A 56-page report from the United Nations University Institute for Water, Environment and Health (UNU-INWEH) finds that the growth of AI-driven data centres will sharply increase energy, water, land and carbon footprints. According to Reuters and the UNU report, global data centres used 448 terawatt-hours of electricity in 2025 and produced 189 million tonnes of CO2, and the report projects annual power use could reach 945 TWh by 2030 with AI driving 40% of that demand. The UNU report, cited by Time and AP, projects water consumption could hit 9.3 trillion litres by 2030, roughly matching the basic domestic water needs of 1.3 billion people, and a land footprint exceeding 14,500 square kilometres by 2030. The report includes warnings from UNU researchers about tradeoffs between cutting carbon and increasing water or land use.
What happened
A 56-page report from the United Nations University Institute for Water, Environment and Health (UNU-INWEH) quantifies the physical footprint of AI and data centres and projects major increases in energy, water, land and carbon use. According to Reuters and the UNU report, global data centres consumed 448 terawatt-hours of electricity in 2025 and produced 189 million tonnes of CO2 in that year. The report projects annual power consumption could rise to 945 TWh by 2030, with AI accounting for 40% of that total, Reuters reported. The UNU analysis also projects data centres could use 9.3 trillion litres of water by 2030 and expand to more than 14,500 square kilometres of land, figures cited by Reuters, Time and AP.
Technical details
Editorial analysis - technical context: The UNU report treats the AI footprint as multi-dimensional, separating energy, water (cooling and electricity production), and land (infrastructure and supply chains). The authors highlight that choices to reduce one footprint can worsen another; for example, Time quotes UNU lead author Miriam Aczel: "What surprised us most is how often the choices that look greenest from a carbon perspective end up worse for water or for land." Reuters and Time note that a growing share of data centre energy use is attributable to AI workloads, and that cooling infrastructure plus upstream electricity generation are major drivers of water demand.
Context and significance
UNU director Kaveh Madani, the report's lead author, framed the issue in physical terms: "The public debate still often treats AI as software, but AI is also physical infrastructure: data centres, electricity generation, cooling systems, transmission networks, chips, minerals, land and water," Reuters reports. The scale reported-data centres using more electricity than many countries and potentially matching the drinking-water needs of 1.3 billion people by 2030 (Time)-places AI infrastructure squarely in debates about resource allocation, regional equity and industrial planning. Bloomberg reporting from 2025 complements the UNU findings by showing a concentration of new data centres in water-stressed regions, implying local competition for scarce resources.
For practitioners
Observed patterns in similar transitions: Organizations expanding compute capacity typically encounter tradeoffs across sustainability metrics. Public reporting and the UNU analysis suggest that shifting to low-carbon electricity may reduce CO2 but could increase water and land footprints depending on the energy source, a point that is especially relevant for architects designing large-scale AI deployments.
What to watch
- •Regional siting decisions and permitting processes, especially in water-stressed areas covered by Bloomberg and Time.
- •Reporting and disclosure standards for data centre water and land use, since the UNU report underscores gaps in how sustainability is measured.
- •Technology and cooling choices that alter water intensity; the UNU report highlights tradeoffs that make single-metric accounting (carbon only) misleading.
Editorial analysis: For data scientists and ML engineers, the report raises supply-side constraints that could influence long-term operational costs and regulatory environments. Industry practitioners should monitor infrastructure siting, regional resource limits and evolving environmental disclosure requirements, all of which could affect capacity planning and total cost of ownership for large models.
Scoring Rationale #
The UNU report quantifies large, directly measurable resource impacts of AI infrastructure that matter for capacity planning, deployment location and sustainability engineering. The story is notable for practitioners but not a paradigm shift in models or algorithms.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.