Authored by Sasha Rubel, Head of AI Policy for Generative AI at AWS EMEA

More than half of European businesses (54%) now use AI, up from 33% two years ago, according to the AWS commissioned report Unlocking Europe’s AI Potential 2026, based on a survey of over 34,000 businesses and citizens. Investment is moving the same way, up 26% year on year overall and up 35% among startups. Despite this momentum, just 22% of adopters have reached advanced use, where AI is part of how the business operates rather than something sitting alongside it. Everyone else is still largely experimenting, a chatbot here, a pilot there, without much change to how work gets done.

Europe’s position in the AI economy depends less on how many companies say they’re using AI, and more on how many are prepared to transform their businesses using it. Some challenges are familiar, but one is becoming harder to ignore.

One rulebook, 27 interpretations

Europe moved early on AI governance, but the EU AI Act doesn’t exist in isolation. It sits alongside national data rules, sector-specific regulation and different enforcement practices across EU countries, so a single market can feel like 27 separate compliance exercises.

The cost isn’t abstract; 42% of IT budgets now goes towards compliance rather than building. Businesses citing regulatory uncertainty as their biggest obstacle expect to invest 48% less in AI over three years, and many high-growth startups have already delayed their plans. The IMF estimates this to be equivalent to a 110% internal tariff on innovation.

Businesses need clear rules and consumers need confidence AI is developed responsibly. The opportunity is making those rules work consistently across borders.

The missing piece may be leadership

Talent is another constraint, though not quite in the way we usually describe it. Nearly half of European businesses cite talent shortages as an obstacle for AI adoption, up from 39% two years ago, and it takes around 5.5 months to fill an AI-related role.

We need more people who can build and work with AI. But there’s another shortage that gets less attention: people who know what to do with it. The hard part isn’t getting access to a model, it’s working out which parts of the business should change because that model exists. That takes leadership: someone who can look at a 20-year-old process, ask whether it still makes sense, and own the answer when it isn’t obvious. Without that, AI remains a pilot that gets reported on, but not something the business genuinely relies on.

That challenge will only grow as agentic AI, systems that plan and carry out tasks with less human oversight, becomes more capable. The companies that benefit won’t be the ones with the best models, but the ones that know where to put them to work and the countries and regions that diffuse adoption across their economies.

The capital problem comes later

Europe has also made progress on funding AI. Early-stage investment has improved, but the bigger gap sits further down the road: scale-up capital, which takes a promising company from working product to global business. Those larger, later-stage rounds matter most for deeptech, where proving the technology takes years before revenues catch up.

This is where Europe can lose companies it helped create. A founder may raise an early round here and build a team, but at the next stage of growth, deeper capital elsewhere becomes hard to ignore. Around four in ten European AI startups say they’d consider relocating to scale faster, rising above half among the fastest-growing. Each departure costs more than the company itself: the jobs, tax revenues, suppliers, and people who might have started the next generation of companies.

Europe already has some of the answers

None of this means the European AI story is failing, quite the opposite. The latest AWS Pioneers Project cohort makes the case well. Proximie connects surgeons across countries with live, ultra-low-latency video, reaching operating rooms that otherwise lack specialist expertise. Iktos combines AI and robotics to speed up drug discovery for rare diseases. Paebbl turns captured CO₂ into building materials, and Quandela develops photonic quantum processors that run at room temperature over ordinary fibre-optic networks.

These companies aren’t successful despite being European. Safety, sustainability, privacy and human dignity are part of why their technologies earn trust in the first place. But trust is only one part of the equation, Europe also needs to make it easier for them to grow.

So what would help?

First, make the AI Act feel like one set of rules, not 27 compliance journeys, so businesses can build once and operate across the single market. Second, treat AI skills as basic economic infrastructure: giving managers, public servants, citizens, and workers enough understanding of AI to know what it can do and how their organisations need to change. Third, get more capital behind companies once they’ve proved the idea and are ready to scale, directing more of Europe’s pension and sovereign wealth toward its next generation of global technology businesses.

Europe has a choice

Europe doesn’t lack the ingredients: world-class researchers, an educated workforce, a technology sector worth close to $4 trillion, and a regulatory approach that puts trust at its centre. What it needs now is to connect those pieces: a single market that actually feels single, AI capability spread through businesses rather than confined to specialist teams, and capital that lets successful companies scale without looking elsewhere.

European founders aren’t asking for the rules to disappear, or for special treatment. They’re asking for the chance to stay, scale and build global companies from Europe.

Photo courtesy of AWS