Contrary to common beliefs, a United Nations report cautions that artificial intelligence (AI) may increase its electricity consumption twofold by 2030, accounting for about 3% of global power usage. This surge could result in greenhouse gas emissions equivalent to the United Kingdom’s output. The report also highlights the risk of AI using more water for cooling than the world’s annual drinking water requirements.
The report mentions the “Jevons paradox,” where technological advancements enhancing resource efficiency paradoxically lead to increased resource consumption rather than reduction. Economist William Stanley Jevons first observed this phenomenon in 19th-century England with coal usage, where cost savings from efficiency gains drove up coal demand.
As AI models become more cost-effective and widespread, new applications and increased usage are expected, offsetting any efficiency gains. To address this, the report outlines a roadmap for responsible AI deployment based on principles like transparency, efficiency, equity, global cooperation, and sustainability.
Data centers, consuming electricity equivalent to Saudi Arabia’s usage last year, could face a substantial increase in electricity demand by 2030. This growth would necessitate planting 6.7 billion trees over a decade to offset the associated carbon footprint. Additionally, data centers might need about 9.3 trillion liters of water and land nearly ten times the size of Mexico City to support this expansion.
The report also raises concerns about a digital and environmental gap, with only 32 countries hosting AI-specific cloud infrastructure, predominantly in the US and China. This concentration could lead to an unequal environmental impact on nations utilizing AI services due to mineral extraction and e-waste. The report stresses the need for comprehensive environmental disclosures and full governance across the AI value chain, from mineral sourcing to recycling and disposal.
