At this year’s World Economic Forum, leaders acknowledged that the impact of AI remains highly uncertain. Yet amid the unanswered questions, Europe may actually be entering this phase from a stronger position than many expect, writes Marieke Blom,

Marieke Blom is Chief Economist and Global Head of Research at ING

Geopolitics, competition, and the global AI race We’ve learnt quite a bit since last year: China does get access to advanced Nvidia chips, much of the recent US economic growth stems – directly or indirectly – from AI, and corporates are still wondering how to make a return on their investments. An uncomfortable number of AI discussions at this year’s World Economic Forum ended with the same conclusion: “We don’t know yet.” Yet surprisingly, Europe may be better positioned in this phase than many assume.

Last year, AI chiefs were unequivocal: prevent China from accessing the most advanced chips. That message has clearly not become reality. As Howard Lutnick described in a recent Bloomberg interview, there was an intense debate inside the Trump team, with ‘strong arguments on both sides’. The room remained unconvinced by each other’s reasoning, yet the president ultimately decided the exports would go ahead. Whether this was driven by economic calculations, strategic optimism, or simply the belief that China would prefer Nvidia over developing domestic alternatives – we still don’t know.

But we do know the outcome: for now, access to advanced chips will not stop Chinese AI development.

AI leaders described today’s competitive landscape as “ferocious” and “difficult to keep up with.” Both Dario Amodei and Demis Hassabis emphasised this, with Amodei noting (without hiding his regret) that if the US had blocked Nvidia chips for China, “it would have only been between the two of us.” In other words: without China, the race would have narrowed to a near-duopoly. Meanwhile, the technical trend is clear: models are becoming smaller and more targeted, designed to reduce compute and energy needs. Prices are falling – or more precisely, for the same price, users are getting higher quality. And despite relying heavily on open source approaches, Hassabis estimated the Chinese players are only around six months behind.

Corporate reality, economic effects, and Europe’s strategic position

Across the consultant-hosted sessions, the dominant theme was corporate return on investment. Many companies feel they must adopt AI to stay competitive, yet they cannot confidently point to where the value will come from. This was obvious from the conference stages and even more so in private conversations. Consultants stand ready to help them navigate the landscape – often by shopping more intelligently among a growing list of models – but the fundamental question remains open: Where will AI create the big shifts? Right now… we don’t know yet.

Whether AI is already visible in labour market data remains debated, mostly because it’s still early. But coding is the exception: fewer people are needed, and tasks that used to require full teams can now be done by much smaller groups. Anecdotes were plentiful. I ran into some lawyers who told me that the tedious administrative work – copying, pasting, searching through legal texts – is now largely done by AI. Their clients have noticed, and procurement departments are increasingly demanding lower prices. As an economist, this uncovers something important: we may not see the AI effect in productivity statistics, but we may see it in lower inflation.

One striking shift from last year: AI leaders now focus their messaging on their relative strengths, revenue growth, and pitching their products to corporates. The US-China rivalry came up, but only briefly. Why this shift? It could be that geopolitics is no longer a winning narrative now that China has access to chips. It could also be that the industry has entered a new phase – commercialisation and scaling – where selling matters more than strategy debates. Likely both reasons play a role, and both make it rational to focus on the market.

A key uncertainty is whether the major players will develop models with strong pricing power. Will quality differences become significant? Will interoperability remain low? Will switching costs rise? And will scale advantages – such as access to compute or data – raise entry barriers? Or will we enter a world of many, smaller models which continue that ferocious competition?

We don’t know yet. This implies we don’t know exactly to what extent the model builders will be the main beneficiaries of this new technology.

A striking macro contrast emerged at Davos. In the US, AI is now a major source of economic momentum: data centre construction, equity valuations, and the wealth effects tied to them all fuel demand.

Europe, by contrast, is primarily a buyer of AI, not a producer. That means the economic risks are lower, because Europe is less exposed to the massive capital expenditure cycle. Secondly, if scale turns out to be less of an issue, some European model builders may catch up. But most importantly, it also implies the possibility of benefiting from productivity gains, without needing to pay monopoly-level rents to the providers. Being a user rather than a producer may sound passive, but in this phase – where so many critical questions remain unanswered – it may actually be a relatively comfortable place to be.