Europe’s race is deployment, not models.

When Anthropic raised sixty-five billion dollars this month, the tech press asked whether Europe is cooked. The model race runs on capital and compute Europe doesn't have. But the models are already good enough, and you can rent them from a handful of vendors.

Europe has won this kind of race before. ASML has never made a chip; it makes the machines that make chips possible, out of Veldhoven, with precision nobody else can match. Europe wins where precision meets industry. The same thing is starting in AI. Mistral is moving into heavy industry — a "physics-aware" stack, built and deployed with Airbus, BMW and Stellantis for design, simulation and safety work in regulated fields. It's catching up fast, and for many services it has already cleared the only bar that matters: good enough for this one.

Ahead of HumanX Amsterdam this September, eleven founders building Europe’s AI were asked what the region does differently. None of them starts with model size. They talk about how AI works inside real companies: explainable, governed, tied to measurable outcomes from day one. One founder’s rule: build for the world’s strictest regulator first, and every other market is the easy part. The world copied Europe’s privacy rules without being asked. Done right, regulation is infrastructure.

Now the honest part. Europe’s weakness is hesitation. Roughly one in seven European enterprises has put AI to work. The Draghi report counted our dependency on foreign digital infrastructure at eighty percent and called the trajectory decline. Celonis’s chief executive said this month: adoption hasn’t gone as quickly as anyone expected.

But look at what moved this year. The AI Act was simplified and re-timed where it bound too early. Two hundred billion euros is on the table; AI factories are coming online. The supply side is being built. What’s missing is demand: organisations deploying inside their own services. That is design and operating-model work, and no amount of compute does it for you: legible enterprise knowledge, trust built as structure, role definitions for the humans in the loop. Less glamorous, but decisive.

This is the race worth running, because it’s wide open. It’s won at the service desk, on the factory line, in the radiology reading room.

And the organisations best placed to win it are rarely the giants. A mid-sized company has enough scale for the work to matter and little enough legacy to redesign it — fewer systems fused together, decisions that take a meeting instead of a quarter. It also feels the pressure first: an AI-native competitor doesn’t go after a multinational’s market share, it goes after the mid-sized firm’s. Urgency and room to move, in the same place. Europe, and the Netherlands especially, is dense with exactly these companies.

A large organisation can build the same conditions. Pick one service and give it the mid-sized treatment: a team that owns the whole workflow end to end, the people who run the systems in the room, a budget it controls, and an approval path that fits in a week. Large organisations already know how to do this; every successful scaling programme I’ve worked in started as exactly such a pocket. And the scale that slows the giants down becomes the advantage the moment something proven needs to travel.

If you lead part of a European organisation, the position is yours to take: choose the one service where trust decides everything, and make it the best-deployed AI service in your sector this year. The models are rentable. The rules are stabilising. What’s scarce is operators who can make it land.

That’s the conversation I want at HumanX in September — and before then. Cooked? Europe has barely entered its own race.

Sources: Sifted — “Is Europe cooked?”, Sifted — Mistral’s industrial AI push, Sifted — Mistral’s European AI cloud, Sifted — Celonis on adoption, HumanX Amsterdam — speaker previews, The Draghi report, EU AI Act — the May 2026 omnibus, AI Continent Action Plan