Artificial intelligence has a growing environmental footprint because the data centers running AI models consume electricity, require cooling, and depend on equipment that takes energy and materials to manufacture.
The three main environmental concerns are carbon emissions, water consumption, and the resources needed to build and operate AI infrastructure.
However, measuring AI's environmental impact is difficult because most data centers support both AI and traditional computing. Companies also do not always disclose how much electricity or water is consumed specifically by AI workloads.
According to the International Energy Agency (IEA), data centers worldwide consumed approximately 415 terawatt-hours (TWh) of electricity in 2024, representing about 1.5% of global electricity consumption. In its 2025 outlook, the IEA projected that demand could reach approximately 945 TWh by 2030.
These figures include all data center activities, not AI alone. Nevertheless, they illustrate why the environmental consequences of expanding computing infrastructure deserve attention.
How Damaging Is AI to the Environment?
AI contributes to environmental pressures through electricity consumption, water use, hardware manufacturing, and data center construction. The severity depends on where and how the infrastructure operates.
For example, a data center connected to a grid heavily dependent on fossil fuels can produce more operational carbon emissions than a similar facility using lower-carbon electricity.
Likewise, water-intensive cooling systems can create greater environmental pressure in drought-prone regions than in areas with abundant water resources.
The main environmental impacts include:
| Environmental Impact | What Causes It | Why It Matters |
|---|---|---|
| Carbon emissions | Electricity generation, construction, and equipment manufacturing | Contributes to climate change |
| Water consumption | Cooling systems and electricity generation | Can increase pressure on local water supplies |
| Material demand | Chips, servers, buildings, and electrical equipment | Requires mining, manufacturing, and transportation |
| Electronic waste | Replacement and disposal of computing hardware | Creates recycling and waste management challenges |
Not every AI application has the same environmental impact. A short text response generally requires less computing than generating a complex video, although actual consumption varies between systems.
Even if individual AI tasks become more efficient, total environmental pressures can increase when usage grows faster than efficiency improves.
How Much Water Does AI Use?
AI uses water primarily through data center cooling and indirectly through the electricity generation required to operate computing equipment.
Powerful computer chips produce heat. Data centers must remove that heat to maintain reliable operation.
Some cooling systems use evaporation, consuming water in the process. Others rely on air cooling or closed-loop systems that can reduce direct water consumption.
There is no universal amount of water consumed by an AI request. The result depends on:
- The data center's location and climate.
- The cooling technology used.
- The efficiency of the computing hardware.
- The electricity sources supplying the facility.
- The complexity and duration of the AI workload.
Can One AI Request Really Use Water?
Yes. AI requests can be associated with water consumption, although widely circulated estimates require careful interpretation.
A frequently cited academic study estimated that approximately 10–50 GPT-3 responses could be associated with consuming roughly 500 milliliters of water, depending on deployment conditions.
That estimate was based on specific assumptions about computing operations, cooling, and electricity generation. It should not be treated as a universal measurement for every chatbot or modern AI model.
Different hardware, cooling technologies, and operating conditions can produce substantially different results.
For consumers and communities, the total water requirements of large data centers are generally more relevant than assigning a fixed amount of water to an individual prompt.
Why Location Matters More Than a Global Average
The consequences of water consumption depend heavily on where it occurs.
A data center using significant amounts of freshwater in a drought-prone region may compete with households, agriculture, and other industries for limited resources.
In other locations, reclaimed water, alternative cooling technologies, or more abundant supplies can reduce pressure on drinking water systems.
For example, two facilities consuming the same amount of water can have very different environmental consequences if one operates in a water-stressed region and the other in an area with reliable supplies.
Environmental assessments should therefore consider local water availability, cooling methods, and competing demand rather than relying only on global averages.
How Much Carbon Does AI Produce?
AI creates carbon emissions primarily through electricity consumption and the manufacturing and construction of computing infrastructure.
The operational emissions of a data center depend largely on the electricity generation sources supplying its power.
Coal-fired electricity generally produces more greenhouse gas emissions than electricity generated from wind, solar, or nuclear power.
Natural gas typically produces fewer direct carbon emissions than coal during electricity generation, but it remains a fossil fuel and can also involve methane emissions.
