UN report warns AI could soon consume 3% of global electricity and vast water resources

A new United Nations report challenges the assumption that AI will become more energy-efficient over time, instead projecting that AI’s global electricity consumption could double by 2030, reaching 3% of worldwide demand. This surge would generate emissions equivalent to the UK’s annual output and require 9.3 trillion liters of water—nearly ten times Mexico City’s annual drinking needs—primarily for cooling. The report cites the Jevons paradox, where efficiency improvements paradoxically increase total resource use, warning that cheaper AI models will drive greater demand rather than reduce it. The report highlights the scale of the issue, noting that data centers already consume as much electricity as Saudi Arabia, the world’s 11th-largest consumer. If consumption doubles by 2030, offsetting the carbon footprint would require planting 6.7 billion trees over a decade. Structural inequities are also emphasized, with only 32 countries hosting AI-specific cloud infrastructure, and 90% of that capacity concentrated in the US and China. This creates a digital divide where nations consuming AI often bear disproportionate environmental burdens from mineral extraction and e-waste. To address these challenges, the report outlines a roadmap for responsible AI use, focusing on transparency, efficiency by design, equity, and lifecycle responsibility. It calls for global cooperation and sustainable practices, including mandatory environmental disclosures for AI models and tasks. Governments adopting AI, such as Aotearoa New Zealand and Australia, are urged to integrate these principles into national strategies, despite current gaps in regulations or energy-use tracking. The report stresses that AI’s operational footprint depends on both usage volume and model efficiency, with different tasks—like text, image, or video generation—requiring varying levels of computational effort. It advocates for full value-chain governance, from mineral sourcing to recycling, to minimize environmental harm. The twinning of AI capability with environmental stewardship is framed as essential, requiring projected AI demand to be factored into climate and energy planning. While countries like Aotearoa New Zealand and Australia are advancing AI strategies, the report notes that frameworks like New Zealand’s lack requirements for environmental disclosures or regulatory oversight of energy use. Australia’s national AI plan, which includes projects like the National Film and Sound Archive’s Bowerbird system, also does not address sustainability metrics. The report concludes that without proactive measures, the environmental costs of AI could outweigh its benefits, exacerbating global resource scarcity and inequality.