As part of the European Union's efforts to keep up in the race of developing artificial intelligence (AI), the bloc recently announced a €200-billion ($208-billion) funding program for the breakthrough technology.

It comes as US President Donald Trump launched multibillion-dollar AI projects in January and China made headlines with its own large language model (LLM) called DeepSeek.

Björn Ommer, a professor of artificial intelligence at Ludwig Maximilian University of Munich (LMU), told DW that AI is driving the rapid emergence of new business models. But, he warned, "missing out on this will cost us dearly, so we need to step on the accelerator."

Ommer, one of Germany's leading AI specialists, was the mastermind behind German technology startup Black Forest Labs. The company has developed an image generator that has convinced even US venture capital providers and Elon Musk's social media platform X that integrated the AI tool.

Opportunities 'beyond large foundation models'

But Black Forest Labs is more of an outlier in the world of so-called foundation models, which are dominated by US and Chinese companies. A foundation model is an AI neural network — trained on mountains of raw data, generally with unsupervised learning — that can be adapted to accomplish a broad range of tasks.

The market for large foundation models is very much in the hands of AI heavyweights such as OpenAI's ChatGPT, Google's Gemini, Perplexity, Midjourney and DALL-E. China's DeepSeek is also considered an efficient foundation model for text processing, according to media reports.

German company Aleph Alpha, based in Heidelberg, also attempted to launch a foundation model. However, the results failed to impress, leading the company to pivot toward customized AI applications for organizations, businesses and government agencies.

Ommer believes the opportunities for German companies in AI will lie "outside of foundation models." AI applications for medicine, customer communication and specialized industrial sectors, he said, could mean profitable business in the future.

'At the forefront of research'

Katharina Morik shares this perspective. The now-retired researcher co-founded the Lamarr Institute for Machine Learning and Artificial Intelligence in Dortmund, Germany. She views large language models as a gateway to AI — similar to the emergence of the smartphone which gave users access to multiple applications.

"We use language models as an interface, providing access, for example, to a robot, a booking system, or a recommendation platform," she told DW, adding that Germany's powerful and export-oriented small and mid-sized companies could have an advantage in developing these specialized applications. "When it comes to research, we are at the forefront. We have fantastic talent."

Morik also noted, however, that many of Germany's AI talents are currently leaving for the US after graduating. "We train brilliant minds, but then they leave because we only offer them temporary contracts," she said.

Morik, part of the scientific advisory board of German data analytics firm RapidMiner, said she has gained firsthand experience with the brain drain. The company's analytic tool was developed at the Technical University of Dortmund but later grew in the US where it was ultimately sold.

In Germany, she added, companies just weren't willing to pay for RapidMiner's services because they "expect to get such services for free."

AI has massive potential for manufacturing

Morik believes a broader "cultural shift" is needed to make Germany a global AI hub. "I don't know why there's so little willingness to experiment here," she said.

One area where Germany is already leading is in intelligent, distributed and embedded systems. These involve small sensors that monitor processes and can intervene autonomously, such as adjusting manufacturing operations.

Morik thinks this is "incredibly valuable" for the manufacturing sector, even though it may not "grab headlines."

As an example, she cited a collaboration with Dortmund-based pump manufacturer Wilo where her company identified 82 possible AI applications to improve operations in Wilo's smart factory. "The potential of AI is huge […] if we fail to tap into it, we will become an industrial museum," warned Morik.

In this regard, Morik is also critical of OpenAI's recent announcement to open a Munich office. She wants German talents to "stay here" and "our companies and industries to benefit from AI."

That's why she believes it's crucial for Europe to continue developing its foundation models and to "harness its vast troves of industrial data."

European AI is more trustworthy, she said, which is why successful AI models such as the German Fraunhofer Institute's Teuken-7B model and the AI applications developed by French company Mistral must be expanded upon.

LMU Munich's Björn Ommer also sees national and European initiatives as essential. While Germany may not need to create "the ultimate AI that outperforms everything," he said it's still important to continue work in the field of foundation models.

"We don't know how the big tech companies will behave," he said. "If access to technology suddenly becomes restricted, sovereignty will be crucial."

This article was originally written in German.