The report warns that the infrastructure powering AI could place unprecedented pressure on global electricity, water and land resources over the coming years. The report is raising concerns about whether the benefits of the technology are being pursued sustainably.
Published by the United Nations University Institute for Water, Environment and Health (UNU-INWEH), the report titled Environmental Cost of Artificial Intelligence: Carbon, Water and Land Footprints examine the environmental burden to support AI systems.
While AI is often viewed as a digital innovation, it relies heavily on physical resources including data centres, electricity networks, cooling systems, water supplies and critical minerals.
AI could consume 3% of global electricity by 2030
According to the report, the energy demands of artificial intelligence are expected to rise sharply over the next decade. By 2030, AI-related infrastructure could account for nearly 3% of the world’s total electricity consumption.
“Projected global data centres’ electricity consumption could exceed 945 TWh by 2030, accounting for almost 3% of projected global electricity use—enough to supply residential electricity to all 1.3 billion people in Sub-Saharan Africa for about 5.5 years,” the report said.
Researchers estimate that the resulting carbon emissions could become comparable to those produced by the United Kingdom annually. The report notes that offsetting such emissions would require the equivalent of approximately 6.7 billion trees growing for a decade.
Understanding the ‘Jevons Paradox’
A key concept highlighted in the report is the ‘Jevons Paradox,’ an economic theory suggesting that improvements in efficiency often lead to greater overall consumption rather than reductions in resource use.
Named after economist William Stanley Jevons, the theory emerged from observations that improvements in coal-use efficiency in England ultimately increased coal consumption by making it cheaper and more accessible.
The report suggests a similar trend could unfold with artificial intelligence. As AI systems become cheaper and more attractive, businesses and consumers use them, driving overall demand higher and potentially offsetting any environmental savings generated by technological improvements.
Growing pressure on water and land resources
Beyond electricity consumption, the report highlights the significant water requirements associated with AI infrastructure. Data centres rely heavily on water-based cooling systems to prevent servers from overheating.
The report also points to increasing land demands, estimating that data centre infrastructure could occupy an area nearly ten times the size of Mexico City.
Concerns over global inequality
Researchers also draw attention to the uneven distribution of AI infrastructure around the world. According to the report, only 32 countries currently host AI-focused cloud infrastructure, with approximately 90% of global capacity concentrated in the United States and China.
This imbalance, the report argues, risks widening the digital divide between countries that develop and control AI technologies and those that primarily consume them. Many developing nations bear the environmental costs associated with mineral extraction, electronic waste.
Why responsible AI development matters
The report stresses that the environmental footprint of AI is influenced not only by how often the technology is used but also by the type of tasks it performs. Different applications, such as text generation, coding assistance, image creation and video processing, require varying levels of computational power and energy consumption.
Additionally, different AI models have distinct environmental costs depending on their size, architecture and processing requirements. To address these challenges, the report calls for greater transparency across the entire AI value chain, from mineral sourcing to recycling and safe disposal.
The report ultimately calls for a shift in how societies think about artificial intelligence — not only as a digital tool, but as a resource-intensive system whose environmental impacts must be carefully managed as adoption accelerates worldwide.
“Investors and financial institutions should accelerate the transition to sustainable AI by incorporating environmental footprints into due diligence. They should finance efficient infrastructure and support low-impact products and services. Stewardship should include expectations for disclosure, efficiency improvements, and lifecycle responsibility,” the report said.
