How Data Centres Consume Water, Electricity and Natural Resources and Their Environmental Impact

Quick Answer

Data centres consume large amounts of electricity, water and other natural resources because they operate thousands of servers, storage systems, networking devices and cooling equipment continuously, often 24 hours a day. Electricity powers the IT equipment and supporting infrastructure, while cooling systems remove the heat generated by processors, graphics processing units (GPUs), storage devices and other electronic components. Water may be consumed directly through cooling towers and indirectly during electricity generation and semiconductor manufacturing.

The environmental impact extends beyond the data centre building. Depending on its energy source, cooling technology, location and operational efficiency, a data centre can contribute to greenhouse gas emissions, freshwater depletion, air pollution, land use, mining impacts and electronic waste. The growing demand for cloud computing, artificial intelligence (AI) and digital services makes responsible resource management increasingly important.

What Is a Data Centre and Why Does It Need So Many Resources?

A data centre is a specialised facility that houses computing infrastructure used to store, process, manage and distribute digital information. It supports services such as websites, online banking, cloud storage, video streaming, email, social media, business applications and artificial intelligence.

Unlike an ordinary office or computer room, a large data centre may contain thousands or even hundreds of thousands of servers. These machines operate continuously and generate considerable heat while processing information.

A data centre requires several supporting systems to remain operational:

• Servers and processors: Execute applications, process requests, run databases and perform calculations.
• Graphics processing units (GPUs): Handle demanding computing workloads, including AI training, machine learning and scientific processing.
• Storage systems: Store databases, documents, backups, videos and other digital information.
• Networking equipment: Transfers information between servers, users and other networks.
• Cooling infrastructure: Removes heat produced by electronic equipment and maintains suitable operating temperatures.
• Power infrastructure: Includes transformers, switchgear, uninterruptible power supplies (UPS), batteries and backup generators.
• Building infrastructure: Includes structural materials, lighting, fire protection, security and ventilation systems.

All these systems require energy, materials and maintenance. Therefore, the environmental footprint of a data centre is not limited to the electricity consumed by its servers.

1. How Data Centres Consume Electricity

Electricity is the most important operational resource for a data centre. Almost every activity inside the facility depends on a continuous power supply.

1.1 Electricity Used by Servers

Servers contain processors, memory, storage devices, network interfaces and other electronic components. They consume electricity while executing instructions, handling user requests, accessing databases and transferring information.

Even when servers are not processing intensive workloads, they may continue consuming electricity while running operating systems, monitoring services, background applications and other processes.

AI workloads can increase electricity requirements significantly. Training and operating large AI models often require specialised GPU servers that perform substantial numbers of calculations simultaneously.

1.2 Electricity Used by Cooling Systems

Electronic components convert much of the electrical energy they consume into heat. If this heat is not removed, equipment temperatures can exceed acceptable operating limits, resulting in reduced performance, hardware failures or automatic shutdowns.

Cooling equipment therefore operates continuously in many data centres. It may include:

• Computer room air-conditioning (CRAC) units.
• Computer room air-handling (CRAH) units.
• Chillers and cooling towers.
• Water pumps and circulation systems.
• Fans and ventilation equipment.
• Heat exchangers and liquid-cooling systems.

These systems consume additional electricity to move air, circulate coolant, operate compressors and transfer heat outside the building.

1.3 Electricity Used by Power Backup and Supporting Infrastructure

Data centres must maintain reliable power because an unexpected interruption can affect websites, banking services, business applications and other critical systems.

Supporting infrastructure includes UPS systems, battery systems, power distribution equipment, transformers and backup generators.

UPS systems consume some energy through conversion losses, charging and maintaining batteries. Generators may consume diesel or other fuels during testing and power interruptions. Fuel storage and distribution also require supporting infrastructure.

1.4 How Much Electricity Do Data Centres Consume?

The International Energy Agency (IEA), in its April 2025 report Energy and AI, estimated that global data centre electricity consumption was approximately 415 terawatt-hours (TWh) in 2024. Its base-case projection indicated that consumption could reach approximately 945 TWh by 2030.

