Is Europe Becoming an AI Superpower? Ursula Says So!

I have a weird relationship with Ursula von der Leyen. Every time I think about writing something about her, I decide to leave her alone. She’s a professional politician, and politicians making grand promises is approximately as surprising as finding water in a toilet.

But Ursula has this strange need to comment on almost everything. And when artificial intelligence became the latest shiny object politicians desperately wanted to sound knowledgeable about, of course she had something to say.

Quite a lot, actually.

Apparently, Europe is going to become an AI continent. Excellent. Any minute now.

Europe Isn’t Late. Apparently.

At the AI Action Summit in Paris in February 2025, von der Leyen addressed one particularly inconvenient observation: Europe was behind the United States and China in AI.

Her response?

“Too often, I hear that Europe is late to the race – while the US and China have already gotten ahead. I disagree.”

Well, that’s reassuring.

If you’re losing a race, apparently one possible strategy is simply announcing that you disagree with the position of the other runners.

She also unveiled the great European AI push, with €200 billion to be “mobilized” for AI development, including €20 billion for AI gigafactories.

Two hundred billion euros.

Holy shit.

NVIDIA executives must have started checking whether Brussels had secretly constructed another Silicon Valley while everyone was sleeping.

Except mobilize is one of those magical political words. It sounds a lot like we have €200 billion, while actually allowing public funding, private investment, existing programs, future investment, and assorted financial plumbing to be bundled into one enormous headline number.

Still potentially useful. Still a lot of money.

Just considerably less sexy once somebody removes €200,000,000,000!!! from the PowerPoint slide.

Meanwhile, Back on Planet Earth

There is one unfortunate problem with the EUROPE ISN’T LATE story. Numbers.

Mario Draghi gave a considerably less inspirational description in September 2025:

United States: 40 large foundation models.
China: 15.
European Union: 3.

Three. Not thirty. Three.

And when Draghi discussed strategically important AI built on European intellectual property, his description of progress was wonderfully concise: “minimal.”

That’s a fascinating definition of not being late.

Ursula: WE ARE NOT LATE!

Maybe America and China simply arrived suspiciously early.

Draghi also pointed out that industrial electricity prices in Europe were, on average, more than twice those in the United States and warned that the transition to a high-tech economy would stall unless the gap narrowed.

Tiny problem when your brilliant strategy depends on warehouses stuffed with expensive hardware eating electricity for breakfast, lunch, dinner, and a 3 AM snack.

Don’t Worry. Human-Like AI Is Coming Next Year.

Then Ursula gave us the sentence that convinced me this article needed to exist.

During an EU budget conference on May 20, 2025, she said:

“When the current budget was negotiated, we thought AI would only approach human reasoning around 2050. Now we expect this to happen already next year.”

Next year.

2026.

I love statements like this because they contain everything a good political technology prediction needs: an enormous claim, a wonderfully short deadline, and absolutely no useful definition of what the fuck “approach human reasoning” actually means.

Which human? Einstein? A drunk guy trying to enter his Wi-Fi password?

Someone heating aluminum foil in a microwave because TikTok promised him something shiny? Human reasoning covers quite a range.

And this isn’t predicting that next year’s iPhone will have a slightly nicer camera. Nobody can even agree on a precise measurement of “human-level” general intelligence, yet Ursula cheerfully dropped a 2026 timetable into a speech like she had received the delivery schedule from Amazon.

Human reasoning: arriving next year. Your package has been dispatched.

Beautiful. Maybe understanding political speeches will finally qualify as an AGI benchmark. If an AI can determine what the politician actually promised after the communications department explains what the politician meant, we’ve achieved superintelligence.

Don’t Worry, We Have Gigafactories

Of course, Europe isn’t going to sit around watching American and Chinese AI companies disappear over the horizon. We have a plan. A very European plan.

Factories.

Actually:

AI Gigafactories.

Much better.

The original EU AI plan talked about financing up to five of them. Then, on July 30, 2026, the EU launched a tender for up to seven AI gigafactories, supported by as much as €10 billion in EU and national public funding and intended to unlock at least €20 billion in private investment.

