How are hyperscalers sourcing energy for AI data centres, and what pressure does this place on grids and energy markets?

Hyperscalers are sourcing energy for AI data centers through escalating demands on existing electrical grids, with site requirements growing from 100MW to 1GW-scale facilities, leading to 3-5 year waits for grid interconnections due to capacity constraints [2][4]. To mitigate impacts, companies like Anthropic, Microsoft, and OpenAI have pledged to absorb electricity price increases caused by their data centers, protecting local ratepayers from added costs and pursuing alternative power sources amid the AI boom [7][11][12]. Investments in sustainable infrastructure are also rising, with hyperscaler capital expenditures exceeding $600B in 2026, much of it tied to AI, though build times of 18-36 months exacerbate supply bottlenecks [1][5]. This surge places significant pressure on grids, with data center power demand projected to nearly double to 150GW by 2028, straining architecture and prompting innovations like demand-response reductions in AI workloads to avoid disruptions [3][9][10]. Energy markets face risks of higher residential electricity costs, with 78% of Americans concerned about rising bills from data centers, potentially leading to regulatory backlash and economic fallout if scarcity persists [3][11].