One terawatt-hour equals one billion kilowatt-hours. These figures are estimates based on modelling and available industry data, not a complete direct measurement of every data centre worldwide.

The IEA also estimated that data centres in India consumed approximately 9 TWh of electricity in 2024, equivalent to around 0.5% of India's total electricity consumption at that time. Demand is expected to grow as cloud computing, digital services and AI infrastructure expand.

These figures demonstrate that data centres are becoming an increasingly important source of electricity demand. Their effect can be particularly significant in locations where many facilities are concentrated in the same electricity grid.

2. Why Data Centres Consume Water

Water is primarily required to remove heat from data centre equipment, although the amount varies significantly depending on the cooling technology, local climate, facility design and operating conditions.

Water consumption occurs in two major ways:

• Direct water consumption: Water used at the data centre site, particularly in evaporative cooling systems.
• Indirect water consumption: Water used elsewhere to generate electricity and manufacture the electronic equipment installed in the data centre.

2.1 Water Used in Cooling Towers

Many traditional data centres use cooling towers to remove heat from their cooling systems. These systems rely on evaporation to transfer heat into the surrounding atmosphere.

The process generally works as follows:

1. Servers generate heat while processing information.
2. Air-conditioning equipment collects hot air from the server room.
3. A heat exchanger transfers heat to a chilled-water or cooling-water circuit.
4. A chiller or associated heat-transfer system moves the heat into another water circuit.
5. The heated water reaches a cooling tower.
6. Some water evaporates, removing heat from the remaining water.
7. The cooled water is circulated back through the system.

Water must be added continuously to replace evaporation and other losses. Some water is also deliberately discharged through a process called blowdown to control the concentration of dissolved minerals and impurities in the circulating water.

This is why a cooling tower can consume significant quantities of water even though much of its water circulates repeatedly through the system.

2.2 Water Used in Liquid Cooling

Modern servers, particularly GPU-based AI systems, can generate concentrated heat in a relatively small area. Traditional air cooling may become less practical for certain high-density computing environments.

Liquid cooling transfers heat using a liquid coolant rather than relying exclusively on moving air.

Common approaches include:

• Direct-to-chip cooling: Coolant flows through cold plates attached to heat-generating components such as CPUs and GPUs. Heat is transferred to the coolant and subsequently removed through heat exchangers.
• Immersion cooling: Servers or selected electronic components are immersed in specially designed non-conductive dielectric fluid that absorbs heat.
• Rear-door heat exchangers: Heat is transferred from hot server exhaust air to a liquid circuit through heat exchangers attached to server racks.

These technologies can reduce dependence on conventional air-conditioning and may significantly reduce direct water consumption, depending on the complete cooling system design.

However, liquid cooling does not automatically mean zero water consumption. The facility may still use water in external heat-rejection equipment, and the production of electricity and cooling equipment also has environmental costs.

2.3 Water Used Indirectly for Electricity Generation

A data centre may not use much water at its own premises, but the electricity supplied to it may have required substantial water consumption elsewhere.

For example, many thermal power plants use water for cooling steam-generation and electricity-production equipment. Some of this water is returned to its source, while another portion is lost through evaporation.

Hydroelectric power generation can also be associated with water consumption through evaporation from reservoirs.

The amount of water involved depends on the electricity-generation technology, plant efficiency, cooling design and geographical conditions.

Solar photovoltaic and wind power generally have much lower operational water requirements than many conventional thermal power sources. However, their manufacturing and supply chains also require resources.

2.4 Water Used in Semiconductor and Chip Manufacturing

Data centres depend on semiconductor chips used in processors, GPUs, memory modules and networking equipment.

Manufacturing advanced semiconductor chips requires extremely clean production environments and highly purified water known as ultra-pure water.

This water is used in processes such as wafer cleaning, rinsing and other manufacturing stages where even tiny contaminants can affect chip quality.

Producing ultra-pure water requires treatment processes that remove dissolved minerals, particles and other impurities. The process itself can require substantial water input, and some water is lost during treatment and manufacturing.

Consequently, the environmental footprint of a data centre includes water consumed during the production of the hardware it uses, even when the chip manufacturing facility is located in another country.

