Germany and Europe's AI sector has strong research but few models, limited compute, and regulations that favor US competitors

A scientific advisory body to the German federal government paints a sobering picture of Germany as an AI hub in its 2026 annual report: strong research output, but hardly any homegrown models, too little compute capacity, and a General Data Protection Regulation (GDPR) that slows European developers while US models operate freely on the EU market. A proposed "28th regime" aims to finally open the fragmented EU single market for startups, and the experts are also calling for radical reforms within Germany's armed forces.

The six-member Expert Commission on Research and Innovation (EFI), chaired by Irene Bertschek of ZEW Mannheim, handed the report to Chancellor Merz today. It delivers a comprehensive assessment of Germany's research and innovation system, focusing on three core topics. These are innovation in mid-sized companies, competition in the higher education system, and the development of AI in Germany and Europe.

The commission's central diagnosis boils down to one sentence. Germany and Europe have strong research ecosystems but lag behind the US and China when it comes to developing AI models and turning research into economic value.

Only nine "notable" AI models have come out of Germany

The numbers make the scale of the gap hard to ignore. According to the report, the EU actually leads the US in scientific AI publications, though China dominates the field overall. Germany ranks fifth globally, behind the United Kingdom, and produces more AI publications than any other EU country.

But the picture shifts when it comes to patents. The US and China dominate transnational AI patent filings by a wide margin, with Germany trailing behind Japan and South Korea. The commission confirms a pattern it first identified in its 2022 report. Germany's relative weakness in AI and microelectronics isn't about research quality. It's about the inability to turn new ideas into real-world products and applications.

The gap becomes especially stark when looking at market-facing indicators. An analysis of data from the nonprofit Epoch AI on "notable" AI models found that only nine came from Germany between 2023 and 2025. The US published 250 during the same period. According to the report, Germany's notable models primarily come from research institutions that frequently collaborate with US partners but rarely with organizations from other EU member states.

26,000 H100 equivalents against US dominance

The technological divide becomes most visible when looking at AI infrastructure. For 2025, only about 26,000 H100 equivalents of compute capacity in large data centers are publicly documented in Germany. For comparison, the US has roughly 1.4 million H100 equivalents in 2025, according to the Epoch AI data used by the commission.

There's also a structural problem. Most of Germany's AI compute capacity is publicly operated, and EU state aid rules mean private companies can only use it for pre-competitive purposes. The commission recommends the EU set a goal of providing 10 to 15 percent of global compute capacity within the next five years and that the private sector needs to be empowered to drive this expansion.

Concrete obstacles stand in the way, the commission says. These include insufficient grid capacity, electricity costs that are high by international standards, and long wait times. Getting a grid connection can take an average of up to seven years. France and the Nordic countries offer more favorable conditions.

Investment gap shows the EU spends less than half of what the US puts into AI

The report estimates EU spending on AI at around 130 billion euros (0.75 percent of GDP) in 2023, based on OECD figures. The US comes in at 310 billion euros (1.22 percent). Germany, with an estimated 20 billion euros, falls below the EU average and far behind the US. The gap has actually widened in 2025 due to sharply rising US investment.

The structure of Germany's AI spending stands out. A full 49 percent goes to research and development, a high share by international standards. In the US, R&D accounts for only 29 percent, with the bulk going to data and equipment. Venture capital shows a similar gap. EU investors put roughly 13.1 billion dollars into AI in 2025, compared to 119.8 billion from US investors.

In terms of AI companies, Germany counts 1,980 firms according to a Crunchbase analysis, placing it seventh globally. The US has 21,719. One bright spot is that Germany's startup share sits at 31.3 percent, nearly on par with the US at 32.9 percent.

A "28th regime" to fix Europe's fragmented market

The commission identifies the fragmentation of the single market as one of the central reasons European startups struggle. Young companies looking to expand have to navigate 27 different national legal, tax, and administrative systems.

The commission strongly advocates for a so-called "28th regime." This would be a new, unified European legal form for companies that would exist as an optional 28th alternative alongside national legal frameworks.

This regime would allow startups and scale-ups to operate under a single, harmonized set of EU-wide rules covering corporate law, insolvency, labor, and tax law.

The commission warns against a half-hearted implementation. The 28th regime must be introduced through an EU regulation to take direct effect. A directive, which would need to be transposed into national law in each member state, would defeat the purpose of unification. If no EU-wide majority can be found, the commission advises the German government to push the project forward as a "coalition of the willing" or as a joint German-French initiative.

GDPR slows European AI development while US models benefit

One of the report's sharpest findings concerns the GDPR. According to the commission, many experts see limited data availability as the single biggest obstacle for AI in Europe, even more so than limited compute capacity.

The commission describes specific cases where legal concerns around the GDPR delayed German foundation models from research by several months, ultimately forcing publication under a research-only license that excluded commercial use. A strict reading of the GDPR effectively prohibits the use of personal data for training or operating AI systems.

The report names the resulting paradox. US providers successfully market their AI models in Europe, even though their training practices don't align with GDPR requirements. The commission also doesn't see the European Commission's Digital Omnibus as a solution. It cites legal experts Härting, who argue that the "tangled web of exceptions and exceptions to those exceptions in Article 9(5) GDPR" will lead to the "safest path" being to avoid specially protected personal data entirely during AI training and operation.

The commission recommends amending the GDPR to make training foundation models easier. It also proposes introducing a misuse principle that would generally permit the use of data as long as no abuse occurs.

EU's digital sovereignty is "severely limited"

The commission's analysis of digital sovereignty is equally blunt. It concludes that Europe's digital sovereignty is severely limited, with the EU one-sidedly dependent on US companies for AI chips, AI models, and cloud infrastructure.

