The Great AI Infrastructure Heist: How Four Boring Companies Just Became Tech's New Billionaires
While everyone watched Nvidia, these century-old European firms quietly cornered the market that actually matters
The Misdirection Play: Everyone's obsessing over which AI model will achieve AGI first. Meanwhile, four companies that make electrical sockets and cooling systems just pulled off the heist of the century. They're not building AI—they're building the world that AI lives in. And that world is worth a lot more than anyone realized.
The Accidental Tech Giants
It's 2006. Schneider Electric, a French company founded in 1836, decides to buy American Power Conversion for $6.1 billion. Wall Street thinks they're insane. "Who pays that much for backup generators?"
Fast-forward to today. Schneider is valued at €127.9bn, overtaking French oil major TotalEnergies last year in a sign of the world's shift from fossil fuels to electricity. An electrical equipment company is now worth more than one of Europe's biggest oil companies.
That's not luck. That's what happens when you accidentally become essential to the future.
The same story played out across Europe. Legrand, still headquartered in the same French town where they started making porcelain light switches in 1900. ABB, the Swedish-Swiss engineering giant. Siemens, the German industrial powerhouse. Since the launch of ChatGPT in November 2022, the market valuations of Schneider Electric, Siemens AG, ABB and Legrand have grown a combined €151bn, rising more than 60 per cent in the case of German-listed Siemens AG.
They're not AI companies. They're infrastructure companies that became AI companies by accident.
The Power Problem Nobody Talks About
European data center energy demands are set to triple by 2030, with power needed to fuel European data centers growing from 62 TWh today to more than 150 TWh by the end of the decade. That's not just a big number—it's a crisis in the making.
European data centres will expand capacity by 22% this year and yet struggle to meet demand. In Dublin and Frankfurt, if you want to build a new data center, the time required to supply power to new data centres can exceed three to five years, with lead times for electrical equipment alone often surpassing three years.
This isn't about having enough servers. This is about having enough electricity to run them.
The Cooling Crisis
Every ChatGPT conversation generates heat. Lots of it. Around 35-40% of a hyperscaler's energy consumption is from cooling, and that percentage is growing as AI chips get more powerful.
For context, hyperscalers are the tech giants like Amazon Web Services, Microsoft Azure, and Google Cloud that operate vast global networks of data centers. They manage massive amounts of computing power and storage, providing infrastructure as a service to businesses worldwide. When a hyperscaler—a company that might operate hundreds of data centers globally—says cooling eats up 40% of their power budget, that's not just an engineering problem. That's a business crisis.
Nvidia's latest Blackwell chips are so power-hungry they need liquid cooling instead of air. Schneider acquired a 75 per cent controlling interest in Motivair, a specialist in liquid cooling, for $850mn specifically to serve customers like Nvidia.
Think about that: a company founded before the light bulb is now the go-to supplier for the most advanced AI chips in the world.
The European Advantage
While America dominates AI software, with some exceptions, the nuts and bolts of the infrastructure are dominated by European players. This isn't an accident—it's the result of decades of industrial expertise that suddenly became incredibly valuable.
Schneider Electric cornered the critical power market through that expensive 2006 acquisition. Data centres accounted for about 24 per cent of its orders in 2024, an increase from 23 per cent in 2023 and 19 per cent in 2022. They're not just growing—they're accelerating.
ABB saw data center orders grow from 9% to 15% of their electrification business in just two years. Between 2019 and 2023, data centre orders grew at an average annual rate of 24 per cent, accelerating further in the past year as the AI race has intensified.
Legrand doubled their data center exposure. Data centres orders made up 20 per cent of sales in 2024, double the rate in 2019. Their CFO thinks it could hit 25% by 2030.
These aren't growth rates. These are hockey sticks.
The DeepSeek Reality Check
When China's DeepSeek claimed they trained their AI model for a fraction of GPT-4's cost, tech stocks crashed. Everyone panicked that AI infrastructure would become worthless overnight.
However, preliminary analysis suggests that these types of efficiency gains will likely be offset by increased experimentation and training across the broader AI market. Translation: cheaper AI means more AI usage, not less infrastructure.
It's the efficiency paradox. When Netflix made streaming cheaper, people didn't watch less TV—they binged entire seasons. Same logic applies here.
What This Means for Business Strategy
The Infrastructure Imperative
Hyperscalers, including major cloud service providers, are at the forefront of this growth, currently driving up to 70% of the anticipated demand by 2028. If you're building an AI strategy without thinking about infrastructure, you're building on quicksand.
The smart companies are already moving. Legrand has made 10 small acquisitions in the past year, six of which have been in the area of data centres. They're not just riding the wave—they're building it.
The Geographic Play
Nordic countries are attractive for data center investments, boasting cool climates, renewable energy, and government incentives. While everyone fights over Silicon Valley real estate, the real action is moving to places like Stockholm and Helsinki.
Milan, Warsaw and Berlin are expanding the fastest in 2025, but firms are increasingly looking outside cities. The future of AI might be powered by data centers in the Swedish countryside.
The Investment Angle
European AI startups have secured around 55% more year-on-year investment in Q1 2025, according to data from Dealroom. But the real money isn't going to AI startups—it's going to the companies that power them.
Scaling data center infrastructure at an unprecedented pace is capital intensive and will, we estimate, require more than a trillion dollars in investment across the ecosystem. Someone has to finance that trillion dollars. The companies that control the infrastructure will control the returns.
The Uncomfortable Truth
"We're not putting $70bn or $80bn on the table like Microsoft and Meta," said Franck Lemery, chief financial officer of Legrand. "[But] we [provide] the components and our business is growing relative to that [spending]."
That's the key insight. While tech giants burn billions trying to build the next breakthrough, these infrastructure companies are collecting tolls on every AI breakthrough that gets deployed.
The AI revolution isn't being won by the companies with the smartest algorithms. It's being won by the companies that control the boring stuff—the power, the cooling, the electrical infrastructure that makes AI possible.
The bottom line: Everyone's watching the AI race. But the real winners are the companies building the racetrack. And right now, four European companies that most people have never heard of are building the only racetrack that matters.