3. How Much Water Do Data Centres Consume?

The IEA's 2025 Energy and AI report estimated that global data centre-related water consumption was approximately 560 billion litres annually. Its base-case projection indicated that this could rise to approximately 1,200 billion litres annually by 2030.

This estimate includes water associated with electricity generation, direct cooling, primary energy supply and semiconductor manufacturing. It should not be interpreted as the amount of water physically consumed inside data centre buildings.

The report identified electricity generation and primary energy supply as major contributors to the overall water footprint, with direct cooling and semiconductor manufacturing accounting for additional consumption.

As an illustration, the IEA estimated that an average 100-megawatt hyperscale data centre in the United States consumes approximately two million litres of water per day when direct and indirect water consumption are considered. More than 60% of this estimate was indirect water use.

Actual consumption can be considerably different between two facilities of similar computing capacity because they may use different cooling systems and electricity sources.

Water Withdrawal Versus Water Consumption

These two terms are important when evaluating environmental impact because they describe different quantities.

• Water withdrawal: The total amount of water taken from a source such as a river, reservoir, lake or groundwater supply.
• Water consumption: The portion of withdrawn water that is not returned to the original water system in the same usable condition or location, commonly because it evaporates or is otherwise removed from the immediate water cycle.

For example, a cooling system may withdraw 100,000 litres of water and return some of it after use. If 30,000 litres evaporate or are otherwise consumed, the consumption figure is 30,000 litres, not the entire withdrawal amount.

This distinction is essential when comparing water consumption figures published by different data centre operators.

4. Other Natural Resources Used by Data Centres

Water and electricity are not the only resources required to operate digital infrastructure. Building and maintaining a data centre requires substantial quantities of construction materials, metals, minerals, electronic components and fuel.

4.1 Copper

Copper is widely used in electrical wiring, transformers, power distribution systems, circuit boards, motors, cables and cooling infrastructure.

Large data centres require extensive electrical infrastructure to distribute power safely and reliably. Mining and refining copper consume energy and water and can generate mining waste and environmental pollution if not properly managed.

4.2 Aluminium and Steel

Aluminium and steel are used in server racks, structural components, electrical equipment, cooling systems, enclosures and building infrastructure.

Producing these materials requires mining, processing, transportation and industrial energy. Primary aluminium production, in particular, can be electricity intensive.

4.3 Silicon and Semiconductor Materials

Silicon is a fundamental material used in semiconductor manufacturing. It is processed into wafers from which integrated circuits are manufactured.

Modern semiconductor production also depends on various specialised materials, chemicals and metals, including materials used in chip fabrication, packaging and electronic connections.

Manufacturing these components involves industrial processes that consume energy, water and other resources.

4.4 Rare Earth Elements and Other Critical Minerals

Some data centre components and supporting electrical systems use rare earth elements and other specialised minerals.

Depending on the equipment, these materials may be used in magnets, electronic components, power systems and other high-technology applications.

The IEA has highlighted the importance of minerals such as copper, silicon, rare earth elements and gallium in the expansion of data centre infrastructure.

Mining and processing these materials can disturb land, generate waste, consume water and affect nearby ecosystems. The extent of the impact depends on the mining method, location, environmental controls and recycling practices.

4.5 Concrete and Construction Materials

Data centres require large buildings with strong foundations, structural support, security infrastructure and specialised equipment rooms.

Concrete and steel are commonly used in their construction. Cement production is an important source of industrial carbon dioxide emissions, while steel manufacturing also requires substantial energy.

Additional materials include glass, plastics, insulation, roofing materials, pipes, cables and construction chemicals.

Therefore, environmental damage can occur before a data centre begins processing its first customer request.

4.6 Diesel and Other Fuels

Many facilities maintain backup generators to supply electricity when the normal grid supply fails. Diesel generators are common in some installations, although other technologies and fuels may also be used.

Fuel is consumed during testing and actual operation. Burning diesel releases carbon dioxide and air pollutants, including nitrogen oxides and particulate matter.

The extent of local air pollution depends on generator size, operating hours, fuel quality, emission controls and proximity to surrounding communities.

