By the editorial team
In this era of new rivalry between the great powers, the European Union continues to come up against a fundamental flaw: it is a union primarily in name, without the economic and political cohesion required to be competitive.
With the advent of artificial intelligence, this flaw could be fatal if left unaddressed.
If recent years have proven anything about trade relations, it is that over-reliance on others can be dangerous. Nevertheless, the EU has allowed itself to become deeply dependent amid the dizzying growth of artificial intelligence, a technology that promises to transform everything from work to war.
The main European AI provider, Mistral AI, lags significantly behind global leaders Anthropic PBC and OpenAI. The EU has only 5% of the advanced computing capacity needed to train breakthrough models and produces practically none of the necessary chips. It relies on companies subject to U.S. law, such as Amazon.com Inc., Alphabet Inc.'s Google, and Microsoft Corp., for most of the cloud storage.
Such dependencies pose a significant risk: The U.S. demonstrated that it exists when it denied foreign governments and companies access to Anthropic's groundbreaking "Mythos" model — a model that can carry out sophisticated cyberattacks at unprecedented speed and, therefore, is crucial for defending against such attacks. The U.S. has also put itself in a position to decide who gets access to future, potentially riskier models first.
The EU seems to recognise the threat. However, her reaction so far has been disappointing. The proposed package of measures on "technological sovereignty" would, at worst, amount to self-destructive isolationism – as if requiring civil servants (or, say, systemically important banks) to use something like Mistral's "Le Chat", which runs on less advanced locally produced chips, would somehow enhance the competitiveness of anyone, instead of leaving them vulnerable to higher-model cyberattacks. In its most ambitious version, the plan includes public incentives to attract around €300 billion of private sector investment in computational AI over 10 years. Even if it materializes – a big "if", given the fate of previous efforts – this amount is much less than what US companies will invest this year alone.
All is not lost. Europe has a number of companies crucial to the AI supply chain – notably the Dutch ASML Holding NV, specialising in chip lithography – and, when it comes to breakthrough technologies, the frontrunners do not necessarily remain dominant. What the Union desperately needs is a much larger and more dynamic capital market, which would be much more efficient than the Brussels bureaucrats in identifying and financing niche areas where domestic companies can become indispensable. In addition to the direct economic benefit, this would provide the EU with much-needed bargaining power when its trading partners threaten to cut off the supply of chips or computing power.
The obstacle, as always, is the EU's internal policy. Its leaders have long publicly called for a capital markets union. However, for various reasons – fear of internal reactions, desire for control, protection of influential local industries – they have been reluctant to move towards the deeper regulatory, legal and tax integration required to make this a reality. As a result, European markets remain too small and fragmented to handle US-scale investments. Promising startups are heading to the other side of the Atlantic, where the money is – and where regulatory hurdles are less daunting.
The EU has so far managed to cope with its flawed way of operating, gradually losing ground to the US and China. AI emerges as the accelerator that makes this approach unsustainable. EU leaders should be honest with their citizens about the nature of the challenge and act accordingly before their Union falls far behind.
BloombergOpinion
