A new United Nations report warns that arguments suggesting artificial intelligence (AI) will become more energy-efficient and thus reduce demand are a trap. The report, which quantifies the environmental costs of AI, estimates that by 2030, AI's energy use could double to consume 3% of the world's electricity, produce emissions equal to the UK's, and deplete more water for cooling than the global population's annual drinking water needs.
The Jevons Paradox
The report anticipates that AI will follow the Jevons paradox, an economic principle where technological improvements that increase resource efficiency lead to a rise, rather than a fall, in total consumption. Named after economist William Stanley Jevons, who observed this effect with coal in 19th-century England, the paradox suggests that as AI models become cheaper and more attractive, new uses and higher volumes of use will erode or erase any savings from efficiency advances.
To avoid this trap, the report lays out a roadmap for responsible AI use based on transparency, efficiency by design, equity, lifecycle responsibility, global cooperation, and sustainable use.
Scale of the Problem
Last year, data centres consumed as much electricity as Saudi Arabia, the world's 11th largest electricity consumer. If electricity use doubles by 2030, the associated carbon footprint would require 6.7 billion trees grown over ten years to offset it. Data centres would also need 9.3 trillion litres of water and land nearly ten times the size of Mexico City.
The report also highlights structural inequity: only 32 nations host AI-specific cloud infrastructure, with 90% of that capacity in the US and China. This widens the digital divide, with consumer nations bearing disproportionate environmental burdens from mineral extraction and e-waste.
Responsible AI Use
Two main forces shape AI's operational footprint: how much we use it and how we use it. This includes all tasks from text generation to video processing, each requiring different computational effort. Model choice also matters, as each AI system has distinct energy and environmental costs.
The report argues for full value-chain governance, from mineral sourcing to recycling and safe disposal. It calls for environmental disclosures to become routine in AI development, at both model and task levels, and for incorporating projected AI demand into climate and energy planning.
In Aotearoa New Zealand, the government has launched a national AI strategy and a public service AI framework, but there is no requirement for environmental disclosures or a regulator compiling energy use or emissions. Similarly, Australia's national AI plan focuses on improving public services, such as the National Film and Sound Archive's Bowerbird transcription engine and a proof-of-concept tool for the Department of Veteran's Affairs. Both countries take a light-touch, principles-based regulatory approach, which risks overlooking the growing environmental cost of AI.
The natural environment is foundational to the economy, culture, and wellbeing. It should be at the centre of our thinking. It is time to rethink the AI innovation playbook and shift focus toward a sustainable tech future.
