National Grid has 30 gigawatts worth of data centers queued up for power connections across England and Wales—equivalent to two-thirds of Great Britain's peak demand. The queue tripled in size since the UK designated data centers "critical national infrastructure" in late 2024, with applications "far exceeding even the most ambitious forecasts," according to energy regulator Ofgem. Grid operators across Europe face similar bottlenecks as AI labs race to secure compute capacity, but the infrastructure to transport power simply isn't there.
This isn't just a capacity problem—it's becoming an existential threat to Europe's AI ambitions. Data center projects are collapsing because they can't get grid access, according to Taco Engelaar from grid optimization company Neara. While AI companies pour hundreds of billions into compute infrastructure globally, European projects are stuck waiting 7-14 years for new transmission lines. The UK's geography makes this worse, concentrating renewable generation in areas far from where data centers want to locate.
Grid operators are scrambling with stopgap measures: switching metals in power lines, rerouting around congested areas, and dynamically adjusting power flows based on weather conditions. "There's no one simple solution," admits Steve Smith from National Grid Partners. "What you have to do is a lot of everything." But these optimizations can only squeeze so much extra capacity from aging infrastructure originally designed for a pre-AI world.
For developers banking on European cloud capacity, this grid crisis means longer deployment timelines and higher costs. The compute shortage everyone expected from chip constraints is happening at the power grid level instead—and it's going to get worse before it gets better.