Artificial intelligence is having a run. Within just a few years, it has developed into a cross-sectional technology that leaves practically no sector of the economy and no area of society untouched. “This meaning didn’t even exist back when we awarded the prize for the first time,” recalled Moritz Seyffarth, deputy editor-in-chief of the PREMIUM GROUP at Axel Springer, in his introduction to the awarding of the eighth German AI Prize in Berlin.
At the time of the first award ceremony, AI was primarily something for initiates and nerds, but today it has become part of many people's everyday lives. According to the Federal Ministry of Economics, 260 billion US dollars in venture capital flowed into the industry worldwide in 2025. “This makes it clear what expectations are associated with this technology,” emphasized Economics Minister Katherina Reiche in her keynote.
However, these expectations are primarily aimed at the USA and China. Almost three quarters of the 260 billion US dollars were invested in the United States, while the People's Republic now registers three quarters of all AI patents in the world. Europe seems hopelessly left behind.
The Economics Minister didn't quite see it that way. “It is a task for us in Europe and for us in the Federal Republic of Germany to tackle our own opportunities more decisively,” said Reiche. “We have the brains, we have the talent, we also have the capital and we have the technologies.” Europe should work on its sovereignty and create the regulatory conditions so that AI applications can develop. When asked about a possible AI tax, Reiche said: “We can talk about that when things are going well.”
In her opinion, events such as the awarding of the German AI Prize serve to demonstrate the possibilities of the technology. For Europe and its industrial center Germany, these primarily exist where AI is translated into concrete applications - less in the race for the most powerful basic model.
The USA and China have largely staked out this territory for themselves. “The economic success of AI is also determined by what comes from it,” says Reiche: “Better products. More efficient processes, new business models and additional added value.”
Reiche sees Germany and Europe's great opportunity in this physical AI, when AI is integrated into manufacturing machines and processes and improves performance and efficiency or enables completely new products and solutions. “One in seven machines worldwide comes from Germany,” said the minister, “and this contains engineering art and a great treasure of industrial data, historical, technical, process-oriented knowledge.” Politicians from large companies and medium-sized companies hope to use AI to uncover this treasure and make it usable.
This year's user prize was awarded to Robert Bosch GmbH, one of the technological heavyweights of the German economy, precisely for such use of AI in manufacturing. The company received the award because it not only demonstrated an AI application in a pilot project, but also implemented it on a large scale in manufacturing. “It is already running in over 1,000 production lines and creates corresponding added value there,” emphasized Marco Huber, university professor in Stuttgart and scientific director for digitalization and AI at the Fraunhofer Institute for Production Engineering and Automation (IPA), in his laudatory speech.
Specifically, it is an application for detecting anomalies in the manufacturing process that indicate possible problems before they actually occur. “Production lines are now optimized so that there are typically very few errors,” explained Huber. This is exactly what makes it difficult to train an AI with a sufficient number of examples of errors. Bosch's AI model learns how a normal manufacturing process works based on continuously collected sensor data. “If she can do this well enough, then she can notice strong deviations at an early stage,” says Huber. This helps to further reduce the production of rejects.
“We are improving our products and services, but we are also increasing productivity,” explained Tanja Rückert, Managing Director of Digital Business and Services at Robert Bosch, in her acceptance speech. The group wants to optimize the entire value chain from product development to production to logistics using AI. Rückert referred to the number of AI patents the company has, 2,000 in the past five years alone. For Rückert, this is a sign of the Germans’ unbroken innovative strength: “People shape the technology.”
The handling of limited training data also plays an important role in this year's innovation award. Prize winner Frank Hutter, computer science professor at Freiburg's Albert Ludwig University and managing director of the start-up Prior Labs, co-developed an AI model that can provide good predictions even if there are only a few training examples available for a new task.
“We have incredibly strong talent, we have incredibly strong start-ups,” explained Frank Hutter in his acceptance speech. “If we boldly invest in cutting-edge research and then simplify spin-offs, then the hope will also become a plan to achieve this,” he said, alluding to the Federal Minister of Economics’s hope for AI growth for Germany.
Hutter received the innovation award for his research in the field of AI-supported evaluation of structured data, such as tables. "The currently most advanced AI models have been trained primarily on texts, on unstructured data. This also explains why these models are particularly good at generating unstructured data and supporting us in working with them," explained Antonio Krüger, computer science professor and managing director of the German Research Center for Artificial Intelligence (DFKI) in Saarbrücken, in his laudatory speech. These AI models are very powerful at interpreting and producing text or images. However, when evaluating or building structured data, they often reach their limits.
The model that Hutter developed with scientists at the University of Freiburg is different: TabPFN was trained in advance with a huge amount of artificially generated data sets to recognize patterns and connections in tables. It can then transfer this knowledge to new tasks without having to undergo extensive training. It can therefore read and understand tabular data with comparatively little effort and derive predictions from them, which is important, for example, in medical diagnoses or financial analyses, for scientific experiments or operational decisions.
A particularly important point for the jury was that TabPFN also works with relatively little data. This can be crucial, especially in medical research or specialized scientific experiments.
This year's award winners are representatives of an innovative AI scene in Germany. Europe may be well behind the USA and China when it comes to basic models, but the continent can still exploit its strengths in other areas of AI application.
“The Foundation Models train may have left,” says DFKI Managing Director Antonio Krüger, “but a lot of trains are leaving the station at the moment.” It will be crucial that Europe chooses the AI fields that suit its strengths.