5. How Data Centres Cause Environmental Damage

Environmental damage does not arise simply because a building contains servers. It results from the resources consumed throughout the data centre's lifecycle and the environmental effects of obtaining, processing and disposing of those resources.

5.1 Greenhouse Gas Emissions and Global Warming

One of the most important environmental concerns is the greenhouse gas emissions associated with electricity consumption.

When a data centre receives electricity generated from coal, natural gas or other fossil fuels, the associated power generation releases greenhouse gases, particularly carbon dioxide.

These gases contribute to the greenhouse effect, increasing global temperatures and influencing climate patterns.

The environmental impact depends heavily on the electricity supply. Two data centres consuming the same amount of electricity can have substantially different carbon footprints if one operates on a fossil-fuel-intensive grid and the other uses lower-carbon electricity.

Emissions can also arise from:

• Diesel generators and other onsite fuel consumption.
• Manufacturing servers, GPUs, storage devices and networking equipment.
• Producing concrete, steel and other construction materials.
• Mining and processing raw materials.
• Transporting equipment and construction materials.
• Replacing electronic equipment and disposing of obsolete hardware.

These additional emissions are often described as embodied emissions or lifecycle emissions. They are important because measuring only electricity consumption during operation does not capture the complete environmental footprint.

5.2 Freshwater Depletion and Water Stress

Data centres located in water-stressed regions can create additional pressure on local freshwater resources, particularly when they depend on freshwater for evaporative cooling.

Water stress occurs when the available water supply is insufficient to meet the needs of households, agriculture, industries and ecosystems.

When a large industrial facility consumes significant quantities of freshwater in such an area, it can increase competition for limited water resources.

Potential consequences include:

• Increased pressure on groundwater supplies.
• Reduced availability of water for other local users.
• Greater competition between industrial, agricultural and domestic water requirements.
• Increased pressure on local water infrastructure.
• Additional stress on rivers, lakes and other freshwater ecosystems.

These consequences are not automatic. They depend on local water availability, the source of water, seasonal conditions, the facility's actual consumption and how water resources are managed.

A facility using recycled wastewater in a water-abundant location may have a substantially different local impact from one extracting groundwater in a drought-prone region.

5.3 Air Pollution

Data centres can contribute to air pollution through onsite generators and the electricity supply chain.

Diesel generators produce nitrogen oxides, particulate matter and other combustion-related pollutants. Fossil-fuel power stations also release air pollutants, depending on their fuel and pollution-control systems.

These emissions can affect air quality and contribute to respiratory and cardiovascular health risks when exposure levels are significant.

The impact is generally more important in locations with high generator usage, inadequate emission controls or a concentration of industrial and power-generation facilities.

5.4 Damage to Land and Natural Habitats

Constructing large data centres requires land for buildings, electrical substations, cooling equipment, backup generators, access roads and other infrastructure.

Depending on the site, construction may involve clearing vegetation, modifying natural drainage patterns and disturbing existing habitats.

Additional land disturbance can occur in the supply chain through mining, mineral processing, fuel extraction and electricity infrastructure development.

In areas with sensitive ecosystems, poorly planned construction can affect wildlife habitats, soil quality and natural water flows.

5.5 Electronic Waste and Hazardous Materials

Data centre hardware has a limited useful life. Servers, GPUs, storage devices, batteries, power supplies and networking equipment eventually require repair, replacement or disposal.

Rapid hardware development and increasing demand for computing performance can encourage frequent equipment upgrades.

Discarded electronic equipment becomes electronic waste, commonly known as e-waste.

Some electronic components contain hazardous substances, while others contain valuable materials such as copper, aluminium, gold and other recoverable metals.

If obsolete equipment is dumped, burned or processed through unsafe recycling methods, hazardous substances may contaminate soil, air and water.

The World Health Organization identifies poorly managed e-waste as a significant environmental and public health concern. Its guidance also highlights the risks of exposure to hazardous substances released during informal recycling and disposal.

5.6 Pressure on Local Electricity Infrastructure

Large data centres require a reliable supply of electricity, often at a scale comparable to that of major industrial facilities.

When several data centres are developed in the same geographical area, their combined demand may require additional electricity generation, transmission lines, substations and grid upgrades.

