Europe is falling behind in the AI race. U.S. private investment in AI rose to $285.9 billion in 2025, according to Stanford University figures, dwarfing the $20.9 billion invested in Europe. The U.S. has higher AI adoption rates, a greater number of data centers, and the most valuable AI companies in the world.
In total, eight U.S. tech companies have surpassed trillion-dollar valuations, with OpenAI and Anthropic poised to join their ranks. Currently, the closest Europe comes to a trillion-dollar business is ASML, which has a market cap of $670 billion.
Europe’s struggle to build a trillion-dollar market-cap company has been on the mind of Peter Koerte, CEO of Siemens’s Smart Infrastructure division—the most profitable of Siemens’s four core businesses. “We have really smart people in Europe,” he says. “But the U.S. and China have a massive advantage because of their huge domestic markets. In Europe, it’s disproportionately harder to scale because of the different languages, regimes, and political systems.”
Europe’s industrial data advantage
One place where Europe could have an upper hand, he says, is in the development of AI for industrial processes. Siemens’s Industrial Foundation Model, which is still in development, has the potential to shorten engineering cycles by up to 40%, according to Koerte, and he foresees use cases in the automotive and aerospace industries. Siemens plans to invest more than €1 billion in industrial AI over the next three years.
Unlike LLMs, which are primarily trained on language, industrial models are taught on manufacturing and engineering data for use in those fields. “In engineering, you need to be precise. If the AI hallucinates or makes up a calculation, you will have a problem,” says Koerte. “When we try to use large language models in engineering and production, it doesn’t work—words are much more imprecise.”
But securing sufficient training data for industrial models can represent a challenge. Koerte says this is where the longevity of Europe’s leading manufacturers could be an advantage. More than half of the companies on the Fortune 500 Europe are over 100 years old. “We have the largest install base. All the data is available within the pharma, automotive, and chemical companies that have been operating in Europe for decades,” Koerte says.
Persuading these companies to part with their data may prove more difficult. “We’ve been very good at educating organizations that data is the new oil,” Koerte says. “BMW or AstraZeneca is not going to put their data on the internet, so they need to trust others to train these models using their data.”
This is where Koerte believes Siemens, which was founded in 1847, has another edge. “We have decades-old relationships with most of the companies we work with,” he says. “They trust us to take utmost care of their data and can rely on us if something breaks.”
So far, Siemens has managed to persuade companies to share their data in return for access to its AI model and the ability to request new use cases. The data is used only for training, and ownership is retained by the company that created it, Koerte explains.
Talent, partnerships, and the limits of sovereignty
The other issue is talent. U.S. institutions employ 59% of the world’s elite AI researchers, according to research platform MacroPolo Archive’s Global AI Talent Tracker. But last year, Siemens poached Amazon’s vice president of generative AI, Vasi Philomin, to be its head of data and AI. Manu Parbhakar, a former Amazon Web Services director, also joined Siemens to lead its Silicon Valley–based strategy and partnerships team. “You have to write a bigger check than you usually would in Europe,” Koerte says. “But you need to have the experts to build this.” U.S. partnerships have also been crucial. Siemens is currently working with Nvidia to build an industrial AI operating system.
Siemens is not the only company attempting to apply AI to physical challenges. Amazon founder Jeff Bezos’s latest startup Prometheus is using AI to assist engineers in the design and manufacturing of a range of devices—from computers to automobiles to jet engines. “Speed is of the essence. But I think we are in a good position,” Koerte says.
Siemens’s tighter ties with the U.S. come at the same time the European Commission is pushing for a split from American Big Tech. The EU relies on non-EU countries for 80% of its digital infrastructure and services, and the Commission’s tech sovereignty package is aimed at supporting the creation of homegrown alternatives.
Although Koerte believes Europe needs to develop greater AI capabilities, complete sovereignty remains a pipe dream. “To put it simply, there’s no sovereignty. Not in the U.S. Not in China. Not in Europe,” he says. No country has all the resources required to develop AI. “If you want to create high-performance chips, you need the ultraviolet machines made by ASML, in the Netherlands. It sources many parts from German companies, and those businesses’ suppliers are spread across the world,” Koerte adds.
Siemens CEO Roland Busch has warned that a “digital iron curtain” risks slowing technological progress as governments seek greater control over digital supply chains. Rather than pursuing technological isolationism, Koerte advocates greater collaboration across borders. Only through such cooperation, he argues, can we fully realize AI’s promised efficiency gains.
