The AI battle has moved beyond model benchmarks. CNBC reports that Europe is struggling to compete on AI data centers because electricity is too expensive.
The numbers are stark
According to the International Energy Agency, European industrial electricity prices last year were double those in the US and 50% higher than in China and India.
Real estate research firm CBRE predicts that in Frankfurt, London, Amsterdam, Paris, and Dublin—the five largest European data center markets—capacity prices will rise another 12% in 2026.
This is not cyclical. CNBC cites the International Data Center Authority (IDCA), noting that global data centers now consume 2% of total electricity, up from 1.7% in 2024. In data center hubs like Texas, Virginia, Slough (UK), and Paris, regional electricity prices have been driven up 20% to 40% by AI alone.
Europe is falling behind on benchmarks and squeezed on electricity costs—US power is cheaper, Chinese power is cheaper, and even Indian power is cheaper.
Paulson speaks plainly
Yesterday Bloomberg aired a video in which former US Treasury Secretary Hank Paulson made a striking comment: the US still leads in AI technology, but electricity shortages will become the next real ceiling—and China holds that card.
He didn't touch geopolitics, just stated that data centers need not just 'available power' but 'affordable, always-on, and scalable for the next decade.'
Paulson is not the first to say this. Goldman Sachs, Bloomberg, and Stanford have published similar analyses—China has been ramping up power generation and grid capacity to match AI data center demand. But Paulson's remarks signal that this view has moved from analyst circles into Washington's decision-making.
Europe's dilemma: squeezed from within and without
Adding to the pain, CNBC attributes the surge in European electricity prices to the US-Iran war, which has driven up energy import costs. The entire European industrial sector is struggling, with data centers being the most sensitive segment.
The result: AI investments already in Europe or planned are being reassessed. CNBC states that 'major data center projects are being redirected away from Europe.'
Where are they going? Some to North America—Texas and Virginia, though prices are rising, are still cheaper than Europe. Others to the Middle East and Southeast Asia, where Saudi Arabia and the UAE have built gigawatt-scale parks to capture customers pushed out of Europe.
The real problem
Counterintuitively, AI data center location decisions are no longer based on 'proximity to users and low latency' as they were five years ago, but on 'where power is cheapest, most stable, and scalable for the next decade.'
Over the past two years, EU regulators have focused on the AI Act, risk classifications, and copyright lawsuits. But what will determine whether Europe can retain the AI industry is the power grid—a slower, harder regulatory challenge that will take a decade.
The next question is not 'when will Europe produce an OpenAI,' but 'will new data center parks break ground in Europe?' The metric has shifted from GPU count to megawatt capacity. Paulson has already taken this argument to television.
Sources: High energy prices could derail Europe's AI race with U.S. and China (CNBC); CocoLoop; China's Energy Boom Could Give It the AI Edge (Bloomberg)