AI infrastructure also creates emissions before a facility begins operating.
Manufacturing advanced chips, producing steel and concrete, constructing buildings, and transporting equipment all contribute to its environmental footprint.
These are known as embodied emissions: greenhouse gas emissions associated with producing and constructing infrastructure rather than operating it.
Because AI-specific emissions are not consistently disclosed, there is no single reliable figure for the total carbon footprint of all AI systems worldwide.
Are Technology Companies' Emissions Increasing Because of AI?
Some major technology companies have reported rising greenhouse gas emissions alongside expanding computing infrastructure.
Google's 2024 Environmental Report stated that its total greenhouse gas emissions in 2023 were 48% higher than in 2019. The company identified growing data center energy demand and supply-chain emissions among the challenges affecting its climate goals.
Microsoft's 2024 Environmental Sustainability Report also reported higher total emissions compared with its 2020 baseline, with data center construction and hardware-related supply chains contributing to the increase.
These figures describe company-wide emissions, not emissions caused exclusively by AI. They also refer to historical reporting periods rather than current emissions levels.
Nevertheless, they illustrate an important challenge: expanding computing capacity can increase environmental pressures even when technology companies invest in cleaner electricity and more efficient infrastructure.
Is AI Worse for the Environment Than Other Industries?
AI has a growing environmental footprint, but its overall energy use remains smaller than that of major global sectors such as transportation, manufacturing, and buildings.
However, comparisons must distinguish electricity consumption, greenhouse gas emissions, water use, and other environmental impacts. These are different measurements and cannot be directly substituted for one another.
| Industry or Activity | Main Environmental Pressures | How It Compares With AI |
|---|---|---|
| AI and data centers | Electricity, cooling water, and hardware manufacturing | Smaller overall energy footprint than major industrial sectors, but rapidly growing electricity demand |
| Transportation | Fuel consumption and greenhouse gas emissions | Much larger global direct energy consumption and emissions |
| Steel and cement production | Industrial heat, fuel consumption, and manufacturing emissions | Much larger direct industrial emissions |
| Agriculture | Land use, freshwater consumption, and greenhouse gas emissions | Much larger global freshwater footprint |
| Cryptocurrency mining | Electricity consumption from specialized computing | Similar dependence on computing electricity, but different technologies and economic purposes |
These comparisons do not mean AI's environmental impact is insignificant.
A relatively small share of global electricity demand can still create serious regional problems when consumption grows rapidly or becomes concentrated in specific locations.
For example, a cluster of new data centers may account for a modest share of national electricity consumption while requiring substantial upgrades to a local power grid.
The distinction matters because AI services operate globally, but their environmental consequences are often concentrated in the communities hosting the infrastructure.
Can Renewable Energy Make AI Environmentally Sustainable?
Renewable electricity can reduce the operational carbon footprint of AI infrastructure, but it cannot eliminate every environmental impact.
Technology companies are purchasing renewable electricity and investing in energy infrastructure to support expanding data center operations.
However, buying renewable electricity is not the same as operating entirely on carbon-free power every hour.
A company may purchase enough renewable electricity over a year to match its annual consumption while still drawing electricity from a grid using fossil fuels during certain periods.
Providing low-carbon electricity around the clock can require a combination of renewable generation, storage, nuclear power, and other dependable resources.
Even when electricity-related emissions decline, data centers still need cooling systems, buildings, computing hardware, and materials.
Renewable electricity therefore addresses an important part of AI's environmental footprint, but not the entire problem.
For an explanation of AI electricity demand and the costs of supporting infrastructure, see AI Energy Explained: How Much Electricity AI Uses and Who Pays for It.
Could AI's Environmental Impact Slow Its Expansion?
Yes. Environmental constraints can delay data center construction, increase operating costs, or influence where technology companies invest.
Large computing facilities need electricity, land, cooling resources, and supporting infrastructure. Securing those resources can become difficult in areas experiencing rapid development.
Projects may face opposition when communities are concerned about freshwater supplies, electricity costs, land use, or emissions.
Governments and regulators can respond through permitting requirements, environmental reviews, or restrictions on new developments.