The report cites well-known examples. In May 2025, Microsoft blocked the email access of the chief prosecutor of the International Criminal Court under pressure from the US government. In November 2025, President Trump announced his intention to block the sale of Nvidia's latest GPUs to foreign buyers. Even EU data centers operated by European companies but running US technology can't fully guarantee data security because of the US CLOUD Act.

The consequences go far beyond technical issues, the commission says. As an example of the threat to European values, the report points to Grok's Holocaust denial, the AI model from xAI. The EU's trade deficit with the US in largely digital services reached 148 billion euros in 2024.

As a countermeasure, the commission recommends the German government support a private-sector European collaboration to develop an open-source foundation model. The EU and its member states should serve as long-term anchor customers for this model, since staying competitive requires continuous iterative development.

Mid-sized companies show hidden innovation strength under crushing bureaucracy

The report devotes significant attention to Germany's mid-sized companies (Mittelstand), which make up roughly 90 percent of all businesses in the country. The analysis reveals a nuanced picture. Measured by revenue, mid-sized firms spend less on innovation than comparable non-mid-sized companies and are less likely to conduct continuous R&D. But when you control for structural differences like size and industry, mid-sized companies actually innovate more frequently than their non-mid-sized counterparts.

They're especially successful at incremental improvements and customer-specific solutions, the formula behind Germany's famous "Hidden Champions." The study also shows that when mid-sized firms invest in both innovation and digitalization simultaneously, they see especially strong productivity gains.

The commission highlights one bright spot in the tax-based research allowance. Uptake has surged. While only 2.7 percent of research-active companies applied in 2020, that number reached 17.9 percent by 2023. The tool is especially attractive for companies applying for government support for the first time.

But the barriers remain high. Beyond the skilled labor shortage, the report names bureaucracy as the main brake on progress. The commission calls for full implementation of the "once-only" principle, meaning companies should only have to submit data to the government once. It also recommends "practice checks" for new legislation to realistically assess the burden on businesses before laws take effect.

Universities face a fight for talent held back by outdated structures

Germany's higher education system also comes under scrutiny. Competition has intensified, not just for research funding but increasingly for students as well. While public universities see stagnating enrollment, private institutions continue to grow.

One positive finding is that Germany remains highly attractive to international talent. The share of international students has risen to 14 percent, and Germany has one of the highest retention rates among OECD countries. Nearly half of international students from non-EU countries still live in Germany five years after graduating, a critical factor in addressing the skilled labor shortage.

The commission is critical of internal structures, though. Hiring procedures for professorships in Germany often take more than a year and are extremely formalized. This makes it very difficult to attract top talent in international competition. The experts call for radically faster and simpler hiring processes. They also criticize the lack of transparency around teaching quality. Prospective students often choose universities based solely on research reputation because comparable data on teaching simply doesn't exist.

German research fuels foreign innovations instead of domestic growth

Closely linked to the university situation is the question of technology transfer. A study commissioned by the panel shows that German research has become significantly more relevant for global patents in recent years. The downside, however, is a massive outflow of this knowledge. Between 70 and 90 percent of patent citations of German publications come from patent applicants outside Germany.

The report breaks this down by technology field. In biotechnology, mechanical engineering, and aerospace, around 70 percent of patent citations of German publications go to foreign applicants. In artificial intelligence, the situation is even more dramatic, with that share reaching roughly 90 percent.

In concrete terms, this means the basic research happens at German universities, but the commercial exploitation through patents takes place almost entirely abroad. According to the commission, this suggests that a substantial portion of the economic returns tied to these research results don't end up in Germany either. Comparable European countries like France and the United Kingdom show even higher shares, while the US has significantly lower ones. This can only partly be explained by market size.

At the same time, IP transfer negotiations at German universities take more than 18 months on average. The commission recommends a "Transfer Time" initiative featuring transfer and founding sabbaticals along with standardized IP contracts to radically speed up the process.

German armed forces need a "founding era" for soldiers

Given the geopolitical situation, the commission also takes a close look at the innovation capacity of Germany's armed forces (Bundeswehr). The report welcomes the Bundeswehr Procurement Acceleration Act (BwBBG), passed on January 15, 2026, which strengthens instruments like innovation partnerships and enables functional performance specifications. This means innovations are no longer automatically subordinated to the lowest price.

But the commission wants more. To drive genuine breakthrough innovations in defense, it recommends expanding the Federal Agency for Disruptive Innovation (SPRIND) or establishing a dedicated military-focused innovation agency modeled after SPRIND. Waiting for a European agency would take too long given current threat levels.

The commission also proposes tapping the potential within the troops themselves. Career soldiers nearing the end of their service should receive a "founding period." During this phase, supported by the Bundeswehr through prototype testing facilities and similar resources, they could launch companies developing solutions for military needs. The commission sees the Bundeswehr's new Innovation Center in Erding as the nucleus for connecting troops, research, and startups.

R&D intensity stagnates as Germany misses its 3.5 percent target

The commission also sees a need for action on the broader metric of research spending. Germany's R&D intensity, the share of R&D expenditure relative to GDP, has stagnated at 3.13 percent for several years (2024 figure). The 3.5 percent target set by previous governments for 2025 was not met. The commission urges the current government to pursue this goal with renewed determination.

For implementing the High-Tech Agenda, which is funded with 18 billion euros through 2029, the report calls for greater transparency about the government's financial commitments, a dedicated future spending quota in the federal budget, and an Innovation Freedom Act to cut through funding bureaucracy. Ambitious research and innovation policy alone won't be enough, the report notes. "Indispensable prerequisites" include attractive general conditions for entrepreneurial activity.