SpaceX, Elon Musk’s space-exploration company, has just completed the largest stock-market flotation of all time. It’s worth noting that the IPO also included SpaceX’s sizable AI division, xAI. ChatGPT owner OpenAI and its rival Anthropic, the firm behind large-language model Claude, are, meanwhile, eyeing flotations of their own, with the markets currently valuing the former at $850 billion and the latter at $965 billion. For comparison: their largest European competitor Mistral, developer of the chatbot Vibe (previously known as Le Chat), is deemed to be worth just €11 billion.

There’s been much talk of American AI hype, but there are solid real-world economic reasons for the gold-rush mood on the money markets. At its heart is ‘agentic AI’. Up to now, AI tools have primarily been used for research, writing and programming, but the next stage of their development will see them performing tasks autonomously — and not just for private individuals; companies will increasingly be using them too. These AI ‘agents’ will, before long, be running entire business processes, organising workflows, optimising supply chains and laying the groundwork for decisions.

If this trend continues, European firms will find their business processes tied to foreign-owned AI operating systems whose rules are made elsewhere.

Anthropic, in particular, is among those leading the charge, while the losers in this race include German enterprise software company SAP, which has already felt the negative effects, its share price having plummeted over the past year. The predicament of this blue-chip European company is emblematic of the difficulties European economies are set to face. After all, with its ability to boost the productivity of business processes, AI could change the rules of the game when it comes to states’ industrial infrastructures.

Digital sovereignty, therefore, is about much more than just individuals’ social media use or the prevalence of Microsoft software in public sector IT. More broadly, it’s about whether European firms will become reliant on foreign-owned productivity tools and whether this will plunge European countries into a deep economic and political dependency — and not just at the software level. Already, the US’s AI ecosystem feels like a digital spider’s web in which Europe is ensnared.

Amazon and Microsoft dominate the mega-clouds, with European cloud-storage providers being niche operators at best. It’s a similar story with the data centres that power AI processes. In high-performance chips, meanwhile, US giant Nvidia is the clear front-runner, dwarfing Munich-based chipmaker Infineon.

If this trend continues, European firms will find their business processes tied to foreign-owned AI operating systems whose rules are made elsewhere. In the medium term, the European Union could thus become a kind of US AI colony — a digital Greenland, if you will, with the land grab here being virtual rather than physical.

It’s time for digital self-empowerment

So what should it do? Option number one: try to emulate the US. This, however, is highly unlikely to succeed. Europe doesn’t have massive investment funds that can invest billions in high-risk start-ups at the drop of a hat, nor is it a largely unregulated market in which new business models can easily flourish. In addition, Europe hasn’t produced big tech titans like Google, Amazon or Microsoft; those firms have invested heavily in OpenAI, Anthropic and co and know exactly how to strategically exploit global markets.

Option number two: follow China’s lead. According to the Stanford AI Index, the capabilities of Chinese large-language models are only slightly inferior to those of US chatbots. Moreover, Chinese AI firms are highly efficient; the first version of DeepSeek, the best-known Chinese AI model, is said to have cost less than $300 000 to develop. Start-up DeepSeek aside, China’s AI industry almost exactly mirrors that of the US, being dominated by e-commerce giant Alibaba, TikTok parent company ByteDance and social media firm Tencent, which owns the WeChat platform.

These rivals may each be worth billions, but they are also subordinate to the state, i.e. to the Chinese Communist Party. And its leaders have big plans: in the coming years, they want to see AI used across all major industries — everything from waste management to sugar production. In order to achieve this objective, the state is investing heavily in an AI infrastructure comprising low-cost electricity generation, huge data centres, mega-clouds and manufacturing capabilities for powerful chips.

The EU needs to give targeted support to a ‘European champion’.

A second aspect of this strategy is to have Chinese companies buy Chinese AI products. The state is leading by example, with public sector bodies driving up demand, although some of this is via large-scale investment in ‘security programs’. Although parts of China’s AI strategy may seem worth imitating, for democratic Europe, Chinese AI can’t be considered a full-scale alternative.

That leaves only the third option: a strategic, independent industrial policy via which the EU would create a coordinated European AI infrastructure, develop its own operating systems and establish trustworthy democratic standards.

Granted, the EU hasn’t been just sitting back and doing nothing, but much of what it has done has been piecemeal and afflicted by setbacks. The European Commission’s most important initiative to date is InvestAI, a package that it hopes will mobilise €200 billion of AI investment. InvestAI, however, is a public-private partnership under which major EU funding is contingent on the involvement of private investors. To date, though, only Japan’s Softbank has made any significant commitment.

Central to InvestAI is the construction of data centres. A few smaller facilities have already been built, but the far larger giga-factories envisaged by the Commissionso far exist only on paper. Moreover, the EU has been here before: in 2023, it set out to boost chip production, primarily via subsidies, but US chipmaker Intel has since abandoned its plans for a large plant in Magdeburg, a setback that highlights the limitations of this policy and could even be seen as emblematic of its failure.

Despite these initial measures, there’s still a huge amount of ground to make up, first and foremost in funding. The 2024 Draghi report on the future of European competitiveness called for €800 billion of additional public and private investment per year to enable EU member states to compete economically and technologically with the US and China. That might seem like a fanciful sum, but it emphasises that the EU has hitherto been thinking on much too small a scale.

What Europe needs now is to come up with a genuine grand plan that would secure funding for an effective strategic industrial policy. We’ve already seen the rise of US and Chinese AI giants that have been able to exploit their size advantage at home and abroad, which is why the EU needs to give targeted support to a ‘European champion’. The only realistic candidate for that title right now is Mistral AI.

Support for a second European champion, in the field of high-performance chips, should also be considered. Here, the most promising candidate would be Infineon. Although the company has so far focused on specialised chips such as those used in the car industry, its size makes it the best-placed option. Moreover, it would provide a German counterweight to France’s Mistral.

Europe-made AI could become the gold standard for transparent models, fair data protection and clear liability rules.

Another key point to consider is that scepticism towards AI is noticeably increasing. According to a Gallup survey from April of this year, almost a third of Gen Z respondents in the US expressed anger towards AI. Criticisms include data theft, hallucinated results, negative impacts on creativity, the danger of dumbing down and the risk of political manipulation.

Scepticism is also rising on this side of the Atlantic. That should prompt us to view regulation not as a bureaucratic obstacle but as an opportunity, and thus a competitive advantage. Europe-made AI could become the gold standard for transparent models, fair data protection and clear liability rules as well as a byword for reliability and industry-appropriate security.

For that to happen, however, there needs to be a fundamental shift in the mentality of European politicians. Time and again, we’ve heard legislators in Brussels and Strasbourg argue that the main race is already lost, that Europe should instead concentrate on industry-specific AI for sectors such as the car industry or mechanical engineering. Granted, Europe has strengths in those areas. But if it doesn’t address its weaknesses too, then its firms will be perpetually tied into expensive subscriptions for large, general-purpose AI models developed by the likes of Anthropic and Google — as well as for cloud services owned by a Jeff Bezos or a Bill Gates. After all, large language models will, in future, play a defining role in companies’ business processes.

If we value European strength and digital sovereignty, we also have to rid ourselves of the misconception that free trade can just carry on as before, even though it’s long been clear that the US and China are playing by different rules. That’s why we can’t allow control over the digital infrastructure of tomorrow to be ceded to foreign actors, much less to actors who treat trade as a political lever. Acknowledging this is not nationalistic bravado; it is the first step towards economic self-empowerment.