Data centres already consume about 415 terawatt hours of electricity a year, roughly 1.5 per cent of the global total, and the International Energy Agency projects this could reach about 945 terawatt hours by 2030, driving almost half of global demand growth this decade. In a detailed analysis for LSE's Grantham Research Institute, Yanxi Zhou and Tao Zou argue the policy question has shifted from how to power artificial intelligence to who pays for the grid capacity, generation and backup it requires.
The UK picture is stark. Data centres currently use about 2.5 per cent of grid electricity, but proposed AI-driven projects could lift that to 8.8 per cent by 2030, and South Wales alone is projected to host 54 per cent of National Grid Electricity Distribution's additional capacity by 2050. Evidence from the United States shows what is at stake for consumers: existing data centres have raised national wholesale prices by 2 to 6 per cent, with rises of up to 25 per cent in some zones of the PJM grid.
The authors argue that the UK's proposed VAT cut on household electricity is a short-term demand-side fix that leaves the structural costs untouched. They favour cost allocation in which the entities creating new costs pay for them, citing Ofgem's proposed commitment fee of 2.5 to 7.5 per cent of average data centre project costs and Virginia's new per-kilowatt-hour charge, alongside dedicated large-load tariffs, direct clean power procurement and local benefit requirements for host regions.