How AI Impacts the Environment: Water, Energy, and Carbon

Artificial intelligence affects the environment mainly through electricity consumption, water use, carbon emissions, and the resources needed to manufacture computing equipment.

AI systems run on powerful processors housed in data centers. These facilities consume electricity to perform calculations and operate cooling systems that prevent equipment from overheating. Manufacturing chips, servers, and data center infrastructure also requires energy, water, and materials.

The environmental impact depends on how much computing power is used, where electricity comes from, how facilities are cooled, and how efficiently equipment operates.

For investors, these pressures matter because they can influence operating costs, infrastructure spending, regulatory requirements, and company profitability.

How Does Artificial Intelligence Affect the Environment?

AI's environmental footprint extends beyond the electricity consumed when someone uses a chatbot. It includes manufacturing equipment, constructing data centers, operating computing systems, and eventually replacing or recycling hardware.

The main environmental effects include:

• Electricity consumption: AI processors require energy to train models and respond to users.
• Water consumption: Some data centers use water to remove heat from computing equipment.
• Carbon emissions: Electricity generation, equipment manufacturing, and construction can produce greenhouse gases.
• Resource consumption: Semiconductor production and infrastructure construction require materials, energy, and water.
• Electronic waste: Servers and other equipment eventually need to be reused, recycled, or disposed of.

These effects are connected. For example, a data center might reduce water consumption by installing cooling equipment that requires more electricity. Whether that improves its overall environmental performance depends partly on how the electricity is generated.

More efficient AI chips can also reduce energy consumption per task while encouraging greater AI usage. As a result, total industry electricity demand may increase even when individual calculations become more efficient.

How Much Electricity Does AI Consume?

AI requires substantial computing power, particularly for training large models and operating applications used by millions of people.

However, electricity consumed by AI should not be confused with electricity consumed by all data centers. Data centers also support websites, cloud storage, streaming, business software, and other digital services.

According to the International Energy Agency (IEA), data centers consumed approximately 415 terawatt-hours (TWh) of electricity worldwide in 2024, representing about 1.5% of global electricity consumption.

In its 2025 Energy and AI analysis, the IEA projected that global data center electricity demand could reach approximately 945 TWh by 2030 under its base-case scenario.

That would represent more than double the 2024 level. However, the projection covers all data centers, not AI alone. Actual demand will depend on AI adoption, computing efficiency, and infrastructure availability.

Why Does AI Require So Much Energy?

AI models perform calculations using specialized processors, including graphics processing units (GPUs). Training a large model can involve thousands of processors working together.

Once trained, a model continues using electricity whenever it processes a request. This activity is called inference. As more people and businesses use AI applications, inference can become a major source of computing demand.

Data centers also consume electricity for cooling, networking, and power distribution. Their environmental footprint therefore depends on the entire facility, not just the AI processors.

For a detailed breakdown of electricity consumption and operating costs, see How Much Electricity Do Data Centers Consume?.

How Much Water Does AI Use?

AI data centers can consume water when their cooling systems use evaporation to remove heat from computing equipment.

Some facilities rely on cooling towers or other evaporative systems. Others use dry cooling technologies that consume little water on-site but may require more electricity under certain conditions.

There is no single reliable figure describing the water consumption of every AI data center. Usage varies with facility size, climate, cooling technology, and operating conditions.

Does Every AI Data Center Use Large Amounts of Water?

No. Water consumption depends heavily on the cooling system and how the facility releases heat.

For example, a direct-to-chip liquid cooling system can circulate fluid through a closed loop. The presence of liquid cooling does not automatically mean that large quantities of water are consumed. Water use depends partly on how heat is ultimately transferred outside the facility.

It is also important to distinguish between two measurements:

• Water withdrawal: Water taken from a source, some of which may be returned.
• Water consumption: Water that is not returned to the original source in a usable form, including water lost through evaporation.

For investors, water availability can affect data center locations, construction costs, operating expenses, and project approvals. Facilities in water-stressed regions may face additional restrictions or require more expensive cooling solutions.

For specific consumption examples and cooling comparisons, read How Much Water Does an AI Data Center Use Per Day?.

How Does AI Contribute to Carbon Emissions?

AI contributes to greenhouse gas emissions through electricity generation, semiconductor manufacturing, and data center construction.

