By 2030, AI water consumption will match the needs of 1.3 billion people. Its energy consumption will triple the annual consumption of almost 650 million. Scientists from the United Nations University warn this. AI threatens the world's natural resources. One report says global data centers powering AI will consume 945 terawatt-hours of electricity. It is three times the combined annual electricity consumption of Pakistan, Bangladesh and Nigeria.
It is detailed in the report 'Environmental cost of AI energy use: carbon, water and land footprint'. This was done by the United Nations University Institute for Water, Environment and Health (UNU-INWEH). The environmental costs of AI and data centers cannot be understood through carbon emissions alone. To do this, they quantify the carbon, water and land footprint of AI electricity consumption around the world. They highlight the large differences between these footprints in the 20 largest data centers in the world.
carbon emissions
The environmental cost of AI is being systematically mismeasured. They focus on the carbon emissions associated with training complex models. But it carries a water footprint, derived from cooling and energy generation. And a land footprint, derived from energy infrastructure and supply chains.
Energy consumption per query varies considerably depending on the task. A typical chat query is about 200 times more energy intensive than a basic text classification. Generating a single image using AI can require around 1,450 times that amount. A single short AI-generated video can consume as much electricity as 200,000 spam classifications.
The benefits and burdens of AI's massive global expansion are highly uneven. In Ireland, data centers accounted for 21% of total measured electricity in 2023. They exceed the consumption of all urban households. AI infrastructure could generate up to 2.5 million tonnes of e-waste each year by 2030.
Unequal
Only 32 countries in the world host data centers specialized in AI. 90% of that capacity is concentrated in two countries. More than 150 countries currently have little or no access to sovereign computing for AI. It's not just like an economic gap, but also like an environmental justice issue. Excluded countries bear the burden of critical mineral extraction and electronic waste. Strategic benefits go elsewhere.
AI threatens the world's natural resources. But the report concludes that AI, within planetary boundaries, is feasible. The central argument of the report is constructive. Capacity and management can be developed together, but only with measurement, transparency and shared responsibility across the ecosystem.
Leave a Reply