If infrastructure expansion does not keep pace with demand, the additional load can contribute to local capacity constraints and make electricity planning more difficult.

In some circumstances, the need for additional power generation can prolong the use of fossil-fuel-based electricity or increase dependence on onsite generators.

The effect on household electricity prices is not uniform. It depends on electricity-market rules, infrastructure investment, how new capacity is financed and whether data centre operators pay the full costs associated with their demand.

6. Why AI Is Increasing Data Centre Resource Consumption

Artificial intelligence is an important driver of new data centre investment because AI applications require substantial computing resources.

Traditional workloads such as email, web hosting and file storage also consume electricity, but large AI workloads can require specialised high-performance hardware and high-density server infrastructure.

6.1 AI Training Requires Substantial Computing Power

AI training involves processing large datasets repeatedly to adjust the parameters of a machine learning model.

Large models may require thousands of GPUs or other specialised accelerators operating together. These systems consume electricity and generate considerable heat during intensive computation.

The training process can run for extended periods, requiring continuous power and cooling.

6.2 AI Inference Also Consumes Electricity

Training is not the only energy-consuming activity. Inference is the process of using an already trained AI model to generate a response, prediction or other output.

Every AI request requires computing resources. The energy consumed depends on the model, hardware, input and output size, workload, utilisation and serving efficiency.

When millions of users access AI services, the combined demand can become substantial even if each individual request requires relatively little energy.

6.3 High-Density GPU Systems Increase Cooling Requirements

AI servers often contain multiple GPUs and high-performance processors in a relatively small physical space.

This increases rack power density and heat concentration. Conventional air-conditioning may become less efficient or insufficient for certain configurations.

Operators may therefore deploy direct-to-chip cooling, rear-door heat exchangers or immersion cooling.

These systems can improve heat removal, but their total energy and water performance must be evaluated at the facility level.

6.4 More Computing Demand Can Offset Efficiency Improvements

New processors and cooling technologies can perform more work using less energy per unit of computation. However, overall resource consumption can still increase if the demand for computing grows faster than efficiency improves.

This is an important distinction between efficiency and total consumption. A data centre can become more efficient per calculation while consuming more electricity and water overall because it performs many more calculations.

7. Understanding Data Centre Efficiency Measurements

Data centre operators use specialised measurements to understand electricity and water efficiency. Two commonly used metrics are Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE).

7.1 Power Usage Effectiveness (PUE)

PUE measures the total electricity consumed by a data centre facility compared with the electricity consumed by its IT equipment.

The formula is:

PUE = Total Facility Energy Consumption ÷ IT Equipment Energy Consumption

For example, assume a data centre consumes 1,000,000 kWh of electricity annually, of which 700,000 kWh is used by servers, storage and networking equipment.

Its PUE would be:

PUE = 1,000,000 ÷ 700,000 = 1.43

This means the facility consumes approximately 0.43 kWh of additional electricity for every 1 kWh consumed by its IT equipment.

A PUE closer to 1.0 indicates that a smaller proportion of facility electricity is being consumed by supporting infrastructure. However, PUE does not measure the environmental impact of the electricity source or the efficiency of the computing workload itself.

7.2 Water Usage Effectiveness (WUE)

WUE measures the amount of water consumed at a data centre site in relation to the electricity consumed by its IT equipment.

The commonly used site-based formula is:

WUE = Annual Site Water Consumption in Litres ÷ Annual IT Equipment Energy Consumption in kWh

For example, if a facility consumes 5,000,000 litres of water annually and its IT equipment consumes 10,000,000 kWh of electricity:

WUE = 5,000,000 ÷ 10,000,000 = 0.5 litres/kWh

This indicates that the facility consumes 0.5 litres of onsite water per kWh of IT energy consumption.

Site WUE does not automatically include indirect water consumption associated with electricity generation or hardware manufacturing. Therefore, it should not be treated as a complete measurement of the data centre's water footprint.

7.3 Comparing PUE and WUE

Metric	What It Measures	What a Lower Value Indicates
PUE	Total facility electricity compared with IT electricity	Less supporting electricity relative to IT energy
WUE	Onsite water consumption compared with IT electricity	Less direct water consumption per kWh of IT energy

Neither metric alone establishes whether a data centre is environmentally sustainable. A complete assessment should also consider carbon emissions, electricity sources, indirect water use, equipment manufacturing, land use and electronic waste.