When a data center uses electricity generated from fossil fuels, its operations are associated with carbon emissions. Electricity from lower-carbon sources generally produces fewer operational electricity-related emissions for the same amount of energy consumed.

However, the actual carbon footprint depends on the electricity mix, when electricity is consumed, and how emissions are measured.

Why Does the Electricity Source Matter?

Consider two hypothetical data centers consuming identical amounts of electricity.

The first operates on a grid heavily dependent on fossil fuels. The second operates on a grid with a greater share of low-carbon electricity.

Even with identical computing equipment, their electricity-related carbon emissions can differ substantially.

Technology companies may purchase renewable electricity through power purchase agreements (PPAs), which are contracts to buy electricity from energy producers. These agreements can support investment in renewable generation.

However, purchasing enough renewable electricity to match annual consumption does not necessarily mean a data center uses carbon-free electricity every hour.

For investors evaluating corporate sustainability claims, the distinction between annual renewable electricity purchases and the actual carbon intensity of electricity used is important.

Does Manufacturing AI Chips Also Harm the Environment?

Yes. AI's environmental impact begins before computing equipment is installed.

Manufacturing advanced semiconductors requires electricity, water, chemicals, specialized materials, and sophisticated industrial facilities. Producing servers, electrical equipment, and construction materials also consumes resources and generates emissions.

Environmental assessments generally distinguish between:

• Operational impacts: Energy, water, and other resources consumed while equipment is running.
• Embodied impacts: Environmental costs associated with manufacturing, construction, transportation, and producing equipment.

For example, replacing older servers with more efficient models may reduce operating electricity consumption. However, manufacturing replacement servers creates additional environmental impacts.

The overall benefit depends on the energy saved, equipment lifespan, and how the replaced hardware is reused or recycled.

Can AI's Environmental Impact Be Reduced?

Yes. More efficient computing equipment, improved cooling, lower-carbon electricity, and better software can reduce environmental impacts. However, no single solution eliminates them.

More Efficient AI Chips

New processors can perform more calculations using less electricity per calculation. This can reduce energy consumption for individual AI workloads.

However, total electricity demand may still increase if AI usage expands faster than efficiency improves.

Better Data Center Cooling

Efficient cooling systems can reduce the electricity required to manage heat. Operators may use air cooling, liquid cooling, or combinations of technologies.

Some designs prioritize electricity efficiency, while others reduce water consumption. The best environmental outcome depends on facility requirements and local conditions.

Lower-Carbon Electricity

Using electricity from lower-carbon sources can reduce emissions associated with AI operations. Renewable energy, nuclear power, and other low-carbon generation technologies can contribute.

However, grid capacity, electricity availability, reliability, and infrastructure costs influence which options are practical.

More Efficient AI Models

Smaller models, optimized algorithms, and improved inference systems can reduce the computing resources required for certain tasks.

For businesses operating AI applications at scale, even modest efficiency improvements can reduce electricity consumption and operating costs per request.

Longer Equipment Lifetimes and Recycling

Extending the useful life of computing equipment can reduce the need for replacement manufacturing. Responsible recycling can recover valuable materials and reduce electronic waste.

However, operators must balance equipment lifespan against improvements in the efficiency of newer hardware.

Can AI Help the Environment Instead of Harming It?

AI can also support activities that improve environmental efficiency.

Examples include:

• Electricity forecasting: Predicting demand and renewable energy generation to support grid management.
• Industrial efficiency: Identifying equipment problems and reducing unnecessary energy consumption.
• Building management: Optimizing heating, ventilation, and cooling systems.
• Weather analysis: Improving the processing of complex environmental and meteorological information.

However, potential efficiency improvements should not automatically be treated as proven environmental savings.

An AI system may reduce the energy required for a particular process while making that process cheaper and encouraging greater usage.

To determine whether an application produces a net environmental benefit, its measurable savings must be compared with the resources required to develop and operate it.

Why Are AI's Environmental Costs Important for Investors?

Environmental pressures can become financial pressures when companies face higher electricity bills, water restrictions, infrastructure requirements, or delays in building data centers.

These costs can influence the profitability of AI services and the returns companies earn on infrastructure investments.

