joe Biden's presidency. AGI stands for "Artificial General Intelligence," meaning artificial intelligence (AI) that is cognitively equal to or superior to humans in virtually every respect. Buchanan describes how his team in the White House worked to prepare the American government for AGI. And that he expects AGI to arrive during Donald Trump's presidency.
Artificial intelligence that achieves at least human-level capabilities within the next four years - this is a scenario that many in expert circles now consider possible. Nobel laureate and AI researcher Geoffrey Hinton shares this view, as does his colleague Yoshua Bengio and the CEOs of the world's leading AI companies. If they are right and this scenario occurs, would Germany be prepared? Is the federal government preparing for it?
It doesn't look like it. A stringent political AI strategy is not yet apparent. Instead of actively shaping the transition to the AI era and preparing for what is possibly the most profound technological transformation since the industrial revolution, the new federal government has so far treated artificial intelligence more like an exciting future topic among many digital topics. Elsewhere, however, AI-based value creation has not only been taking place for a long time. Among decision-makers in Washington, London, or San Francisco, the topic of AGI is also omnipresent.
Many political strategies and business models there are based on the assumption that intelligence will soon be available, replicable, and scalable on an almost unimaginable scale. The impact of such a scaling of artificial intelligence on the economy, scientific progress, and global power relations will be enormous.
There is uncertainty about exactly when AGI will be achieved. Some experts are less optimistic than Hinton, Bengio, and entrepreneurs like Sam Altman of Open AI or Dario. Amodei from Anthropic. They expect that it will take at least one more major technological innovation before machine intelligence will catch up with human intelligence in every respect. But even if it takes another ten to twenty years, Germany is completely inadequately prepared.
While other countries are mobilizing enormous sums to expand AI infrastructure, far too little is happening in Germany, and it's happening far too slowly.
In the USA, for example, $500 billion is to be invested in American AI infrastructure over the next four years as part of the so-called Stargate project alone. Fear of competition from China, currently in second place in the global AI race, is a driving factor. At the same time, Donald Trump is concluding deals for massive data centers with the United Arab Emirates and Saudi Arabia, which are likely to soon join the ranks of the most important AI nations.
Europe is a bystander, and even within Europe, Germany is still one of the more sluggish AI players. The AI gigafactory announced in the coalition agreement – a data center worth three to five billion euros, with some of the financing still unclear – lags far behind the efforts of other countries. For example, Great Britain is currently introducing "AI growth zones" where data centers can be built particularly quickly, and in France, President Macron has made AI a top priority and announced investments in at least the tens of billions.
There is hope that important competencies are being pooled in the new Digital Ministry and that there is a spirit of optimism. Through smart and decisive action, Germany could find and maintain its place in the AI world.
Three priorities would have to be at the heart of such an AI policy.
1. Strong and resilient infrastructure
AI relies on enormous amounts of computing power. Data centers with special AI chips such as GPUs (Graphics Processing Units) are thus becoming a core strategic resource. Germany currently has individual flagship projects such as the JUPITER supercomputer at the Jülich Research Center, which is equipped with approximately 24,000 GPUs. By international comparison, however, this is more of a third-league level: large tech companies in the USA often have several hundred thousand GPUs alone.
The coalition agreement of the new federal government now envisages expanding at least one supercomputer into an "AI gigafactory" with 100,000 GPUs. This would be a step towards more AI sovereignty. However, for this infrastructure expansion to be truly successful, several key points must be considered.
First, computing capacity is needed not only for the development of new AI models ("training"), but also for the operation of existing AI models ("inference"). The latter will be essential, especially in a world with AGI, in order to be able to widely use the available, extremely powerful AI for all possible processes in industry and administration. Countries whose infrastructure is equipped for such scaling and can bring intelligence into comprehensive applications could leave the rest of the world behind. Therefore, Germany needs as many data centers as possible that can be used for inference – and that make the country more independent of foreign cloud providers.