8. Can Data Centres Operate Without Consuming Large Amounts of Water?

Yes. Data centres can reduce their dependence on freshwater through different cooling designs and better water-management practices. However, eliminating direct water consumption does not necessarily eliminate environmental impact.

8.1 Air-Cooled Systems

Air-cooled systems use fans, heat exchangers and air-cooled chillers to remove heat without relying on evaporative cooling towers.

They can substantially reduce direct water consumption. However, depending on the climate and system design, they may require more electricity than water-based cooling systems.

8.2 Recycled and Treated Wastewater

Facilities that use evaporative cooling can investigate treated wastewater or other suitable non-potable water sources instead of relying entirely on drinking-quality freshwater.

Water must be treated to a quality appropriate for the cooling equipment. Mineral concentration, corrosion, scaling, biological growth and discharge requirements must be considered.

Using recycled water can reduce pressure on freshwater supplies, but it does not eliminate water consumption through evaporation.

8.3 Direct-to-Chip and Immersion Cooling

Liquid-cooling technologies can transfer heat more efficiently from high-performance components and may reduce the need for conventional air-conditioning.

When combined with appropriate dry heat-rejection equipment, these systems can reduce direct water consumption substantially.

However, their total environmental performance depends on the coolant, heat-rejection design, pump energy, maintenance and the electricity supply.

8.4 Natural Air Cooling and Economisers

Some facilities use favourable outdoor temperatures to reduce the need for mechanical refrigeration.

Air-side economisers introduce filtered outside air when environmental conditions permit. Water-side economisers use suitable external conditions to reduce chiller operation.

These methods can save electricity and, depending on the system, water. Air quality, humidity, temperature and equipment operating limits must be considered.

9. How Data Centre Operators Can Reduce Environmental Damage

Reducing the environmental footprint of data centres requires improvements in facility design, electricity sourcing, cooling systems, hardware management and resource reporting.

9.1 Use Lower-Carbon Electricity

Operators can reduce electricity-related greenhouse gas emissions by sourcing electricity from renewable and other low-carbon generation technologies.

Solar, wind and suitable hydroelectric or nuclear power can help reduce dependence on fossil-fuel generation. The actual environmental benefits depend on the electricity supply arrangements, grid conditions and lifecycle impacts.

Operators should distinguish between purchasing renewable energy certificates or contractual electricity products and physically receiving low-carbon electricity at the facility.

9.2 Improve Server Utilisation

Unused or underutilised servers can consume electricity without delivering proportional computing value.

Operators can improve efficiency through:

• Consolidating workloads onto fewer, appropriately sized servers.
• Virtualisation and containerisation.
• Removing unnecessary or obsolete services.
• Using energy-efficient processors and storage devices.
• Monitoring server utilisation and idle power consumption.
• Automatically scaling computing resources according to workload requirements.

These measures can reduce unnecessary electricity consumption and delay the need for additional hardware.

9.3 Optimise Cooling Operations

Cooling systems should be designed according to actual equipment requirements and local climate conditions.

Useful improvements include:

• Implementing hot-aisle and cold-aisle separation.
• Preventing hot and cold air from mixing unnecessarily.
• Optimising fan speeds and pump operations.
• Maintaining suitable temperature and humidity set points.
• Using economisers where environmental conditions permit.
• Monitoring cooling-system performance continuously.
• Maintaining cooling towers to minimise unnecessary water losses.

Cooling temperatures should not be changed without checking equipment specifications, manufacturer requirements and the facility's operating limits.

9.4 Monitor Direct and Indirect Water Consumption

Operators should measure water consumption at the facility and assess the water footprint associated with electricity generation and hardware manufacturing.

Water-management plans should consider the local watershed, seasonal water availability, groundwater dependence and competing community requirements.

Where suitable infrastructure is available, treated wastewater and recycled water can reduce dependence on freshwater.

9.5 Extend Hardware Life and Improve Recycling

Server hardware should be maintained, repaired and reused whenever technically and economically appropriate.

