9th May 2026

The World is Wrong about European AI.

I’m British, and despite what some people seem to believe, Britain is part of Europe. I’ve been obsessed with AI research and development across the globe since my early teens, and now that I’m almost 30, I’m lucky enough to actually work in development related to my area of interest, alongside some genuinely brilliant researchers and engineers. (I know there are a lot of people with strong grievances about AI around here on the old web - I don't personally work directly in AI, but I work alongside people who do.)

But this just makes it all the more frustrating when I see posts like this tweet. The user was listing all the major AI companies, and they pointed out that not one of the companies listed was European. Their question: “what does Europe have?” I see this pervasive belief that Europe is a technology desert, that contributes nothing towards the future other than regulations, from anyone who has a vague interest in AI, right across the map from East to West.

To be honest, it is exhausting to see this come up again and again. Seeing specific companies and countries take credit for long-term global research is frustrating to me (especially when everyone seems to have their blinkers pointed at LLMs as though that is the be-all and end-all of AI, though that’s maybe a complaint for another time). The tweet treated those who control the capital as though they created the field, when many only moved in once there was profit to be seen. Europe’s problem is not that it has done nothing in AI, but that its research, talent, and companies have too often been absorbed into US-based technology giants with deeper pockets, larger infrastructure, and a stronger influence on the public.

Once a logo is pasted on, the history behind the work starts to disappear. People just look at who owns the most visible companies. Europe’s reputation suffers because contribution gets overshadowed by capital.

First case study: DeepMind

DeepMind was founded in London in 2010, and became well known for groundbreaking work in deep reinforcement learning. One of its early breakthroughs was DQN, a system that learned to play dozens of Atari games from raw pixels. DeepMind rapidly became one of the most important AI research labs in the world, but they weren’t exactly a household name. In 2014 Deepmind was acquired by Google. Shortly after this they shared the victory of AlphaGo over European Go world champion Fan Hui, which was around when I began following their work. In 2023, in response to OpenAI’s success with ChatGPT, DeepMind was merged with Google Brain to become Google Deepmind, marking the end of their almost 10 year struggle to remain a distinct organisation.

Now when people see major breakthroughs from DeepMind, like AlphaFold or Gemini, they just see the Google logo and assume it’s more Silicon Valley tech. But this is a British company which is still shaping the modern field of AI. Despite the merge, DeepMind is still headquartered in London, with research labs across both Europe and the US. But to anyone unaware of the history, all they can see is Google’s identity as a US technology giant.

DeepMind is the clearest example of a European AI success story being absorbed into an American giant. But even now, important AI research is still being carried out in Europe under American corporate ownership.

Second case study: TurboQuant

In March 2026, Google Research published “TurboQuant: Redefining AI efficiency with extreme compression”. This research focused on reducing memory bottlenecks in AI systems through advanced quantisation methods.

Research like this pushes back against the idea that every AI problem should be solved through brute force scaling. If AI systems can do more with less memory, that eases at least some of the pressure for endless expansion. Maybe we can even save some water for humans.

When I looked into the researchers who published the work, I found that one of them, Amir Zandieh, is based in Switzerland.

So once again, this is significant AI research being carried out within Europe, but under the branding of an American company. Because the logo on top is Google, the work is much more easily read as a story of American strength than as evidence of European contribution.

For me, DeepMind and TurboQuant feel like two sides of the same coin. On one side, a British AI lab was absorbed into an American giant. On the other side, important research is still being carried out in Europe, but under American corporate branding. For both cases, the result is similar: ownership and visibility overshadow where the work was actually done.

Third case study: Mistral

Mistral is one of the clearest examples of Europe trying to build and retain a serious AI company of its own, without being absorbed into a US company (yet). Founded in France in 2023 by Arthur Mensch, Guillaume Lample, and Timothée Lacroix, Mistral quickly became one of Europe’s most visible AI startups. Its founders demonstrate how international and interconnected this field really is: Mensch previously worked at Google DeepMind, while Lample and Lacroix previously worked at Meta AI.

In 2025, Mistral launched Magistral, described by Reuters as Europe’s first AI reasoning model. A few months later, Dutch semiconductor equipment giant ASML led Mistral’s €1.7 billion funding round, making Mistral the most valuable AI company in Europe and giving it economic backing from within Europe rather than from another American tech giant.