Second, the construction of data centers, whether for training or inference, must be carried out at high speed. Approval procedures must be accelerated, and the necessary energy infrastructure must be expanded with the highest priority. This is because AI data centers require enormous amounts of electricity and maximum grid stability. Currently, the German energy grid would be overwhelmed by the operation of several such data centers. While efficiency gains are expected in the coming years, reducing the energy consumption per AI deployment, increasing demand for AI is likely to lead to a further, possibly rapid, increase in energy demand overall. If the German government is serious about its goal of making Germany an "AI nation," it must – as announced by Economics Minister Katherina Reiche – prioritize expanding the power grid and ensuring affordable energy. The "AI Growth Zones" announced by the British government demonstrate how this can work in practice: Together with local authorities, the government identifies promising locations that are particularly suitable for the construction of data centers and the strengthening of the necessary energy infrastructure. Simplified planning processes then help increase the available power capacity locally.
Thirdly, new AI data centers should be built with a particularly high level of security. AI models will increasingly become the target of massive cyberattacks financed by hostile states in the coming years. Even the most powerful AI models are essentially just a collection of many numbers, so-called parameters. If hackers succeed in stealing these parameters, they have access to the same AI capabilities as the developers – and can use them for military purposes, for example. Secure data centers are therefore of outstanding economic and geostrategic importance. However, according to a report by the consulting firm Gladstone, no AI data center is currently adequately protected against sabotage by foreign states. Establishing effective protection could take years. The security of highly sensitive information must therefore be planned from the outset when building AI data centers.
Last but not least, a strong ecosystem of companies is needed that convert computing power into innovation and growth. The fact that leading German corporations, including Siemens and SAP, want to join forces to build an AI gigafactory for the development and use of AI is a positive sign. Overall, however, the German economy is often still hesitant. The German AI startup scene could also be more vibrant, despite positive examples such as DeepL, Parloa, and Helsing. Policymakers can play a proactive role here: Regulations and real-world laboratories implemented with a sense of proportion help simplify innovation and put AI into practice. Startups and small and medium-sized enterprises also need easier access to capital and public computing power. Above all, the federal government should facilitate access to private capital, for example, by providing financial incentives for pension funds and companies to invest in AI startups.
If these points are taken into account, Germany can lay the foundation for an AI future with new data centers in which hardly any economic process can function without AI. This is explicitly not about self-sufficiency, as that is simply not realistic: there is currently no way around non-European components from countries such as the USA or Taiwan. The federal government must therefore work with the EU to secure the supply of high-performance chips through trade policy. The White House reportedly intends to renegotiate with each country soon about whether and how AI chips will be exported from the US. A positive outcome for Germany should be a priority for the new federal government.
2. Leveraging economic strengths
Even if all of the measures mentioned so far were consistently implemented, Germany as a location for AI would lag far behind the rest of the world’s leading AI nations for the foreseeable future. It is all the more important that Germany manages to compete above its own weight class. This requires cleverly leveraging its existing strengths. German companies can occupy important niches and thus make themselves irreplaceable at key points in the global AI value chain. Three core areas offer themselves:
First, the broadly diversified German industry – from mechanical and plant engineering to robotics and medical technology – is likely to be very well suited to transforming the "intelligence on demand" available in an AI world into real value creation, even if the strongest AI models continue to be developed in the USA. Even if an American AI system is used to design novel vaccines or high-tech devices, prototypes still need to be built, improved, and products mass-produced. Germany, in particular, has a competitive advantage over countries like the USA thanks to its industrial capacity. The more digital the production processes of German companies are and the more closely they are integrated with AI, the more effectively this trump card can be played.
Second, German companies have a wealth of industrial data. With this data, they can develop highly specialized models – as Siemens recently did, for example – that make industrial processes more efficient and resource-efficient. There is enormous potential here, especially in energy-intensive sectors.