Responsible hardware management includes:

• Extending the useful life of servers where performance and security requirements permit.
• Replacing individual components rather than entire systems when practical.
• Reusing or refurbishing suitable equipment.
• Using certified electronic-waste recycling providers.
• Recovering valuable metals and other reusable materials.
• Ensuring secure data sanitisation before equipment is reused or recycled.

Proper disposal is especially important for batteries, electronic circuit boards and equipment containing potentially hazardous substances.

9.6 Reduce Construction-Related Environmental Impact

Environmental planning should begin before construction. Operators can assess the site's ecological sensitivity, water availability, electricity infrastructure and transportation requirements.

Using lower-carbon concrete, recycled steel, efficient structural designs and locally sourced construction materials where appropriate can reduce embodied emissions.

Facilities should also be designed for maintainability, future equipment upgrades and eventual decommissioning.

9.7 Improve Transparency and Environmental Reporting

Data centre operators should publish meaningful information about their resource consumption and environmental performance.

Useful reporting indicators include:

• Annual electricity consumption in kWh or MWh.
• Electricity sources and associated greenhouse gas emissions.
• PUE and the measurement period used.
• Direct water withdrawal and consumption.
• WUE and its calculation boundary.
• Use of freshwater versus recycled or treated water.
• Backup generator fuel consumption and emissions.
• Electronic equipment replacement and recycling rates.
• Embodied carbon associated with construction and hardware.

Transparent reporting helps customers, regulators and surrounding communities understand the actual environmental footprint rather than relying only on general sustainability claims.

10. What Does This Mean for India?

India is becoming an increasingly important market for data centre development because of expanding internet access, digital payments, cloud computing, enterprise technology and AI adoption.

Major urban and commercial regions require digital infrastructure to support financial services, government platforms, telecommunications, online businesses and enterprise applications.

However, data centre development in India also needs to consider local environmental conditions.

10.1 Water Availability Is an Important Site-Selection Factor

Some Indian regions experience seasonal water shortages, groundwater stress and significant competition for freshwater.

Building a large water-intensive facility in such an area can increase pressure on local water resources if the cooling design depends heavily on freshwater supplies.

Data centre developers should evaluate the availability and sustainability of the proposed water source before selecting a location.

10.2 Electricity Demand and Grid Capacity

Large data centres require a reliable power supply. Their development may require additional substations, transmission infrastructure and electricity-generation capacity.

Where electricity is generated using fossil fuels, the additional demand can increase associated greenhouse gas emissions.

Renewable electricity procurement, efficient computing infrastructure and appropriate grid planning are therefore important considerations.

10.3 Environmental Compliance and Groundwater Management

Data centre development and operation must consider applicable environmental, water, pollution-control and groundwater regulations.

In India, groundwater extraction is subject to the guidelines issued by the Central Ground Water Authority under the Ministry of Jal Shakti, including applicable amendments and requirements.

Water use and wastewater discharge may also be subject to the Water (Prevention and Control of Pollution) Act, 1974, and relevant state pollution-control requirements.

The exact permissions and compliance obligations depend on the project, location, water source, capacity and applicable regulations. Developers should verify current requirements with the relevant authorities rather than assuming that all facilities follow identical rules.

10.4 Sustainable Data Centre Development

India can expand its digital infrastructure while reducing environmental pressure through:

• Prioritising locations with suitable electricity and water infrastructure.
• Using treated wastewater where technically and legally appropriate.
• Adopting water-efficient cooling technologies.
• Increasing renewable and other low-carbon electricity sourcing.
• Improving server and cooling-system efficiency.
• Monitoring groundwater extraction and local water availability.
• Recycling and responsibly disposing of electronic equipment.
• Publishing transparent resource-consumption reports.

The objective should be to develop reliable digital infrastructure without placing unnecessary pressure on local communities and natural ecosystems.

11. Does Using Cloud Computing or AI Have an Environmental Cost?

Yes. Cloud computing and AI services operate on physical infrastructure, even when users access them through a browser, mobile application or remote desktop.

When a user uploads a file to cloud storage, streams a video, sends an email or submits a question to an AI service, computing equipment in one or more data centres processes the request.

