\n\n
Developments: 2026‑27 U.S. power demand is projected to break records, rising from 419.5 billion kWh in 2025 to 428.8 billion kWh in 2026 and 435.6 billion kWh in 2027, with AI data centres identified as the primary driver. Commercial sales, which include data centres, are set to hit a historic 154.9 billion kWh. Global forecasts echo the trend – Gartner sees data‑centre consumption climbing 26 % to 565 TWh in 2026 and potentially topping 1,200 TWh by 2030, while AI‑optimised servers will account for roughly 31 % of that use.
Why it matters: The surge means electricity access is becoming a strategic asset for AI developers. In the United States, Berkeley Lab warns AI‑heavy data centres could consume up to 11.8 % of national electricity by 2030, reshaping grid planning and pricing. On a global scale, the IEA predicts data‑centre power needs will more than double by 2030, matching Japan’s total current electricity demand, underscoring the urgency for greener power sources and more efficient hardware.
What to watch next: 1) Grid‑capacity battles – With a U.S. interconnection queue of ~2,600 GW, firms will vie for priority grid slots, potentially inflating the value of AI‑ready sites (valued at ~$27 M/MW vs. <$3 M for idle mining capacity). 2) Efficiency breakthroughs – Per‑query energy use remains modest (0.24 Wh for a Gemini text prompt), but agentic tasks can hit 50 Wh; advances that cut these figures could blunt demand growth. 3) Policy shifts – Legislative moves, such as bipartisan energy bills fast‑tracking AI data‑centre permits, will shape how quickly new capacity can be added and whether renewable‑heavy supplies can meet the rise.
Overall, the accelerating power appetite of AI is setting new benchmarks for electricity consumption, making energy strategy a pivotal factor in the next wave of AI competition. [[1, 2]]