So credit where it’s due. The number didn’t shrink.

It grew.

Seven biiiiiig AI factories.

Now we’re talking.

And naturally, because this is the European Union, all seven will be powered exclusively by solar panels and wind turbines.

If there’s no wind at night, the AI will simply stop thinking until breakfast.

Cloudy Wednesday?

Sorry, ChatGPT Europe is currently unavailable due to weather.

Training resumes when Ursula sees her shadow.

Before anyone sends me an angry email: that’s a joke. The EU has not announced that the gigafactories will run exclusively on solar and wind.

But the joke works because Europe really is trying to solve two rather awkward problems simultaneously: massively increase energy-hungry computing capacity while maintaining its environmental and decarbonization goals.

Nothing wrong with cleaner energy. Nothing wrong with efficient data centers either.

The problem is physics doesn’t give a shit about political slogans.

AI infrastructure needs enormous amounts of electricity, reliably.

Draghi estimated European data-center electricity demand would increase around 70% by 2030, with power already accounting for as much as 40% of data-center operating costs.

An H100 doesn’t become more patriotic because the electricity price is European.

America understands that cheap, abundant power matters.

China definitely understands it.

Europe understands it too, somewhere underneath seventeen strategies, four working groups, and a sustainability workshop.

Before Building the Gigafactory, Please Complete Form AI-GIGA-27B

And then we reach another tiny European specialty.

Bureaucracy.

Building gigantic AI infrastructure is already difficult. You need land, enormous grid connections, cooling, networking, processors, financing, construction, engineers, permits, and enough electricity to make the local power company start sweating.

In Europe you can probably add:

a framework, a consultation, a roadmap, a working group, an advisory forum, an environmental assessment, a stakeholder dialogue, a code of practice, a committee to review the code of practice, and a conference celebrating publication of the roadmap for the consultation about the framework.

Nothing excites a certain species of politician and academic administrator quite like bureaucracy.

Engineers look at a problem and ask:

“How do we build this?”

Politicians look at it and ask:

“Who should sit on the steering committee?”

Academics look at it and ask:

“Could this steering committee support a three-year research project?”

And six months later everyone has a PDF.

The fucking server still hasn’t been plugged in.

This isn’t entirely a cheap joke either. The European Commission itself is now running what it describes as a major simplification drive, with targets to reduce reporting requirements by at least 25% for companies generally and 35% for SMEs, while also talking about faster permitting and reduced administrative burdens.

That’s rather revealing.

Governments don’t launch enormous campaigns to simplify things because everything was beautifully simple already.

WELCOME TO EUROPE.

We have now created a regulation simplifying the regulations created to implement the previous regulations.

Please attend the webinar.

Regulation First, Server Later

Europe does have excellent researchers, universities, engineers, and technology companies.

That’s what makes the situation particularly annoying.

The problem isn’t that Europeans are too stupid to build AI.

Europe has plenty of brains.

What Europe often lacks is the ability to turn those brains into enormous technology companies at the speed and scale achieved in the United States.

And regulation is part of that story.

Even Draghi’s follow-up assessment said the next stage of the AI Act created uncertainty and argued that Europe needed to remove barriers preventing companies from scaling. His wording was considerably more diplomatic than mine because nobody invited me to the conference.

Europe seems addicted to making sure every technological revolution has received the appropriate paperwork before we’ve figured out whether we can actually lead it.

America:

Build it. Scale it. Make billions. Deal with the problems afterward.

China:

Build twenty of them.

Europe:

Before proceeding, we need to determine whether the definition of “proceeding” conforms to Article 7 paragraph 3 subsection B.

I’m exaggerating.

Slightly.

AI FIRST!

Von der Leyen later introduced another slogan:

“AI First.”

Her explanation was wonderfully enthusiastic:

“Whenever a company or public office faces a new challenge, the first question must be: How can AI help?”

Beautiful.

Hospital waiting lists?

AI!

Train delayed?

AI!

The municipal sewer is full of shit?

