Environment & Technology

AI Data Centers Leave Goliath-Sized Environmental Footprints, UN Report Warns

If data centers were a country, they would rank sixth-highest in electricity use by 2030. A new UN University report quantifies the environmental cost of AI's rapid expansion.

• Global data centers used 448 trillion watt-hours of electricity in 2025 — more than all but 10 countries — producing about 208 million tons of CO₂.
• By 2030, data center energy use is projected to reach 935 trillion watt-hours, with AI driving 40% of total consumption.
• Water consumption reached 1.2 trillion gallons (4.5 trillion liters) in 2025, and the report warns that water use for cooling was not even fully accounted for.

Every time you ask an AI chatbot a question or generate an image with a diffusion model, you are not just paying for computing power — you are also drawing on water, electricity, and carbon budgets that now rival those of entire nations. According to a United Nations University report released on June 3, the environmental footprint of the world's data centers already rivals some of the world's largest countries, and the situation is projected to worsen dramatically.

Last year, global data centers used 448 trillion watt-hours of electricity. To put that number in perspective: if data centers were a country, they would rank 11th in global electricity consumption in 2025, behind only the largest industrial economies. That electricity use produced about 208 million tons of carbon dioxide — roughly the same annual emissions as Argentina. Producing that much energy consumed about 1.2 trillion gallons (4.5 trillion liters) of water.

The report was led by the United Nations University Institute for Water, Environment and Health in Canada. "If you look at these numbers, we're seeing scales comparable to nations," said study co-author Kaveh Madani, a water scientist and director of the institute. "The demand is enormous."

The growth trajectory is even more concerning. By 2030, data centers will account for nearly 3% of the world's projected electricity use, with consumption reaching 935 trillion watt-hours. If data centers were a country, they would rank sixth-highest in power use. Carbon emissions would reach nearly 440 million tons.

Much of this growth is being driven by artificial intelligence. About 20% of data center energy is currently attributable to AI, but that share is expected to grow to 40% by 2030. AI workloads require specialized hardware — particularly graphics processing units (GPUs) — that are far more power-hungry than conventional server chips. The training of large AI models consumes enormous amounts of electricity in a single location over weeks or months, and the inference phase — every time a user interacts with a deployed model — adds continuous ongoing demand.

The report focused on energy use and did not fully examine the water consumed to cool data centers, meaning the actual environmental cost may be even higher. Data centers require massive cooling systems to prevent servers from overheating, and many facilities use water-intensive evaporative cooling methods. With AI expansion driving construction of new data centers in water-stressed regions, the competition for water resources is becoming a growing concern.

The findings raise a question that the technology industry has been reluctant to confront directly: can the environmental cost of AI be sustained? Efficiency improvements in chip design and cooling technology are making each computation slightly less resource-intensive, but the sheer scale of AI deployment — more models, more users, more applications — is outstripping those gains. The report does not offer simple solutions, but it makes clear that the environmental footprint of AI is no longer a niche concern.