Stay informed with free updates
Simply sign up to the Technology sector myFT Digest — delivered directly to your inbox.
It’s a common view that EU growth policy amounts to stifling the first spark of entrepreneurial initiative with the dead hand of risk-averse technocracy (or “the dreaded EU working group”, as US Treasury secretary Scott Bessent quipped). It’s at least as true to say that EU policymaking is shaped by something more visceral but not necessarily better: an intense sense of Fomo.
That fear of missing out is certainly the case for digital tech policy in general, and specifically for AI, the hottest purported growth lever in town. You can hardly turn around in Brussels or an EU capital for someone telling you Europe is “losing the race” or that “competitiveness” is at risk because it cannot match the US or China in developing homegrown large language models.
It is worth taking a few deep breaths. Fomo is a poor motivation for good policy. What matters in the end is not “competitiveness” but productivity growth — companies’ ability to keep improving what valuable output they can get out of a given amount of resources.
It is as yet completely unknown whether foundation models are where the economic value of AI, whatever it may be, will be reaped. Any lift to productivity growth will presumably come from deploying the technology smartly across swaths of economic activities — an opportunity not restricted to those now spending fortunes on model training. Indeed many industrialists think Europe’s manufacturing tradition is an advantage in getting AI to yield real growth.
And there are reasons to think that, as we move from training AI to deploying it, many more companies than the “hyperscalers” can compete for a share of the pie, in areas ranging from chips suited for model training to software applications making the best use of the AI capacity that hyperscalers bring online.
The real challenges the AI revolution poses for European economies are at once more mundane and more dramatic than losing some “competitiveness race”. On the mundane side is a simple matter of price and the risk of paying over the odds in the absence of European alternatives to US (or Chinese) LLMs.
Already the Eurozone has seen what was a €100bn-a-year bilateral surplus vis-à-vis the US at the start of the decade turn into a deficit of more than €50bn last year. A large part of this comes from rising payments for business services and intellectual property charges. This paints the picture of an economy in the grip of foreign suppliers who are using their market power to extract increasing economic rent. There is reason to think that greater use of US-owned AI models will reinforce this pattern.
At the more dramatic end, technological dependence risks geopolitical subjection. The threat is ultimately to European freedoms.
Europe’s lag in AI has the same causes as its lag in other digital tech: no deep capital markets to channel the bloc’s ample savings into its equally ample entrepreneurial talent, persistent market fragmentation that makes it harder to scale up than across the Atlantic and under-developed pathways and incentives to go from scientific insights to commercial products. The solutions are known, as Europe’s leading tech CEOs recently highlighted, but they require political leadership to be delivered. Looking for brand-new AI-specific policies, meanwhile, risks distracting energy from addressing the basics.
So does harping on about regulation standing in the way of innovation. Specific rules can undoubtedly be streamlined, and constantly changing regulations is bound to slow things down. But setting standards for products (digital or otherwise) and allocating liability for failure is no greater burden for a business than a customer’s product specifications or contractual terms.
There is one newish policy tool that does deserve greater space in the European toolbox: using public procurement to create markets, as Luc Frieden, prime minister of Luxembourg, rightly argued in a speech last week. The public sector makes up a sizeable chunk of demand in Europe’s economies. On top comes private activity that enjoys public subsidies. Both are tools that should be used much more consciously to reserve predictable markets for made-in-Europe digital tech. Military and sensitive public data processing and research and development are obvious places to start — for security reasons, but also to let companies invest in the expectation that they will be able to sell.
The American venture-driven model may be the Field of Dreams: build it, and they will come. The European approach could be the reverse: come (with your money) and they will build it.
[email protected]
