AI patent race reveals Europe's digital sovereignty challenges: concerns lurk behind the lead

Now here's the translation:

In 2025, the global innovation landscape is mirroring itself through patent applications, mapping out the power shift in the digital economy. According to the latest data from the European Patent Office (EPO), computer technology remains the technical field with the highest number of patent applications in Europe, growing by 6.1% year-on-year; among these, artificial intelligence (AI) patent applications grew by 9.5%, while quantum computing achieved a rapid growth rate of 37.9%. These figures may appear to be mere fine-tuning of technical indicators, but what they reflect behind the numbers is Europe's increasingly complex competitive position under the proposition of "digital sovereignty."

AI Leads in Volume, but Growth Is Lackluster

Without a doubt, AI has become the core engine of computing technology transformation. Over the past decade, the number of AI patent applications accepted by the European Patent Office has grown to roughly 24 times what it was ten years ago. However, when we break down the 2025 growth figures, the situation is not entirely optimistic. Among the AI patent applications accepted by the EPO, applications submitted by U.S. applicants grew by 26.4% over the past year, while the growth rate for European applicants was only 2.6%. In other words, although Europe still maintains a slight lead in total volume with a share of more than one-third, that advantage has become razor-thin.

Within this overall picture, performance inside Europe is also highly uneven. Germany remains the backbone of European AI patent applications, despite a slight decrease of 1.1%; meanwhile, France (+19.0%) and the United Kingdom (+28.8%) have shown strong rebounds. This divergence serves as a reminder that European innovation is not a monolithic bloc; differences in industrial structures and research traditions across countries are shaping distinct technological paths.

At the same time, major economies from Asia are also reshaping the competitive landscape. Applicants from South Korea grew by 16.5%, China by 3.9%, while Japan fell by 8.6%. The three-way tug-of-war among the United States, Europe, and Asia is vividly reflected in the EPO's applicant rankings. Bosch, Siemens, Ericsson, Nokia, Philips, Thales, and the newly listed Orange together form the deep foundation of the European camp.

Segment Breakdown: Europe's Strengths and Weaknesses

Using aggregate data to obscure internal differences often leads to overly broad conclusions. AI is not a homogeneous "technology basket," but rather a complex landscape composed of different technical subfields, each with its own capital intensity, research cycles, and competitive structure.

In the field of neural networks—the "heavyweight" segment accounting for more than half of AI patent applications—Europe's share has long remained at around 35%, but 2025 saw some subtle changes: as U.S. applicants renewed their efforts, Europe's share dipped slightly. However, in terms of growth trends, after experiencing a plateau since 2022, this field turned upward again in 2025. Among the leading applicant companies, Alphabet and Huawei are advancing with great momentum, while Bosch remains one of the European applicants with the largest number of patents.Machine learning is another key domain, underpinning the intelligentization of health diagnostics, infrastructure management, financial risk management, and even public governance. Innovation in this domain appears to have peaked in 2023, with growth slowing since then. Europe's share of machine learning patents has stabilized at nearly 30%, but it continues to lag behind the United States—except in 2019 and 2021. This is not accidental: the industrialization of machine learning generally relies on massive data resources, cloud computing infrastructure, and robust channels for translating scientific research into applications, and American companies have long occupied these ecological niches.

Pattern recognition is the basis for accurately identifying fraudulent transactions, cybersecurity threats, or disease signs in AI applications. Although this field currently accounts for less than 4% of EPO AI patent applications, its strategic value is extremely high. In particular, in recent years, the United States has overtaken Europe in patent applications for pattern recognition.

What truly reassures Europe is image and video recognition. This subfield has now become the second-largest segment of EPO AI patent applications, showing rapid expansion since 2021. Europe leads significantly over the United States in this area, although the growth rate of U.S. applicants is accelerating. Image recognition is a key technological piece for autonomous vehicles and has also played a role in medical imaging diagnostics, especially early cancer screening. Europe's existing advantages in medical devices and the automotive industry positively echo this AI subfield.

