The UK’s broken energy system could burst its AI bubble (www.techradar.com)

As the UK positions itself as a global AI hub—backed by policy deals and major private investment—the industry’s biggest bottleneck isn’t chips or talent but electricity. Global data‑centre electricity demand is forecast to jump 165% by the end of the decade (Goldman Sachs), and training a single GPT‑scale model can consume as much energy as 100,000 homes in a year. The consequence is already visible in the US, where grid strain around data hubs has driven local electricity prices up as much as 267% in five years, and risks forcing widespread subsidies or higher bills for small businesses if unchecked growth continues. The root cause, the author argues, is the UK’s wholesale electricity market: legacy trading logic, dozens of intermediaries and price links to global gas markets that make local renewables’ costs volatile and expensive. That design extracts massive transactional inefficiencies (claimed at ~$1tn globally) and gives no incentive for generators or consumers to optimise for AI’s variable, high‑density demand. The proposed fix is not just more renewables but a rebuilt transaction layer—transparent, modern markets that link compute demand, local generation and pricing so every megawatt of AI demand reduces waste and cost. Firms like tem are already offering routes to immediate savings without long PPAs; without systemic reform, the UK risks losing its competitive edge to cheaper‑powered rivals.