Jakarta, INTI - The rapid expansion of artificial intelligence (AI) is expected to significantly increase global demand for electricity while placing growing pressure on water resources and land use. A recent study by the United Nations University (UNU) projects that global AI data centers could consume approximately 945 terawatt-hours (TWh) of electricity annually by 2030, nearly three times the combined electricity consumption of Pakistan, Bangladesh, and Nigeria, which together are home to around 650 million people.
According to the report, the environmental impact of AI extends beyond energy consumption. AI infrastructure requires substantial amounts of water for cooling data centers and extensive land resources to support power generation facilities and associated supply chains.
Researchers estimate that by the end of the decade, global AI-related water consumption could equal the annual water needs of 1.3 billion people. Land requirements for AI infrastructure are projected to reach approximately 14,500 square kilometers, nearly twice the size of the Greater Jakarta metropolitan area.
Published on the official website of the United Nations, the report highlights environmental challenges associated with AI that have often received less attention than carbon emissions. While previous studies have focused primarily on greenhouse gas emissions from data centers, the growing pressures on water resources and land use have been less widely examined.
Researchers warn that efforts to reduce emissions through AI infrastructure development may unintentionally create new environmental challenges. Some energy sources may lower carbon emissions but require significantly larger amounts of water and land. This trade-off could pose risks for regions already facing shortages of water resources or productive land.
Daily AI Operations Drive the Majority of Energy Consumption
The report reveals that routine AI usage accounts for the largest share of energy demand. Operational activities are expected to represent approximately 80–90 percent of total AI energy consumption, far exceeding the energy required to train AI models.
Researchers noted that the scale of AI deployment is already substantial, with a widely used AI service estimated to process around 2.5 billion prompts per day while consuming hundreds of gigawatt-hours of electricity annually.
Energy requirements also vary considerably depending on the type of AI service. AI-generated images require significantly more energy than text-based interactions, while AI-generated video content demands even higher levels of electricity consumption.
The study further cautions that technological efficiency improvements alone may not reduce overall energy demand. A phenomenon known as the "rebound effect" can lead to increased AI adoption as technologies become cheaper and more efficient, ultimately driving total resource consumption higher.
Under such conditions, overall demand for electricity, water, and land may continue to rise despite advances in energy-efficient technologies.
Uneven Environmental Impacts Across Countries
The environmental burden of AI development is not expected to be distributed equally. Some countries may face increasing pressure on electricity grids due to expanding data center capacity, while others could experience growing challenges related to water availability.
In addition, the AI industry is projected to generate approximately 2.5 million tons of electronic waste by 2030. Much of this waste may ultimately be processed in lower-income countries that often have limited waste management capabilities.
The study also highlights the growing demand for critical minerals needed to support AI infrastructure. Increased mining activities associated with these materials could contribute to environmental degradation and widen social inequalities in certain regions.
At present, AI infrastructure development remains concentrated in the United States and China, while more than 150 countries still lack sufficient AI capabilities and infrastructure.
Researchers argue that this imbalance raises important questions regarding environmental equity, as some nations may bear the environmental costs of AI supply chains without receiving proportional economic benefits.
Calls for Sustainable AI Development
UNU emphasized that the report is not intended to slow the advancement of AI but rather to encourage technology development that aligns with environmental sustainability.
The organization urges governments to incorporate AI infrastructure requirements into long-term planning for energy systems, water resources, and land use management. Technology companies are also encouraged to develop more resource-efficient systems and consider environmental impacts throughout the entire lifecycle of their products.
At the same time, consumers can contribute by choosing applications and digital services that minimize environmental impacts wherever possible.
Conclusion
As AI adoption accelerates worldwide, its environmental footprint is becoming an increasingly important consideration. Beyond rising electricity demand, AI infrastructure is expected to place growing pressure on water resources, land use, and critical mineral supplies. The UNU study highlights the need for governments, technology companies, and users to prioritize sustainable AI development to ensure that technological progress remains aligned with environmental and resource constraints.
Read more: Digital Edge Expands Data Center Capacity as PLN Doubles Power Supply to 2×725 MVA