Something that’s notable is how Mistral and TurboQuant both push against the brute-force trends of the current boom. TurboQuant does this technically, by reducing memory pressure. Mistral does it more strategically, by publishing work on the environmental footprint of its models, arguing for more frugal AI, and building European infrastructure with an emphasis on low-carbon energy. None of this makes AI environmentally harmless, but it does suggest a different instinct in parts of European AI: progress does not always have to mean scale for scale’s sake.

Overall, it’s hard to say how much of Mistral’s success can be attributed to a widening European anxiety about sovereignty. Yet within Europe, it is still able to hold its own against American and Chinese firms with far more capital, infrastructure, and public attention. Mistral proves that it's possible for AI companies to survive in Europe. But for how long can it continue?

Fourth case study: Aleph Alpha

Aleph Alpha snaps this problem right back into view. Founded in Germany in 2019, it was once seen as one of Europe’s most promising attempts to build a serious AI company outside the American and Chinese giants. Like Mistral, it was tied closely to ideas of European digital sovereignty: AI that could serve businesses and public institutions without simply handing control over to Silicon Valley.

But in April 2026, Canadian AI firm Cohere made an agreement to purchase Aleph Alpha. Reuters described the deal as part of Cohere’s move to expand in Europe, particularly in regulated sectors such as government and business. By the time of purchase, Aleph Alpha had already moved away from the oversaturated language model market, focusing instead on enterprise and public-sector AI applications where trust, compliance, and specific use cases matter more than simply having the biggest general-purpose model.

This demonstrates the brutality of the AI “race”. Aleph Alpha was a company with a strong position within the European sovereignty conversation. But competing with the biggest American firms all alone is a lot to ask.

Of course, there is still a sovereignty claim to be made even with the acquisition - the combined company could become a transatlantic alternative for organisations that do not want to depend entirely on US Big Tech or Chinese providers. But it still underlines the problem Europe has with keeping AI companies independent, well-funded, and strategically important over the long term.

This is why sovereign investment - like that of ASML in Mistral - is so important. Just cultivating research is not enough. If Europe wants to retain the value of what it creates, it needs capital, compute, customers, infrastructure, and public institutions willing to back home grown companies.

How can we fix it?

If Britain and the rest of Europe want to stop being treated as a talent pipeline for larger foreign firms, they need to invest in the companies, infrastructure, and compute capacity that allow AI work to stay here and scale here. Research alone is not enough. Talent alone is not enough. If the capital, servers, customers, and strategic backing all sit elsewhere, then the value will keep flowing elsewhere too. This bleeds the economy like a leech.

This is not just about pride. It is about the economy. It is also about sovereignty. As fears grow around dependence on American tech companies, especially during periods of US political instability, European governments and companies have started paying more attention to homegrown alternatives. Reuters has reported growing European demand for locally based tech services, with companies and governments increasingly using the language of “digital sovereignty” as worries about privacy, politics, and US tech dependence grow.

Some steps are already being taken. The UK’s Sovereign AI fund describes itself as a £500 million sovereign venture fund dedicated to scaling British AI startups, offering capital, compute access, visas, and strategic support. Its own wording is very clearly aimed at the problem I have been describing: backing British AI founders to “start here, scale here, and win everywhere.”

Already, they have started to back companies. Last month, April 2026, the UK government announced support for Ineffable Intelligence, a British AI company founded by David Silver, the former DeepMind reinforcement-learning researcher whose work helped power AlphaGo. The government explicitly framed the investment as part of making sure the UK is not just an “AI taker” but an “AI maker.”

The EU is also trying to move in this direction. Through InvestAI, the European Commission committed €20 billion to support up to five huge computing facilities intended to train next-generation models and strengthen Europe’s technological sovereignty. The European Investment Bank has also taken part, to support financing.

Mistral’s European infrastructure push is also a big part of this same movement. Its Swedish data centre investment was described as a step towards building independent AI capabilities in Europe, and the company is notably trying to keep its technology and cloud servers within Europe, unlike US-based competitors such as OpenAI.

None of this means Europe has solved the problem. Funding announcements are not the same thing as durable companies, and Europe still has to deal with fragmented markets, slower capital, weaker scale, and dependence on foreign chips and cloud providers. But these moves recognise the real issue. If Europe wants credit for its contribution to AI, it has to do more than produce talent. It also has to build the conditions to retain the talent.