- AI data centres may require 1,000TWh annually by 2030 – equivalent to Japan’s current consumption
- Renewable energy alone cannot meet demand, forcing reconsideration of nuclear and grid modernisation
What happened: AI’s staggering energy appetite
The World Energy Council’s new report reveals AI’s electricity demands are growing exponentially, with data centres potentially consuming 10% of global power by 2030, according to their findings. This equals 1,000 terawatt-hours annually – more than Germany and France’s combined usage – driven by increasingly power-hungry AI models. Nvidia’s latest GH200 Grace Hopper superchips, for instance, consume up to 1,000 watts each during intensive training sessions, as detailed in their technical specifications.
The analysis highlights particular strain in markets like Ireland, where data centres already use 18% of national electricity, per EirGrid’s 2024 report. Several US states have begun delaying new data centre approvals until grid upgrades are completed.
Also read: DOE identifies sites for AI data centres
Also read: Oklo and Vertiv join forces for AI data centres
Why it’s important
The AI energy crisis presents a dual challenge for global infrastructure and climate commitments. Current projections indicate renewable energy sources alone cannot meet the massive power demands of AI data centres, which require consistent energy flows that often exceed what solar and wind can provide during low-production periods. This shortfall is forcing governments and tech firms to reconsider previously sidelined options like next-generation nuclear reactors and enhanced geothermal systems, while also accelerating development of advanced energy storage solutions.
The infrastructure implications are equally profound, with ageing power grids requiring fundamental redesigns to handle AI’s intensive load patterns. Traditional grids built for steady industrial consumption now face unprecedented volatility from AI workloads that can spike from 10MW to 100MW in minutes. Without coordinated action between policymakers and industry leaders, the world risks creating an energy bottleneck that could artificially constrain AI innovation while undermining hard-won progress on emissions reductions.