For example, a data center developer may need to install water-saving cooling equipment or finance additional infrastructure before receiving approval to operate.
Such requirements can increase upfront investment and delay the point at which a facility begins generating revenue.
These constraints do not necessarily stop AI expansion, but they can influence project profitability and construction schedules.
The broader role of governments, regulators, and technology companies is explored in Who Controls AI and Can It Be Stopped? What Investors Should Know.
Which Companies Could Benefit From Reducing AI's Environmental Footprint?
Reducing AI's environmental impact requires more efficient cooling, electricity management, cleaner power generation, and better resource management.
These requirements create potential opportunities for several types of businesses.
Cooling and Thermal Management
Advanced cooling systems help data centers manage heat generated by powerful computing equipment.
Vertiv supplies power and cooling infrastructure used in data centers. Demand for higher-density computing systems can support interest in its products, although actual financial results depend on orders, competition, and profitability.
Electrical Efficiency
Efficient power distribution equipment can help reduce energy losses and support reliable data center operations.
Eaton and Schneider Electric operate in electrical infrastructure and energy management markets, supplying technologies used across data centers and other industries.
Low-Carbon Electricity
Power producers with nuclear, renewable, or other lower-carbon generation assets may attract demand from technology companies seeking long-term electricity supplies.
Constellation Energy, for example, operates a substantial nuclear generation portfolio in the United States.
Water Management
Companies supplying water treatment, monitoring, and reuse technologies may benefit when data center operators seek to reduce freshwater consumption.
However, the commercial opportunity depends on local water conditions, cooling technology choices, and the scale of actual investment.
All these businesses serve markets beyond AI. Their financial performance cannot be attributed solely to expanding data center demand.
Investors should also evaluate valuations, competition, profitability, and whether expected infrastructure spending produces measurable revenue.
For a more detailed explanation of ownership and investment exposure, see Who Owns AI Energy Infrastructure and How Retail Investors Can Get Exposure.
What Should Investors Watch?
AI's environmental footprint is becoming an operational and financial consideration for companies building and supplying computing infrastructure.
Five indicators can help investors evaluate the risks and opportunities.
1. Electricity Consumption and Emissions
Corporate sustainability reports can reveal changes in electricity use and greenhouse gas emissions.
Investors should distinguish absolute emissions from emissions intensity. A company may use less energy per computing task while its total electricity consumption continues to increase.
2. Water Consumption
Water disclosures can help identify operational pressures, especially in regions with limited freshwater resources.
Facility location and cooling technology are important when interpreting those figures.
3. Cooling Technology Adoption
New cooling systems may improve efficiency or reduce water consumption, but installation costs and operating requirements can affect profitability.
4. Environmental Regulations and Project Approvals
Permitting delays, stricter environmental requirements, and local opposition can influence construction schedules and infrastructure costs.
5. Actual Investment in Efficiency Improvements
Corporate sustainability commitments become more meaningful when supported by measurable investments and operational results.
For investors, these indicators help distinguish environmental promises from developments that can affect costs, revenue, and long-term earnings.
The Bottom Line: AI's Environmental Impact Depends on How It Grows
Artificial intelligence has real environmental costs. It consumes electricity, uses water directly and indirectly, and requires energy-intensive equipment and infrastructure.
There is no single reliable measurement covering the environmental impact of every AI system. The consequences depend on computing efficiency, electricity sources, cooling technologies, hardware manufacturing, and data center locations.
Although AI's overall energy footprint remains smaller than that of major global industries, rapid infrastructure expansion can create significant local pressures.
For consumers, the most important concerns include electricity generation, water availability, and the effects of new facilities on surrounding communities.
For investors, environmental challenges create both risks and potential opportunities. Data center developers may face higher costs or construction delays, while suppliers of efficient cooling, electrical equipment, and lower-carbon electricity may benefit from growing infrastructure demand.
The central question is whether the AI industry can expand computing capacity while limiting the additional electricity, water consumption, and emissions required to support that growth.
Disclaimer: This article is for informational and educational purposes only and does not constitute financial or investment advice. All investments involve risk, including potential loss of principal.