AI's growing demand for electricity, water, and land projected to soar

The rapid expansion of artificial intelligence is poised to dramatically escalate global electricity demand, with projections indicating a near threefold increase by 2030, reaching 945 terawatt-hours annually. This surge in power consumption is comparable to the combined needs of Pakistan, Bangladesh, and Nigeria, nations housing approximately 650 million people.

However, the environmental footprint of AI extends beyond energy. The infrastructure supporting AI requires vast amounts of water for cooling data centers and extensive land for power generation facilities and their associated supply chains. Researchers estimate that by the end of the decade, global AI water consumption could equal the needs of 1.3 billion individuals, while the land required could span 14,500 square kilometers, nearly double the size of the Jabodetabek region.

A recent United Nations University report highlights these often-overlooked environmental impacts. While previous studies focused primarily on carbon emissions from data centers, the strain on water and land resources has received less attention. The report cautions that efforts to mitigate emissions through AI infrastructure development might inadvertently create new environmental pressures. The adoption of certain energy sources, while reducing greenhouse gases, can simultaneously increase water and land requirements, posing risks to regions already facing scarcity.

“

The scale is staggering: one widely used AI service is estimated to process around 2.5 billion requests (prompts) per day, consuming hundreds of gigawatts per hour of electricity annually.

Furthermore, the report reveals that everyday AI applications are the primary drivers of energy consumption, accounting for an estimated 80 to 90 percent of the total energy demand, significantly more than the energy used for training AI models. The energy intensity varies by AI service, with image generation consuming more power than text-based requests, and video generation demanding even higher consumption.

Despite potential technological advancements, the report warns that increased efficiency may not curb the escalating energy demand. The "rebound effect" suggests that as AI technologies become cheaper and more efficient, their usage will likely increase, potentially leading to an overall rise in resource consumption.

“

In such conditions, total resource consumption can continue to increase even as technology becomes more energy-efficient.