This requires electricity and may contribute to cooling demand, water consumption and hardware wear.

However, the environmental cost of a particular digital activity cannot be determined accurately from the number of clicks, messages or files alone. It depends on factors such as:

• The complexity and duration of the workload.
• The hardware and model being used.
• The efficiency and utilisation of the data centre.
• The electricity generation mix.
• The cooling technology and local climate.
• How resources are shared among users.

Cloud services can also reduce resource consumption in some situations by consolidating workloads onto shared, efficiently operated infrastructure instead of maintaining underutilised servers in many separate locations.

Therefore, cloud computing is not automatically more environmentally damaging than local computing, nor is it automatically environmentally friendly. A meaningful comparison requires consideration of the complete system and the workload being performed.

12. Frequently Asked Questions

Do Data Centres Consume More Water Than Electricity?

No. Electricity is the primary operational energy requirement of a data centre. Water is used mainly for certain cooling systems and indirectly through electricity generation and hardware manufacturing. Electricity and water consumption are different measurements and cannot be directly compared by volume.

Can a Data Centre Operate Without Water?

A data centre can operate with very low direct water consumption by using suitable air-cooled or dry heat-rejection systems. However, indirect water consumption associated with electricity generation and semiconductor manufacturing may still exist.

Why Do AI Data Centres Require More Cooling?

AI workloads often use high-performance GPUs and specialised processors that consume substantial electricity and generate concentrated heat. This increases cooling requirements, particularly in high-density server racks.

Does Renewable Electricity Make a Data Centre Environmentally Friendly?

Renewable electricity can substantially reduce operational greenhouse gas emissions and generally has lower operational water requirements than many fossil-fuel-based electricity sources. However, data centres still have environmental impacts associated with water use, construction, equipment manufacturing, mineral extraction, land use and electronic waste.

Is Liquid Cooling Better Than Traditional Air Cooling?

Liquid cooling can remove heat efficiently from high-performance processors and may reduce cooling electricity requirements. However, its environmental performance depends on coolant selection, water use, external heat rejection, pump energy and system design. It is not automatically the most sustainable option for every facility.

What Is the Difference Between PUE and WUE?

PUE measures total facility electricity consumption relative to IT equipment electricity consumption. WUE measures onsite water consumption relative to IT equipment electricity consumption. Both help evaluate efficiency, but neither provides a complete lifecycle environmental assessment.

Do Data Centres Contribute to Global Warming?

They can contribute to global warming through electricity generated from fossil fuels, onsite fuel combustion, equipment manufacturing, construction materials and other supply-chain activities. The amount of greenhouse gas emissions depends on the electricity source, facility efficiency and complete equipment lifecycle.

Can Data Centre Expansion Be Environmentally Sustainable?

Data centre expansion can become less environmentally damaging through lower-carbon electricity, efficient hardware, responsible cooling, sustainable water sourcing, longer equipment lifecycles, recycling and transparent environmental reporting. However, sustainability requires measuring total resource consumption and local environmental conditions rather than relying on a single efficiency metric.

Sources

• International Energy Agency (IEA) – Energy and AI (2025). Global electricity demand, AI computing requirements, water consumption, semiconductor manufacturing and critical minerals.
• International Energy Agency (IEA) – Powering Artificial Intelligence. Data centre electricity-related emissions, water stress and environmental considerations.
• U.S. Department of Energy – Cooling Water Efficiency Opportunities for Federal Data Centers. Cooling tower operation, PUE, WUE and water-saving practices.
• Lawrence Berkeley National Laboratory – 2024 United States Data Center Energy Usage Report. Data centre electricity consumption and direct and indirect water use.
• UNEP Copenhagen Climate Centre – Environmental Sustainability of Data Centres: A Need for a Multi-Impact and Life Cycle Approach. Environmental impacts beyond operational electricity consumption.
• World Health Organization – Electronic Waste: Digital Dumpsites and Children's Health. E-waste, hazardous substances and environmental and health risks.
• International Telecommunication Union (ITU) and UNITAR – The Global E-waste Monitor 2024. Global electronic waste generation and formal recycling statistics.