Reasons for the increase in AI adoption among EU businesses and the ongoing gap in catching up.

      Last December, Eurostat released information that, if it had been from a different continent, would have dominated the headlines. They reported that 20% of European Union businesses with a minimum of ten employees currently utilize artificial intelligence in some capacity, an increase from 13.5% in the previous year. This represents a rise of six and a half percentage points in just a year. In Brussels, this figure was met with quiet relief. An economist at a Berlin think tank shared it with a colleague, simply stating: “finally.” Meanwhile, an SME owner in a Bucharest co-working space reviewed these same statistics and calculated that Romania's adoption rate stood at 5.2%.

      This range, from 42% adoption in Copenhagen to 5% in Bucharest, illustrates the starting point for any candid discussion on AI adoption across Europe. The continent has not been stagnant. It has seen rapid progress in some areas while others have remained still. The overall figure of twenty percent can be misleading, as it masks the reality of a market that no longer functions cohesively on this issue.

      The prevalent belief that Europe lags behind the United States in enterprise AI adoption is often attributed to regulatory factors. The AI Act is said to create apprehension among corporate boards and bog down legal departments. While there is some truth to this, it doesn't capture the full picture. The underlying issue is that Europe's AI adoption is low for reasons that have historically limited its tech sector: a lack of capital and a shortage of skilled workers.

      The single market exists only in theory, as companies purchasing AI primarily procure it from American providers. Beginning with capital, data released by the OECD in February, which Christine Lagarde referenced in her November speech to the European Parliament, indicates that around three-quarters of all AI venture capital in 2025 was directed toward U.S. firms, totaling roughly $194 billion. In contrast, the European Union as a whole attracted just $15.8 billion. This isn't just a gap; it represents two different scales altogether. Lagarde also noted that about 70% of the per-capita GDP gap between the EU and the U.S. can be attributed to productivity discrepancies, with about two-thirds of that gap stemming from the technology sector since the early 2000s.

      These figures resonate with real-life implications. A French SME considering an AI pilot often first encounters a nonexistent budget, then turns to a service that is predominantly American.

      This brings us to the second structural obstacle: three U.S. companies controlled about 70% of the European cloud infrastructure market in 2025, whereas European providers held approximately 15%. Consequently, any enterprise AI implementation in Europe that does not intentionally navigate this reality relies on U.S. computing resources, billed in dollars, and subject to a foreign court's interpretation of data protection laws. This concern is far from speculative.

      As documented, Mistral’s CEO, Arthur Mensch, has spent the last year advocating for Europe to "own and operate" its AI infrastructure, and the company has invested $830 million in a Paris data center to emphasize this point. Yet, it’s still a long way from realization.

      Within organizations, the key constraint is personnel. According to the OECD's December 2025 report on AI adoption by SMEs, prepared for the G7 presidency, half of the surveyed SMEs identified a skills shortage as their main barrier to adoption, while 40% cited maintenance costs. Other concerns included hardware (32%) and confusion over digital regulations (26%). These results are not from executives who have been intimidated away from AI by regulations from Brussels; rather, they indicate a willingness to embrace AI if they could find someone to implement, manage, and explain it in their own language. This is reflected in the Eurostat statistics: large enterprises in the EU adopt AI at about 55%, while small businesses sit at 17%. The gap is practical, not philosophical—it hinges on having an in-house data engineer or not.

      At this juncture, it may be tempting for an American audience to interpret the AI Act as evidence that Europe has prioritized process over progress. However, the truth is more complicated. The most stringent provisions of the Act, pertaining to high-risk systems, won't take effect until August 2026. The European Commission has already signaled its intent to ease restrictions, proposing in a Digital Omnibus released on November 19, 2025, to reduce the compliance burden by 25% overall and by 35% for SMEs by 2029. Furthermore, it has expanded the simplified framework to include firms with up to 750 employees and €150 million in revenue.

      The Commission seems aware of the same survey data. Whether it has acted in time remains uncertain. Industry analysis indicates that EU and UK developers have reported launch delays in nearly 60% of cases due to the Act, with close to two-thirds of European companies still unclear about their obligations under it.

      While regulation is not the primary factor