AI Growth Could Double Data Centre Power and Water Use by 2030, UN Researchers Warn

The rapid expansion of artificial intelligence is expected to double global data centre electricity and water consumption by 2030, according to United Nations researchers cited by Reuters. The warning adds to mounting evidence that AI is no longer only a software or productivity story. It is also becoming a major infrastructure challenge with direct implications for energy systems, water management, land planning and net-zero strategies.

The United Nations University Institute for Water, Environment and Health said data centres could consume 945 terawatt-hours of electricity annually by 2030, roughly comparable to Japan’s total power use. In 2024, data centres consumed about 448 terawatt-hours globally, more than Saudi Arabia’s total electricity consumption, with AI already accounting for around one-fifth of that figure.

By the end of the decade, AI is expected to represent about 40% of total data centre electricity demand. This reflects the rapid growth of AI model training, cloud services and, increasingly, everyday AI inference, where deployed models are used repeatedly by businesses and consumers.

Water Demand is Becoming a Central Sustainability Concern

The report also highlights the water footprint of digital infrastructure. Data centres consumed about 4.5 trillion litres of water in 2024, enough to meet the basic needs of more than 600 million people in Sub-Saharan Africa, according to Reuters’ summary of the UN findings. By 2030, water consumption could rise to 9.3 trillion litres.

Water use varies significantly depending on data centre design, location, electricity source and cooling technology. Facilities using evaporative cooling can place pressure on local water systems, particularly in hot or water-stressed regions. Even where a data centre uses limited water on site, its electricity demand can still create indirect water impacts through power generation, depending on the energy mix.

This is why the sustainability discussion is shifting beyond power purchase agreements and renewable energy claims. For regulators and communities, the key issue is not only whether data centres are powered by clean electricity, but also where they are built, how they are cooled, how much grid capacity they require and whether local water resources can support them.

Carbon Emissions and Land Use are Also Rising

The UN researchers estimate that data centres generated about 189 million tonnes of carbon dioxide emissions in 2024. By 2030, emissions could rise to 399 million tonnes if growth is not managed through cleaner power, efficiency improvements and better infrastructure planning.

The land footprint is also expected to expand. Reuters reported that data centre-related land use could rise from around 6,900 square kilometres in 2024 to more than 14,500 square kilometres by 2030. This includes not only the physical footprint of data centre campuses, but also associated energy infrastructure, transmission networks and supply chains.

For companies with science-based targets, this creates a more complex accounting challenge. AI tools may help improve logistics, energy management, materials efficiency and climate analytics. However, the infrastructure needed to run these tools can increase Scope 2 electricity emissions and create wider environmental impacts that are not always visible in corporate reporting.

Grid Planning is Becoming a Bottleneck

The International Energy Agency has also projected that global data centre electricity consumption could double by 2030, reaching about 945 terawatt-hours and representing just under 3% of total global electricity demand. While this share remains modest at a global level, the challenge is highly concentrated geographically.

Large data centres tend to cluster around fibre networks, cloud regions, tax incentives, clean power access and major customer markets. This can place significant pressure on local grids, especially where connection queues are already long or where electricity demand is rising from electrification, manufacturing, heat pumps and electric vehicles.

The IEA has noted that energy infrastructure often takes longer to plan and build than data centres. A facility can be developed in a few years, while transmission lines, substations and new clean generation capacity may take considerably longer. This mismatch can increase reliance on fossil fuel generation in some markets if demand grows faster than clean power supply.

Policy Pressure is Increasing

Governments are beginning to respond. The European Union has been moving toward stronger energy efficiency standards and sustainability reporting requirements for data centres, including disclosure of energy performance, water consumption and clean energy use. These measures are designed to make the sector more transparent and to help policymakers understand where digital infrastructure could conflict with climate and energy security goals.

The policy challenge is delicate. AI and cloud infrastructure are central to economic competitiveness, research, industrial digitalization and public services. At the same time, unmanaged expansion could raise electricity costs, increase emissions, strain water systems and create new environmental justice concerns if resource impacts fall on communities that receive limited economic benefit.

For data centre operators, the direction of travel is clear. Greater scrutiny is likely around power usage effectiveness, water usage effectiveness, carbon intensity, grid impact, backup generation, waste heat reuse and equipment lifecycle management. Companies may also face growing pressure to locate facilities in regions with abundant low-carbon power and lower water stress.

Implications for Net-Zero Strategies

For businesses adopting AI, the findings underline the need to include digital infrastructure in sustainability planning. AI procurement should consider not only model performance and cost, but also energy intensity, cloud provider transparency, data centre location, cooling approach and renewable electricity sourcing.

For technology providers, efficiency will become a competitive issue. Improvements in chips, servers, cooling systems, workload management and software design can reduce the environmental footprint of AI services. However, efficiency gains may be partly offset if lower costs drive greater AI use, a rebound effect already familiar in energy economics.

For investors, the sector presents both opportunities and risks. Demand for data centres, clean power, grid infrastructure, cooling technologies and energy management systems is likely to rise. At the same time, projects in water-stressed regions or constrained grids may face permitting delays, community opposition or higher operating costs.

The UN researchers do not argue against AI. Their warning is that its physical foundations must be planned responsibly. Data centres depend on electricity, cooling, land, minerals, chips, transmission systems and waste management. Treating AI as purely digital risks underestimating its environmental footprint.

As AI adoption accelerates, the central question for governments and companies is not whether digital infrastructure will grow, but whether it can grow within the limits of local grids, water systems and climate targets. The next phase of AI development will therefore depend as much on energy and resource governance as on algorithms.

Source: www.reuters.com