AI Data Centres’ Relentless Energy Demands May Exceed Expectations

Download IPFS

Artificial Intelligence (AI) data centres, requiring constant power, are driving an unprecedented surge in energy consumption, potentially outstripping current projections. This article explores the escalating energy demands of AI data centres, their strain on global power grids, and the challenges of balancing technological growth with sustainable energy solutions.

AI data centres, the backbone of modern digital infrastructure, consumed approximately 415 terawatt-hours (TWh) of electricity globally in 2024, about 1.5% of total electricity use, according to the International Energy Agency (IEA). By 2030, this could double to 945 TWh, equivalent to Japan’s annual consumption, with AI-specific tasks driving nearly half of this demand. Unlike traditional data centres, AI facilities rely on power-hungry graphics processing units (GPUs), with a single Nvidia H100 chip drawing up to 700 watts—eight times a typical flatscreen TV. “AI’s energy needs are relentless, running 24/7,” said Eric Masanet, a sustainability expert at UC Santa Barbara, noting that training models like ChatGPT can consume as much energy as powering a city for days.

This voracious appetite is straining power grids, particularly in regions like northern Virginia, where data centres already consume 25% of local electricity. The IEA warns that 20% of planned data centres may face grid connection delays due to infrastructure limitations. While tech giants like Amazon and Microsoft aim for carbon-neutral operations by 2030, reliance on natural gas and inadequate renewable capacity raises doubts about meeting these goals. Posts on X highlight public concern, with users like @KatusaResearch noting AI data centres could soon rival the energy use of 40 million US homes.

Solutions like advanced cooling systems and energy-efficient chips offer some relief, potentially cutting 10–20% of emissions, per MIT’s Vinod Gadepally. Yet, the rapid pace of AI expansion outstrips these gains. Nuclear power, favored for its reliability, faces long construction timelines, leaving natural gas as a stopgap despite environmental costs. As AI’s energy footprint grows, policymakers and industry must prioritize grid upgrades and transparent energy reporting to avoid burdening consumers with higher costs or environmental fallout.