Europe missed out on building the AI labs, and is quietly cashing in anyway. The continent’s established tech and industrial giants have emerged as unexpected winners of the AI boom, according to Reuters, even as its startups struggle to rival OpenAI or Google.

The pattern is a lesson in where the money actually lands. Frontier models grab the headlines, but the profits also flow to the companies that make the tools, run the software, and wire the power, and Europe is unusually strong in all three.

The clearest case is ASML, the Dutch firm whose lithography machines are the only way to make the most advanced AI chips, and which has been closing on a trillion-dollar valuation as demand surges.

Enterprise software is the other engine. SAP, now among Europe’s most valuable companies, sells the systems that big firms are wiring AI into, and every AI feature it adds is one more reason for customers to stay.

Then there is the unglamorous business of power and metal. Siemens has raised its outlook on AI-driven data-centre demand, and electrical firms such as Schneider and Prysmian are riding the buildout that every data centre requires.

The mechanism is straightforward; every dollar spent training a model eventually reaches a chip, and every advanced chip passes through an ASML machine, so the spending that starts in California ends up, in part, in the Netherlands.

Telecoms and utilities benefit too as data centres need connectivity and vast amounts of power, and the European firms that provide both are booking the demand whether or not a single European model competes at the frontier.

That this counts as a surprise says a lot about the narrative. Europe has spent two years being told it lost the AI race, a story that measured the continent only by whether it had produced its own ChatGPT.

By that measure it did fall behind. Europe has fretted openly about its AI sovereignty, dependent on American models and clouds for the frontier work its own firms cannot yet match.

There are exceptions like Mistral, the French champion, who has finally started to make its sovereignty bet pay, but a single lab does not close a gap this wide.

The incumbents’ advantage is that they do not need to win the model race to profit from it. In a gold rush, selling picks and shovels has always been the steadier business, and Europe happens to own much of the hardware store.

Collaborative efforts are trying to fill the rest. A pan-European alliance has built an open LLM as an alternative to American and Chinese models, though it remains a modest counterweight to the giants.

The win is real but narrow. It rests on selling into a boom the incumbents do not control, and if the AI build-out slows, demand for their tools and power slows with it.

It has also reshuffled Europe’s corporate hierarchy. The most valuable companies on the continent are increasingly its AI-adjacent industrials and toolmakers, rather than its banks or luxury houses.

The gains are unevenly spread, too. They cluster in chip tools, industrial software, and electrification, while Europe’s consumer-internet and social platforms remain as absent from the AI story as ever.

For policymakers, the result is awkward. Europe is profiting from AI while remaining dependent on others for the intelligence itself, a comfortable position that is also a fragile one.

The danger is reading the profits as a strategy. Selling into someone else’s boom is lucrative until the buyer builds its own supply, and Europe’s dependence on foreign models is a vulnerability a strong quarter does not fix.

Still, being the toolshed to a global boom is not the worst place to stand. Europe may not have built the future of AI, but it is selling a great deal of what that future runs on.

Get the TNW newsletter

Get the most important tech news in your inbox each week.