HAVE WE ASKED THE LANGUAGE MODEL?

A startup cannot understand fourteen overlapping regulatory requirements?

Perfect.

Perhaps AI can explain them.

Actually, that might finally be Europe’s killer AI application:
an AI powerful enough to understand EU legislation.

Forget human-level intelligence. That thing would be a god.

The Commission now officially promotes this “AI-first” approach as part of its Apply AI Strategy.

Again, using AI more widely is not a bad idea. Of course European companies should use it.

But slogans aren’t infrastructure.

Saying AI FIRST doesn’t manufacture GPUs, lower electricity prices, build data centers, create venture capital, or turn three foundation models into forty.

At Least the Bottle Caps Can’t Escape

And while Europe prepares for its glorious transformation into an AI superpower, I would like to congratulate the European Union on one technological achievement it has unquestionably delivered.

The fucking bottle caps are attached.

Since July 2024, certain single-use plastic beverage bottles sold in the EU have been required to have caps or lids that remain attached. The environmental reasoning is straightforward: caps are less likely to become separate pieces of litter and are more likely to be collected with the bottle.

And no, Ursula didn’t personally invent the attached bottle cap.

I checked before blaming her for that too.

But as a symbol of European priorities, it’s fucking magnificent.

USA: OpenAI, Google, Meta, Anthropic, NVIDIA, Microsoft.

China: Alibaba, ByteDance, DeepSeek, huge domestic infrastructure, and an enormous industrial ecosystem.

European Union:

PLEASE DO NOT DETACH THE CAP FROM YOUR COKE.

There.

Technological sovereignty achieved.

Maybe one of our three foundation models can eventually figure out how to drink from the bottle without the cap rubbing against your nose.

The European AI Superpower Is Coming. Probably.

To be fair, Europe isn’t doing nothing.

The AI gigafactory tender is real. The Commission is putting serious money behind computing infrastructure, and the current tender really does aim for as many as seven giant facilities.

But let’s get real.

I really doubt they’ll succeed. Europe does have strong AI companies, researchers, models, infrastructure, and technology, but I wouldn’t bet on Brussels turning all that into the AI superpower Ursula is selling.

EU plans often seem to have the same life cycle. First comes the “Fuck yeaaah! We’re going to build something HUGE! This is going to change everything!” presentation. Brussels is practically vibrating with excitement. Then the erection disappears. The huge thing becomes smaller, the deadline gets longer, the billions become “mobilized,” the shiny presentation gets harder to find, and eventually everyone moves on to the next exciting strategy.

So yes, I can easily imagine Europe spending the next twenty years paying American companies for half the AI stack while Brussels keeps producing reports about European technological sovereignty. What annoys me is the fucking theater around it.

Europe doesn’t become an AI leader because Ursula von der Leyen stands behind a microphone and says: “Europe isn’t late.”

It doesn’t become an AI continent because somebody invented the phrase AI Continent.

It doesn’t catch America because €200 billion appears in giant letters on a presentation. And AI doesn’t approach human reasoning in 2026 because Ursula announced it in 2025.

Reality doesn’t respond to press releases

The European Union has become very good at being a loser in certain areas while pretending everything is going wonderfully. When Europe falls behind, Brussels has an impressive ability to rename the problem, launch an initiative, produce a strategy, announce billions, invent an inspirational slogan, and declare that we’re now leading in some carefully selected interpretation of the word leading.

Maybe these AI investments will finally change that.

But if Ursula von der Leyen wants to look more credible, which, in many cases she clearly isn’t and never will be, she doesn’t have to comment on fucking everything. And if she has an irresistible need to comment on something as complicated and fast-moving as AI, perhaps she should understand it better first, define what she’s talking about, and be realistic about where Europe actually stands.

Because saying “we aren’t late” while the scoreboard says, once again:

USA: 40
China: 15
EU: 3

doesn’t make Europe look confident. It makes the person holding the scoreboard look ridiculous.

But I know. She’s a politician. And politicians still perform one ancient part of their job remarkably well: making promises. It sucks.