And third, German universities train excellent AI talent and have strong basic research capabilities. Nevertheless, Germany has been losing many AI talents to the US for years. This brain drain must be stopped, and Germany should actively recruit AI talent currently working abroad. For many of them, it's not primarily about the money, but rather about working on exciting projects and making a difference in the world through innovative products. The conditions for this must be created now: through easier spin-offs, more capital, practice-oriented implementation of regulations such as the European AI Regulation, and unbureaucratic access to computing power. With a rapid expansion of computing capacity, German research could also venture into ambitious moonshot projects. To this end, Germany could also join forces with other democratic "middle powers" such as the United Kingdom, Canada, Australia, and other European countries to pool resources, thus working toward a "third pole" on the AI map beyond the USA and China.
3. Governmental Capacity
The fact that Germany has not yet tackled the challenges in the field of AI with sufficient rigor is likely also due to the lack of concentrated technological expertise within the government apparatus. For years, AI was treated as a secondary concern, distributed decentrally across the ministries. The new Federal Ministry for Digital and State Modernization (BMDS), with clear AI expertise, is a step in the right direction and should be able to assume a central AI leadership role within the federal government over the course of the legislative term. But that alone is not enough. AI must be viewed as a core strategic issue that goes far beyond the intersection with state modernization and affects every area of life.
State capacity is needed not only to make Germany competitive again in the field of AI, but also to avoid societal upheaval in the AGI era and safeguard national security. A report recently published by internationally recognized experts on behalf of the United Nations and 30 leading AI nations addresses, in addition to already visible dangers such as deepfakes, a number of other risks that could soon arise: AI-assisted cyberattacks, widespread job losses due to automation, or even a complete loss of control over increasingly autonomous AI systems.
If AGI were truly achieved in the next few years, such risks could become highly relevant during this legislative period. Germany would then be left without any security. And even if no risks to national security materialize, Germany would still be at acute risk of being completely left behind by other countries.
Perhaps the current moment is comparable to January 2020, when the coronavirus began to spread, but many decision-makers were not sufficiently aware that something profoundly disruptive, something world-changing, was happening. How can Germany avoid looking back on the summer of 2025 in a few years and realizing that it should have built the necessary resilience for the looming AGI era much more quickly and decisively.
A look at the United Kingdom shows how this could work. The British government began in 2023 to establish a world-leading AI Security Institute (AISI) as soon as possible, which will research the impact of AI on national security and advise the government. Germany should now follow suit and establish a German AI Security Institute. The new Digital Minister, Karsten Wildberger, could attach such an advisory institute to his ministry, and he should ensure that it is granted sufficient autonomy and is managed in an entrepreneurial and agile manner. Furthermore, "AGI Preparedness" should become one of the main goals of the National Security Council announced by Chancellor Friedrich Merz. It is crucial that the Chancellor makes AI a top priority and personally makes it clear to all stakeholders that he expects this issue to be addressed at the highest speed and given the highest priority. The only reason Great Britain currently has the world's best "state capacity" in the field of AI is because Rishi Sunak made this project a personal top priority and pushed it through, even in the face of resistance.
Germany urgently needs a strategy for the AGI era. It requires infrastructure and a wise use of the few assets our country has in the field of AI. And it requires "state capacity." Above all, however, political leaders must understand that with AI, we may find ourselves in a situation similar to that of January 2020 with Covid. "The US government knows AGI is coming," and many others know it too and are preparing for it. This must now happen in Berlin as well.
Monika Schnitzer is a professor at Ludwig Maximilian University in Munich and chair of the German Council of Economic Experts.
Daniel Privitera is Executive Director of the KIRA Center and lead writer of the International AI Safety Report.” [1]
The mountain gives birth to a mouse: a German AI Security Institute.
1. Wenn Intelligenz in kaum vorstellbarem Ausmaß verfügbar, kopierbar und skalierbar ist. Frankfurter Allgemeine Zeitung; Frankfurt. 16 June 2025: 18. Von Monika Schnitzer und Daniel Privitera
