A UN report warns that AI could consume up to 3% of the world’s electricity by 2030, double current usage, and require vast amounts of water for cooling, potentially straining global drinking water supplies.

AI infrastructure could occupy over 14,500 square kilometers of land by 2030, including space for data centres, power plants, reservoirs, and supply chains.

Electricity demand from AI could reach as high as 945 terawatt-hours annually by 2030, making AI data centres potentially the world’s sixth-largest electricity user and about 3% of global consumption.

National strategies in places like New Zealand and Australia are cited, highlighting a light-touch, principles-based regulatory approach that may overlook growing environmental costs of AI.

Recent policies show gaps: New Zealand’s and Australia’s AI frameworks emphasize governance and public-sector benefits but lack mandatory environmental disclosures or centralized energy/emissions regulators.

Examples include New Zealand’s AI strategy and public-service framework without mandatory environmental disclosures, while Australia pursues a light-regulation approach to AI in government.

The article calls for rethinking AI innovation with environmental stewardship at the core, arguing that sustainable practices must guide AI development and deployment.

Overall, the piece urges a shift in the AI playbook toward a sustainable tech future that places the natural environment at the center of economic and societal planning.

Greener energy paths could bring new trade-offs, such as bioenergy potentially increasing water and land footprints even as carbon emissions fall.

The report highlights the Jevons paradox: efficiency gains in AI could lower costs and spur greater overall use, potentially offsetting savings from efficiency.

Specifically, efficiency gains may lead to expanded AI use, undermining the expected reductions in resource use due to cheaper, more widespread deployment.

Local examples show immediate impacts: Ireland used about a fifth of national electricity for data centres in 2023, with droughts in Uruguay and Mexico raising concerns about water diversion to industry.