Europe's AI Problem Is Not the AI Act, It's the Second Cheque

Este artículo aún no está disponible en español. Se muestra la versión en inglés.

You have read this comment. Possibly you have written it. A European AI company appears in the news, and underneath sits the same sentence in some variation: Europe regulated itself out of the race. It is a satisfying explanation. It has a villain, it requires no further reading, and it flatters everyone who repeats it.

Then you look at the first half of 2026 and the story stops fitting. PitchBook counts about 44 billion euros of European deal value in six months, meaning money actually committed to startups, on pace for roughly 27 percent more than last year. AI took 60.3 percent of it, up from 37.9 percent in 2025. That is not what a continent strangled by its own rulebook looks like.

Europe’s AI money problem is no longer the total, it is the distribution, and the constraints that shape it are financial and commercial rather than technological.

A handful of companies now absorb almost everything, and the rest hit a wall at the second cheque. If the AI Act were the binding constraint, you would lobby. If the second cheque is, you have to change who is allowed to invest and who is willing to buy. Harder, slower, much less fun to post about.

More money, fewer companies

Both things are true at once, and that is the whole story.

In PitchBook’s numbers, the rise is driven by rounds of 100 million euros and above, while the number of deals falls. Crunchbase counted a 44 percent drop in European seed deals, the first outside investment a young company takes, in the first quarter of 2026 compared with a year earlier. Fewer companies get funded. The ones that do get much more.

Mistral, Helsing, Lovable and a few others are doing fine. That is not the same as Europe doing fine, and the two keep getting confused, usually by people quoting the total.

The rules that actually bite are the boring ones

Here the Brussels story is right about the mechanism and wrong about the department.

Ask why European growth rounds are thin and you arrive at the same place: the institutions with the deepest pockets on the continent, insurers and pension funds, mostly cannot put money into startups at scale. Under Solvency II, the EU’s capital rulebook for insurers, shares in companies that are not listed on a stock exchange carry a capital charge of around 49 percent in the standard approach, meaning the insurer must hold enormous reserves against the investment. A gentler category for long term holdings has carried 22 percent since 2019, but the eligibility rules are strict enough that insurers reportedly hardly use it, which tells you something about how the incentives land in practice.

So European venture funds raise less, write smaller cheques, and cannot lead the late rounds. American funds do it instead, and the company’s centre of gravity moves west.

Brussels knows. The Savings and Investments Union work, the EU’s push to get more of Europe’s savings into European companies, and the recent revision of the Solvency II implementing rules are aimed at exactly this. Whether they move enough money is open. The point is that this, not the AI Act, is the regulation with a plausible path from a rule to a missing cheque.

The other half is your employer

Money is only part of it.

Anjney Midha invests from San Francisco and has sat on the boards of Mistral and Black Forest Labs. He was a general partner at Andreessen Horowitz when Sifted asked him in February 2025 what the biggest bottleneck for European AI was, and he did not name regulators. He left the firm later that year to start his own fund. He said the most urgent bottleneck is inaction by the CEOs and executives of Europe’s largest companies, who have been slow to adopt what AI already does. He called it a cultural issue and added the uncomfortable part: if you run a large European enterprise, it can be comfortable to keep doing things the way you always have, until it is too late.

The numbers complicate that, and they should. Eurostat found that 20.0 percent of EU companies with at least 10 employees used AI technologies in 2025, up from 13.5 percent in 2024. That is a jump of half in a single year, and it argues against the picture of a frozen continent.

Read both together and you get something more precise than either. Adoption is growing fast from a low base, and it is wildly uneven: in the same Eurostat survey, 42.0 percent of companies in Denmark, 5.2 percent in Romania. For a European AI company trying to reach the revenue that justifies the next round, a fragmented and mostly still hesitant home market is a real handicap, even while the trend line points up. If large European customers spend two extra years in evaluation, the company raises abroad, sells abroad, or slows down.

The strongest case against this

The best counterargument is not about money at all.

Regulation does not need to block anything to do damage. It can work as a signal. If a growth investor, meaning the kind of fund that writes cheques to companies that already have customers, believes Europe is a place where rules arrive unpredictably, that belief shows up as a lower valuation or a passed deal, and no line item ever says “AI Act”. You cannot measure the companies nobody founded. Absence leaves no data.

That is serious and I cannot fully rebut it. Two things stop me from accepting it as the main story. First, the same investors who call Europe a regulatory swamp are writing very large European cheques right now, which is odd behaviour from people who have written the continent off. Second, the AI Act’s most demanding obligations are not even in force. The high risk rules, the heavy ones covering uses like hiring and credit scoring, were pushed from August 2026 to December 2027 by the EU’s simplification package, with the product related category moved to August 2028. A rulebook whose main chapter starts in 2027 cannot explain a funding funnel that was already narrowing while it was still being written.

Fair caveat in the other direction: compliance work is real, it costs lawyers and engineering time, and it lands hardest on small companies, which is the opposite of what anyone intended.

What this means for you

If you work in a European company, you are in this picture, in a boring way rather than a heroic one. Whether your employer runs a six month evaluation for a tool that costs 20 euros a month is, added up across a continent, the actual industrial policy.

If you buy software, ask whether a European option exists, and whether it is genuinely worse or just less familiar. Sometimes it will be worse, and then you buy the better one. But “nobody got fired for buying the American one” is a real force and it is not a technical criterion. For a way to judge tools that has nothing to do with the flag on the box, see the three levels that actually change what AI can do for you.

And when the Brussels comment appears under a news story, check whether the story is about a company that could not start, or one that could not grow. It is almost always the second.

What would change my mind

Three things, all observable in public data within about two years.

If European late stage rounds move toward US levels while Solvency II and pension rules stay as they are, then capital regulation was not doing the work and I would have to look elsewhere.

If a credible survey of growth investors shows they pass on European AI deals primarily because of the AI Act, rather than fund size, market fragmentation or exit prospects, the signalling argument wins and this piece was wrong.

And if the Eurostat adoption rate keeps climbing at the 2025 pace and the gap between the leading and lagging member states narrows to well under 20 points, while the funding distribution stays as lopsided as it is now, then the customer half of my argument was wrong and this was purely a capital story.

That is the useful thing about a claim that can be wrong.

Sources

• Eurostat: 20 percent of EU enterprises use AI technologies (11 December 2025)
• European Commission: Questions and answers on the Solvency II delegated regulation (29 October 2025)
• Inside Privacy: EU AI Act update, timeline relief, targeted simplification and new prohibitions (May 2026)
• PitchBook: European VC funding narrows around a handful of AI bets (July 2026, Q2 2026 European Venture Report)
• Crunchbase News: AI drives Europe’s second straight quarter of funding gain as deal volume falls sharply (April 2026)
• Sifted: A16z’s Anjney Midha on backing Mistral, Black Forest Labs, and why Europe’s biggest problem isn’t regulation (20 February 2025)