Europe has spent the past two years insisting it will be a serious contender in artificial intelligence, alongside the United States and China. A growing chorus of energy analysts told CNBC this month that the continent faces a problem more basic than algorithms or talent: it cannot supply electricity cheaply enough to run the data centers that AI requires.
The gap is large and getting larger. According to the International Energy Agency, electricity prices for energy-intensive industry in Europe are on average roughly double those in the United States and 50 percent higher than in China and India. In May, the average wholesale price per megawatt-hour was $111.65 in the UK and $88.97 in Germany, against $44.19 in France and just $28 in the US. Because data centers consume power at enormous scale, the cost of a kilowatt-hour now decides where they get built. "If I were making the next $7 billion data center, it would be in the U.S. or China," Franklin Templeton strategist Michael Brown told CNBC.
The consequences are already visible. OpenAI has paused its Stargate project in the UK, citing energy costs and a difficult regulatory environment. The cost of securing data center capacity in Europe's five largest markets, Frankfurt, London, Amsterdam, Paris and Dublin, is set to rise 12 percent this year. And the underlying scale gap is daunting. "The scale of what we're seeing in the U.S. compared to Europe is like 1 to 100," said Olivier Darmouni of HEC Paris. Matching American capacity would require far more investment than soaring energy prices make plausible.
Europe is not losing uniformly. The contest is producing winners and losers inside the bloc. France, with its heavy reliance on nuclear power, and the Nordic countries, with cheap and diverse energy mixes, are pulling ahead. "Nearly every big AI company is in Norway," noted an Nvidia programme manager, and Microsoft has committed billions there along with expansions in Sweden and Denmark. In Finland, winter electricity prices occasionally turn negative, meaning utilities pay customers to use power. Germany and the UK, saddled with the highest costs, are the laggards. One Nvidia manager put it bluntly: "the middle part of Europe has already lost the game."
Darmouni argues the moment should be treated as a "wake-up call" to think about energy as a matter of economic sovereignty, not just a utilities question. But Europe also faces a ceiling on the other side. The International Data Center Authority found that community and political pushback against data centers tends to intensify once they exceed 5 percent of a country's electricity use. The UK is already at 5.8 percent and the US is near 6 percent. Europe is caught in a double bind: too little cheap power to attract the investment, and a public liable to turn hostile if the data centers ever do arrive at scale.
There is a more personal version of the same problem. Darmouni expects AI providers to eventually price their services by region, with prices tracking the local cost of electricity, the true marginal cost of running a model. In that future, a user of a chatbot such as Anthropic's Claude could pay more for the same answer in London than in Lyon. Europe's place in the AI race, in other words, may end up decided not by its researchers but by whether it can build affordable power fast enough to keep them in work.