Quantum computing: a breakout window not yet fully defined

If AI is already a frontline of close combat, quantum computing is more like a blue ocean full of possibilities. Although the base is small, the number of EPO quantum computing patent applications in 2025 surged 37.9% year-over-year. European applicants maintain a leading position and achieved growth of 22.1%; but the United States is closely catching up with 43.3% growth, while Japan has exploded with a 170.8% increase from a lower base.

Quantum computing, AI, and broader computer technology are forming a compound effect that may ultimately reshape the definition of "sovereign digital infrastructure." For the EU, building a patent moat before quantum computing enters large-scale industrialization is a key window for competing for next-generation technology rules.

The dialectical relationship between energy and green

Technological evolution does not happen in a vacuum. The rapid development of computer technology is precisely forming a profound mutual construct with the global energy transition. AI large models and dedicated data centers consume astonishing amounts of electricity: a large AI data center's electricity usage is equivalent to that of 100,000 households; and the largest data centers currently under construction can consume electricity at the level of 2 million households. The International Energy Agency (IEA) has noted in relevant analyses that half of the growth in data center electricity demand can already be met through renewable energy; this makes energy storage, smart grids, and energy-efficiency technologies priority topics running in parallel with AI development.It is precisely for this reason that innovation in more energy-efficient hardware, chips, and quantum computing has evolved from a purely technological track into a fundamental driver of sustainable development. Europe's emission-reduction ambitions under the Green Deal framework are becoming deeply coupled with its industrial strategy in computing technology. AI's "intelligence" and the "greenness" of energy systems need not compete for resources; instead, they can complement each other through patent innovation.

The Micro-Level Vitality of Innovation Ecosystems

Beyond these macro-level data, Europe's innovation vitality is also reflected in the breakthrough creations of specific individuals. Norwegian inventor Esben Beck used image recognition, AI, and laser technology to invent a small underwater robot that removes sea lice from farmed salmon. Without the need for chemical agents, this product, through its application in Norwegian fisheries, demonstrates the value of AI in protecting natural ecosystems. The invention also made him a finalist in the SME category of the 2019 European Inventor Award.

At the other end of the Baltic Sea, a Lithuanian team developed an AI-based enzyme design platform, "Intelligent Architecture™," which can design enzymes for specific industrial applications from first principles, and thereby reached the finals of the 2025 EPO Young Inventors Award. In fields such as precision medicine, sustainable chemistry, renewable energy, and future food, this fusion of AI and biotechnology could become an important driver of Europe's bioeconomy.

These cases cannot be explained by filing volumes alone; rather, they precisely reflect the unique value of the European patent system in supporting individuals, SMEs, and research innovators. They both demonstrate the potential of "small-scale innovation" in the data-intensive era and hint at the differences between Europe's innovation ecosystem and the tech-giant-driven model of North America.

Conclusion

Looking back from the EPO's 2025 Technology Dashboard, computer technology is no longer just a matter for industry itself, but has risen to become an arena of geopolitical competition and strategic autonomy. Whether the EU and its member states can translate the creativity within Europe's knowledge system into globally competitive patent portfolios and standard-setting power will determine the success of Europe's "Digital Decade."

The deeper value of patent application data lies not in keeping a record of the past, but in providing an indication of the future. As the United States' order-of-magnitude advantage gradually becomes apparent and Asia's innovation volume continues to grow, Europe must realize that the decisive factor in the AI race may not depend on one or two leading projects accumulated yesterday, but on whether it can today fill shortcomings and amplify strengths in niche areas, and embed its technological advantages into a low-carbon, digitized real economy. In semiconductor-related innovation, although Europe still retains traditional strength, it also faces signals of momentum loss. This "first glimpse" from the Technology Dashboard may soon become an action guide for European industrial policy.