Artificial intelligence could drive a sharp increase in global water and energy consumption by the end of the decade, with AI-powered data centres projected to use enough water to meet the annual domestic needs of 1.3 billion people, according to a new report from the United Nations University Institute for Water, Environment and Health.
The report warns that the environmental impact of AI extends beyond carbon emissions and includes significant demands on water, electricity, and land resources.
Researchers said public discussions around AI have largely focused on greenhouse gas emissions while paying less attention to the large volumes of water required to cool data centres and support electricity generation for increasingly powerful AI systems.
According to the study, AI-driven data centres could consume 945 terawatt-hours of electricity annually by 2030. The report said this would be nearly three times the combined annual electricity consumption of Pakistan, Bangladesh, and Nigeria.
The study argues that AI relies on a vast physical infrastructure that includes data centres, advanced computer chips, cooling systems, and power networks, all of which carry substantial environmental costs.
Researchers also found that everyday use of AI contributes more to overall energy demand than the initial training of AI models. Daily interactions with AI systems are estimated to account for 80 to 90 per cent of total energy consumption, with image and video generation requiring significantly more energy than text-based tasks.
The United Nations University said the findings are not a call to halt AI development. Instead, the report urges governments, technology companies, and users to consider water, energy, and land-use impacts alongside technological innovation.
The study raises concerns about whether the environmental resources required to support rapidly expanding AI systems can be managed sustainably as the technology becomes increasingly embedded in daily life.