Jonathan Dupont and Rafi Pollack-Joyce / Oct 2026
Image: Shutterstock
In 2024, former ECB President Mario Draghi published his landmark report for the European Commission, The future of European competitiveness. Draghi argued that the productivity gap with the US is largely explained by the tech sector, pointing out that no EU company worth over €100 billion has been built from scratch in the last fifty years, while all six US companies valued above €1 trillion were created in that time.
Will AI be a repeat of history, with the economic centre of gravity shifting to industries in which the EU is just fundamentally less competitive? Is AI more likely to be a tailwind than a headwind for the EU economy?
As a simple metric, over the last thirty years you can see how the average value created by the ICT industry in the US vs EU has diverged. For the foreseeable future, it seems unlikely that any country outside the US or China is going to succeed in gaining significant market share with their own frontier AI lab.
But having a frontier lab isn’t everything. While the way national accounts work make it impossible to do an exact comparison, the 13 US tech giants are only responsible for around a fifth of the labour productivity gap between the US and the EU. Being the creator of a new technology matters - but much more important is how it gets deployed and adopted through the economy.There are different estimates for the overall potential of AI for Europe, but almost everyone agrees that it could be substantial. Our own most recent modelling suggests that it could add $2.8 trillion to EU-27 GDP by 2035, equivalent to one fifth of our estimate for the world as a whole.
But how fast adoption picks up makes a big difference to the growth path. So far, we estimate, Europe has captured only around 14% of AI’s potential, worth about $0.2 trillion a year. Fast adoption would deliver $1.3 trillion a year by 2030, against $0.7 trillion under slow adoption.
Historically, the EU has often lagged the US in business technology adoption: a pattern seen in cloud computing, e-commerce and just the wider deployment of ICT. But interestingly, it's much less clear that this is true with AI.
One way to see this is in overall sentiment. In our research, we’ve seen a repeated two way relationship between overall enthusiasm and both the extent and sophistication of adoption. Workers have to be willing to use these tools, and businesses and governments need public consent to deploy them at scale. PF Global polling in Poland, Germany and France finds that Europeans are generally more enthusiastic about AI than Americans, although France is an exception to this.
Looking at the data released by the labs on adoption, the EU does not look like a noticeable laggard. Generative AI use among working-age adults in the EU-27 already exceeds that in the US, and the median member state ranks 27th globally for ChatGPT messages per person, well ahead of the US in 51st place.
Another way to look at this is the data released by Anthropic, which lets you compare adoption across individual US states to EU countries. Here you can see that Europe is more similar to middle America than it is to Silicon Valley. It is DC, California and New York that are really pulling away.
This doesn’t mean that Europeans are completely content with just being passive consumers of American, or Chinese, AI. Across 24 countries, 47% would back building a home-grown European AI even if it performed less well, compared with 31% who would not. Polling in 3 EU countries shows that 22% are not comfortable with using US AI models, compared to 17% who are. That is better, but not transformatively different than the 33% not comfortable with using Chinese models, against 9% who were.
Having local, sovereign AI is challenging if you don’t have the local infrastructure to support it. At the moment the European public, on average, do not seem to share the growing backlash against data centres we have seen in the US. 41% of Europeans would welcome a data centre locally by 44%, compared to 13% against, a 31 points lead. In the US, the lead is just 2 points.
But this is not true everywhere, and you can already see pockets where public support is much more clear, such as in Finland, Ireland and the Netherlands. It seems likely that a backlash could potentially grow elsewhere if data centre consumption massively increases.
In short, while Europe remains behind the US in frontier labs and data centre capacity that doesn’t necessarily mean that AI can’t still be transformative for its economy.
But, as ever, taking real advantage won’t happen without maintaining public support and ensuring adoption continues to grow.