Electricity Costs and Availability

Electricity is a major operating expense for large data centers. Companies that secure reliable power at competitive prices may have an advantage over operators facing expensive electricity or delayed grid connections.

Long-term power contracts, energy efficiency, and access to electrical infrastructure can therefore affect project economics.

Water Restrictions

Water scarcity can influence cooling system design and construction approvals. Facilities in water-stressed regions may need alternative cooling technologies or additional infrastructure, increasing project costs.

Environmental Regulation

Policies covering emissions, water use, electricity, and infrastructure development can affect data center operations. Requirements vary across jurisdictions and can change over time.

Investors should evaluate the regulations relevant to each company's operating locations rather than assuming identical requirements worldwide.

Capital Spending

Reducing environmental impacts may require investment in more efficient cooling equipment, electrical systems, and computing infrastructure.

These expenditures can create opportunities for equipment suppliers while increasing capital requirements for data center operators.

The financial outcome depends on whether the investments produce sufficient savings, revenue, or other economic benefits.

Which Industries Could Benefit From Greener AI Infrastructure?

Demand for more efficient AI infrastructure may create opportunities across several industries.

• Cooling and thermal management: Suppliers of efficient cooling equipment may benefit from new data center construction and upgrades.
• Electrical equipment: Manufacturers of transformers, power management systems, and distribution equipment can benefit from expanding electrical infrastructure.
• Power generation: Electricity producers may see additional demand, depending on generation costs, contracts, and regulation.
• Water management: Providers of water treatment, monitoring, and efficiency technologies may benefit from efforts to reduce consumption.
• Energy efficiency technologies: Software and equipment providers may find opportunities as operators seek to control electricity costs.

However, growing demand within an industry does not guarantee attractive returns for every company.

Investors should examine actual revenue exposure, profitability, competition, and valuation before assuming that environmental infrastructure spending will translate into stronger stock performance.

What Environmental Risks Should AI Investors Watch?

Environmental constraints can affect different parts of the AI industry in different ways.

• Rising operating expenses: Higher electricity or water costs can reduce data center profitability.
• Infrastructure delays: Grid connection problems and resource shortages can postpone new facilities.
• Regulatory changes: Additional environmental requirements may increase compliance costs.
• Community opposition: Concerns about local water supplies, electricity demand, or environmental impacts can delay projects.
• Uncertain sustainability claims: Reported efficiency improvements may not translate into lower total resource consumption.
• High capital requirements: Investments in more efficient infrastructure may take years to generate financial returns.

These risks do not affect every company equally.

For example, a requirement to install more efficient cooling equipment could increase costs for a data center operator while generating additional orders for a cooling equipment supplier.

Understanding which companies bear the costs and which receive the revenue is essential when evaluating investment opportunities.

What Should Investors Look for in Corporate Environmental Reports?

Corporate sustainability reports can help investors understand how companies manage electricity, water, emissions, and infrastructure efficiency.

Useful indicators include:

These metrics should be considered together rather than individually.

For example, a company may improve its PUE while total electricity consumption increases because it operates more computing equipment. Similarly, water efficiency can improve even when absolute water consumption rises.

Investors should therefore examine both efficiency metrics and total resource consumption.

Reporting boundaries also matter. Companies may disclose emissions from their own operations separately from emissions associated with purchased electricity and suppliers. Comparisons can be misleading when businesses use different reporting scopes.

Will AI Become More Environmentally Sustainable?

AI infrastructure can become more efficient through improvements in processors, software, cooling technologies, and electricity generation.

However, greater efficiency does not guarantee a smaller total environmental footprint. If AI adoption expands rapidly, overall electricity demand, water consumption, and equipment production may continue increasing.

The long-term outcome depends on whether efficiency improvements can keep pace with growth in computing demand.

For investors, this makes environmental performance an increasingly relevant part of evaluating AI companies. Businesses that manage energy costs, resource constraints, and infrastructure investment effectively may be better positioned to sustain profitable growth.

The main takeaway: AI's environmental impact comes primarily from electricity consumption, water use, carbon emissions, and equipment manufacturing. These pressures create financial risks for data center operators but can also generate demand for more efficient infrastructure. Investors should prioritize measurable environmental performance, actual financial results, and realistic growth expectations rather than assuming that every sustainability-related technology represents a profitable investment.

Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. All investments involve risk, including the potential loss of capital.