When EU leaders met at a medieval castle this week to discuss how to ramp up Europe’s economic power, they were confronting a stark reality: Europe is lagging behind the innovation race with China and the US. 

In its 2026 competitiveness report, the European Commission admitted that Europe “risks losing its edge in the race for innovation”. It pointed to labour shortages, problems to scale up, low numbers of patent applications, and limited R&D spending (below the three percent of GDP target).

Europe is strong in AI research papers, but the US dominates in AI companies (with top companies like OpenAI, Google, or Nvidia) thanks to its Big Tech sector and large venture capital flows, while China dominates in rapid deployment and scale, aiming to become an AI superpower by 2030. 

A tortoise vs two hares

Overall, it is estimated that the EU has five percent of the world’s total AI computing power, compared to the current US, which has 74 percent of global high-end AI compute, and China, which has 14 percent.

To catch up, the EU has been rapidly attempting to expand its AI firepower, with the commission pushing money and energy towards developing an EU-wide AI uptake strategy and multiple AI factories across the bloc — while reopening the digital rulebooks through the digital omnibus to make data more accessible.

Other proposals include the Innovation Act and Scaleup Europe Fund to help start-ups and fast-growing businesses.

The EU set a target of mobilising €20bn in AI investment between 2018 and 2020, and €20bn annually in the decade that followed.

The commission pledged to raise its own contribution to €1.5bn for 2018-2020 and €1bn per year for 2021-2027.

But watchdogs have warned that progress has been slow and uneven. 

France and Germany have unveiled the largest national investment plans, while Italy, the Nordics, Spain, and Greece are also ramping up more modest investment focus on supercomputing and SMEs.

Against this backdrop, why has the bloc been struggling so far?

Late start, slow scale-up, fragmented approach

Europe finds itself in this position because it started scaling its AI capabilities much later than other regions, such as the US, which surged in the post-2010s, and China’s state-backed plans from 2017.

And then, when its ramp-up began, the 27-nation bloc did not have sufficient investment in the technology, which has brought scaling and distribution struggles to the continent. 

But the late start doesn’t mean Europe has been incapable of developing its own AI.

“There is a strong research background and engineering talent pool in Europe. We do have some companies even developing these foundational [AI] ]models,” explained Leevi Saari, doctoral researcher at the University of Amsterdam studying the political economy and artificial intelligence.

Mistral AI, a French AI company with a valuation of €11.7bn, is often pointed to as a European champion. 

The issues come from scaling European developments.

“What Europe doesn’t have, which is the key problem in all of this, is that we are lacking the industrial platform or sort of coordinated scalable platform that exists in the US or China,” Saari added.

In the US, for instance, AI development has been advanced by Big Tech, which has access to vast amounts of training data that companies like Google have been hoarding for years, and can distribute their tech across enormous platforms and to their partners. 

Europe currently doesn’t have the hyperscalers, data repositories, and focused industrial flexibility to move and grow as quickly as US or Chinese companies.

With part of this issue also being the piecemeal approach the bloc has taken towards the technology. 

The commission has been attempting AI-related initiatives since 2018, but its approach to coordinating with member states on development and investment has been criticised as fragmented and ineffective.

The European Court of Auditors released a report in 2024 that says: “More than five years after the first plan, the framework for coordinating and regulating EU investment in AI is still a work in progress.”

The report points to the lack of governance tools and to the absence of coordinated, specific funding and development goals. 

Bengüsu Özcan, AI governance researcher at Brussels-based think-tank Centre for Future Generations, explained to EUobserver: “capital, computing power, and talent are still fragmented across member states and regions.”

“Individual member states cannot reach the threshold of resources needed to build the large-scale compute clusters, attract world-class talent, and sustain the long-term investments that cutting-edge AI systems demand,” she added.

Europe will have to increase investment levels significantly in order to compete with the US and China on artificial intelligence (Photo: DARPA)

Market dynamics 

Another part of the AI problem for the bloc is the lack of funding to genuinely compete with the big global players.

“We [the EU] want to get away by not investing enough and still being competitive. But with our current investment level, we should not really be looking at the United States and China, because they operate at another level, if you ask me,” said Robert Praas, data and AI scientist at Brussels-based thinktank CEPS, told EUobserver.

The European Commission launched InvestAI in 2025, a public-private investment initiative aiming to have €200bn towards AI and AI infrastructure.  

Compare this to OpenAI, which alone wants to invest $500bn [€422bn] over the next  four years into AI infrastructure.

“I don’t think it [EU commission] has the resources to fundamentally change the landscape because it has a limited budget compared to American venture capital basically,”  said Praas. 

What Europe is lacking is strong, private, venture capital (VC) investment. 

Much of Silicon Valley AI development happens through VC, where private money invests, often in return for company equity. The US invested €1.33 trillion in VC between 2020 and 2025, 34 percent of which went to AI. 

The EU invested only €252bn in VC over the same period.

To help close the private capital gap, the commission is looking into adapting the bloc’s venture capital capabilities to push more private investment into the technology. 

However, the lifecycle of a venture-funded company is usually to build until you can sell to Big Tech or go public, with the intention of making massive returns for investors.

This could raise problems for European venture capital-funded companies, as the incentive to move to the US market is currently stronger than staying in Europe. 

“If you don’t think about the whole life cycle of those kinds of investments, it’s easy to make, again, counter-productive mistakes where we then create companies that will eventually be bought, or sold to the United States,” added Saari.

The European Court of Auditors warned in 2024 that: “More than five years after the first plan, the framework for coordinating and regulating EU investment in AI is still a work in progress.” (Photo: Unsplash)

But what’s the end goal?

The intense leveraging by the EU to become competitive in AI comes as the bloc is rattled by worrying competitiveness reports and by increasing hostility from longtime technology partners such as the US.

But as the bloc scrambles to develop its own technological sovereignty, the researchers noted that the continent’s AI goals seem vast.

“We want Europe to be one of the leading AI continents, and this means embracing a way of life where AI is everywhere,” said commission president Ursula von der Leyen at the Paris AI summit last February.

The InvestAI initiative, for example, aims to help create infrastructure for next-generation models that aim to boost everything from medical advancements, scientific research, and economic competitiveness. 

And the experts are sceptical that such a strategically huge approach, and trying to directly compete with giants like OpenAI on general models, is going to create real advancements in the way the Brussels executive hopes. Instead, they suggest more strategic clarity.

Saari notes that there is a need to incentivise the choice of European alternatives over foreign models. 

Currently, OpenAI is actively funding initiatives to train 20,000 European small businesses and to sign member state governments up to its platform — demonstrating the intense foreign effort to steal European customers.

“If Europe wants AI sovereignty, whatever that means, more important is to convince the good people in German industry [for example] to choose Mistral over OpenAI,” said Saari.

And Praas believes tailoring investment and development to particular sectors, rather than everything at once, is a stronger approach. 

“If we say, hey, healthcare is so important and we think this technology can help cure more people, help more people, let’s pick this one.

“Let’s create a programme where people really incentivise going into this one,” said Praas.

Suggesting, heavily focusing on a couple of sectors at a time and incubating development could be a way that the EU can compete more effectively.  

“If that’s the case, yeah, then the future can be very bright,” he added.