Viewpoint: Europe must stop chasing US technology hype cycles

Three weeks ago, the low-cost Chinese AI tool DeepSeek sent investors into a panic. On a shoestring budget, by AI standards, it managed to create a model that was comparable with the very best US tech giants can muster. And DeepSeek made the model’s weights open, meaning others can adapt it.  

Suddenly, investors began to worry that the vast investments in AI by the likes of Meta, Google and OpenAI could be bridges to nowhere. What was the point of throwing massive amounts of money at AI if DeepSeek can cheaply undercut any results? Not for the first time, some critics described OpenAI as a grossly overvalued WeWork of AI. 

So it was a strange spectacle this week to see EU leaders solemnly announce that they, too, would like to throw massive amounts of money at AI, as though the DeepSeek panic had never happened. 

Opening the Artificial Intelligence Action Summit in Paris, French president Emmanuel Macron, boasted of a €109 billion package in France, with much of this private cash coming from the United Arab Emirates to build data centres. 

Not to be outdone, European Commission president Ursula von der Leyen reeled off an even bigger set of numbers: the EU would drum up €200 billion of AI investment, including four new gigafactories boasting tens of thousands of AI-focused chips.

Is this a sober investment of money and focus on a technology likely to bring the most economic and strategic rewards to Europe? Or is it a kneejerk reaction to flashy announcements made by the US?

Macron gave us a clue as to which. France’s €109 billion splurge, he said, would be equivalent to the US’s gargantuan $500 billion Stargate AI data centre project, announced shortly after Donald Trump entered the White House. Never mind that DeepSeek’s emergence a few days after Stargate threw the entire project’s logic into question. 

Wider AI doubts

Even before DeepSeek came to the attention of western countries in late January, doubts were growing over the real-world usefulness and profitability of AI. Or, to be more precise, the value of this generation of generative AI, which includes large language models and image generation tools like DALL·E. 

Of course, some technologies – gene editing comes to mind – go through hype and bust cycles before ultimate success. But it doesn’t follow that all technologies that get overhyped will eventually come roaring back. 

ChatGPT is now more than two years old. If this really is the transformative technology of our age, shouldn’t we have seen more use cases by now? The mass displacement of human workers? Or even actual profits for AI companies? 

“Eighteen months after the introduction of generative AI to the world, not one truly transformative—let alone cost-effective—application has been found,” wrote Goldman Sachs’ Jim Covello in June 2024. It’s now been 26 months, and we’re still waiting, I would argue. 

OpenAI boasts 11 million paying subscribers. That sounds like a lot, until you realise it’s about as many as the New York Times, which, unlike OpenAI, makes a decent profit. Where, one might ask, are the giant EU subsidies for newspapers?

Sam Altman’s company was set to lose $5 billion in 2024, and only expects to make a profit by 2029, according to The Information. Altman has admitted that the company is - somehow - losing money on its $200 a month subscription package. 

There’s also plenty of scepticism about whether AI tools will be a shot in the arm for the wider economy. The economist Daron Acemoglu estimates it will add 1% to US GDP over the next decade: a “nontrivial” but “modest” amount, as he puts it. US overinvestment in AI could trigger a tech crash later in the decade, he speculated in the Financial Times earlier this week. 

Overstated potential

I could go on and on: there are plenty of reasons for caution. The point here is not that generative AI won’t lead to any useful tools – it has already become a useful aid for software developers. And other forms of AI could be much more transformative, and consequently potentially dangerous, in the future. 

Rather, the danger is that the EU buys into a generative AI hype machine that massively overstates its potential, blinding Europe to other approaches and technologies. 

Take the EU’s GenAI4EU initiative, which has a budget of €500 million to find real-world uses for generative AI. 

These grants will be focused largely on generative AI, rather than exploring other forms of AI less beset by hallucinations and reliability problems that continue to plague even the latest generative models.

It’s precisely these new approaches to AI, which can logically reason, which the EU should be pursuing, rather than slavishly copying US models, someAI experts have argued.  

Virtual reality

AI is not the first US-driven hype train Brussels has uncritically boarded. In 2016, I started covering the emergence of commercially available, high-quality virtual reality. Like using ChatGPT for the first time, putting on a VR headset is a genuinely wonderous moment, a sense that the future has arrived. 

But that initial “wow” factor does not mean a technology will reach mass adoption. The more I reported on VR, the more sceptical I became. Real life touch, realistic vision, let alone something like smell, was way off. Ready Player One this was not. 

Then, in 2021, Mark Zuckerberg – having rebranded Facebook as Meta – staked the future of his company on building what he called The Metaverse. Strap on a headset, and you’ll be able to do “almost anything you can imagine – get together with friends and family, work, learn, play, shop, create,” he enthused – using Meta’s platforms, naturally. 

Seemingly serious analysts plucked unfathomably huge numbers out of the air. The Metaverse could be worth up to $13 trillion, claimed Citi in 2022, but will require enormous infrastructure investment. Does that sound familiar? 

The EU got caught up in the hype. The Commission still states on its website that the metaverse. or “virtual worlds” as it calls it. has the potential for “860,000 new jobs created by 2025” and will allow you to “meet your friends, study, work or even walk among performers at the opera.” 

The EU is still spending real money on the metaverse, despite vanishingly few signs of real human beings wanting to work and play in it, and Meta losing billions of dollars on the project. There are decent niche uses for VR for sure, in training or gaming, for example, but the dream of the Metaverse seems dead. 

And yet, a leaked Horizon Europe Work Programme for 2025 proposes millions of euros for metaverse projects, including “creative breathtaking divergent worlds and landscapes” and “AI-generated smart digital assistants and 3D chatbots.” This is only a draft and could change, but is it not time for a rethink given the failure of the wider concept?

Blockchain

And remember blockchain? Interest peaked in 2017, but any hopes of mass deployment have since fizzled out. Blockchain's main contribution, I would argue, has been as a kind of intellectual scaffolding for countless cryptocurrency scams.

Yet the EU fell for the blockchain craze, hard. Between 2016 and 2024, Horizon 2020 and Horizon Europe spent €700 million on blockchain related projects. Without the hype that surrounded blockchain and cryptocurrencies in the mid-2010s, it’s impossible to imagine such largess on a solution looking for a problem. 

Born of a cringing inferiority complex vis-à-vis the US, Europe time and time again throws money and attention at whatever is touted as the next big thing across the Atlantic. A more self-assured science and technology policy would ask harder questions of each new tech trend, those who promote them, and the often pliant technology media ecosystem. 

Let’s just hope that in 2030, we won’t look back and ask: they spent how much on generative AI?!

David Matthews is International Editor of Science|Business.