Data Centers

Large Data Centers: Energy Use, Environmental Impact, Sustainability, and Social Value

Large data centers are industrial-scale facilities that house thousands—or even tens of thousands—of servers, along with data-storage systems, networking equipment, backup power infrastructure, and sophisticated cooling systems. In practical terms, they are the physical infrastructure behind much of today’s digital world.

Whenever we use services that may appear “virtual”—such as cloud computing, email, video streaming, search engines, social networks, electronic payments, online storage, scientific computing, or artificial intelligence—computers located somewhere in the world must process those requests and store the associated data.

What Are Large Data Centers?

The largest facilities are often referred to as hyperscale data centers. They are designed to process and store enormous volumes of data while operating continuously and reliably.

A conventional large data center may require approximately 10–25 megawatts (MW) of electrical power, while facilities specifically designed for artificial-intelligence workloads can exceed 100 MW. Some of the largest projects currently under development are being planned on an even greater scale.

Inside a modern data center, the main components include CPUs and GPUs, memory and storage systems, high-speed fiber-optic networks, electrical transformers, uninterruptible power supplies, batteries, backup generators, and cooling systems.

Cooling is particularly important because nearly all the electricity consumed by computing hardware is ultimately converted into heat.

How Much Electricity Do Data Centers Consume?

It is important to distinguish between power, measured in watts, and energy, measured over time in watt-hours.

A 100 MW data center operating near full capacity may require an electrical power supply comparable to that of a small city. Its total annual electricity consumption depends on how continuously and intensively the facility is used.

According to the International Energy Agency (IEA), global data centers consumed approximately 485 terawatt-hours (TWh) of electricity in 2025. Under the IEA’s central projection, this figure could rise to around 950 TWh by 2030, equivalent to roughly 3% of global electricity consumption.

This means that data-center electricity use is already substantial and is growing rapidly. However, its impact is often even more significant at the local level than at the global level, because very large electrical loads can be concentrated in a relatively small geographical area.

In the United States, for example, data centers consumed an estimated 176 TWh in 2023, corresponding to about 4.4% of national electricity consumption. Projections from the Lawrence Berkeley National Laboratory suggest that this figure could rise considerably by 2028, although the range of estimates remains wide because of uncertainty surrounding the future growth of artificial intelligence.

Why Do Data Centers Consume So Much Energy?

Most of the electricity consumed by a data center is used directly by computing hardware: processors, graphics-processing units, memory systems, storage devices, and networking equipment.

A significant additional share is required for cooling and electrical infrastructure.

Artificial intelligence has intensified this challenge. Modern AI systems depend heavily on high-performance accelerators, particularly GPUs, which can perform enormous numbers of calculations in a very small physical space.

This leads to extremely high power density: large amounts of electricity are consumed within a limited area, producing correspondingly large amounts of heat.

According to the IEA, the power density of AI servers increased dramatically between 2020 and 2025 and is expected to continue rising.

For this reason, many new data centers are increasingly adopting liquid cooling systems, in which cooling fluid is brought directly to the processors, rather than relying exclusively on air conditioning.

Environmental Impact

The environmental footprint of large data centers can be divided into several major categories.

Electricity Consumption and Greenhouse-Gas Emissions

The climate impact of a data center depends strongly on how its electricity is generated.

A facility supplied predominantly by coal- or gas-fired power plants may have a substantial carbon footprint. By contrast, one powered primarily by low-carbon sources such as wind, solar, hydroelectricity, nuclear power, or geothermal energy can have much lower operational emissions.

It is therefore not sufficient to consider electricity consumption alone. The carbon intensity of the electricity supplyis equally important.

Manufacturing of Computing Hardware

The environmental footprint of data centers begins before the servers are switched on.

Producing CPUs, GPUs, memory chips, storage systems, batteries, networking equipment, steel structures, copper wiring, and cooling equipment requires raw materials, energy, and water.

Semiconductor manufacturing in particular is technologically complex and resource-intensive.

Electronic Waste

High-performance computing hardware is replaced relatively frequently as newer and more efficient generations of processors become available.

This creates significant quantities of electronic waste and makes equipment reuse, refurbishment, component recovery, and recycling increasingly important.

Local Infrastructure

Very large data centers may also require new substations, high-voltage transmission lines, backup generators, water infrastructure, roads, and additional land.

Their environmental impact can therefore be relatively modest when considered as a percentage of global resource consumption, but highly significant for the specific communities and regions in which they are located.

Can a Data Center Be Sustainable?

There is no simple distinction between “sustainable” and “unsustainable” data centers. Some facilities are substantially more efficient than others.

One of the most widely used indicators is Power Usage Effectiveness (PUE):

PUE = Total data-center energy consumption / Energy consumed by computing equipment

The theoretical ideal is a PUE of 1.0.

For example, a PUE of 1.2 means that for every unit of electricity used by computing equipment, approximately another 0.2 units are required for cooling, power distribution, and other supporting infrastructure.

However, PUE alone does not provide a complete measure of sustainability.

A facility may have an excellent PUE while still being powered by a highly carbon-intensive electricity grid or using large amounts of freshwater.

A more comprehensive assessment should therefore consider several factors simultaneously: energy efficiency, water consumption, the carbon intensity of electricity, hardware life-cycle impacts, and the actual utilization rate of the servers.

How Can Data Centers Become More Sustainable?

A number of strategies can substantially reduce their environmental footprint.

These include the use of more energy-efficient processors and servers, advanced liquid-cooling systems, longer hardware lifetimes, equipment reuse and recycling, and greater reliance on low-carbon electricity.

Another promising strategy is waste-heat recovery.

Because computing equipment produces large quantities of heat, some data centers can transfer this heat to nearby buildings, industrial facilities, or district-heating networks instead of releasing it into the environment.

Facilities can also reduce pressure on freshwater resources by using recycled or non-potable water where technically appropriate.

The geographical location of a data center is also important. Cooler climates may reduce cooling requirements, while regions with abundant low-carbon electricity can significantly reduce operational emissions.

Another important concept is flexible computing.

Not every computational task must be performed immediately. Some workloads—such as AI model training, scientific simulations, or batch processing—can potentially be shifted to times when large quantities of renewable electricity are available.

In this way, data centers could become not only large electricity consumers but also more flexible participants in the electrical grid.

Why Are Data Centers Useful?

Despite their environmental costs, data centers provide infrastructure that has become essential to modern society.

They support banking and electronic payments, telecommunications, online commerce, cloud computing, healthcare information systems, industrial management, weather forecasting, scientific research, cybersecurity, data backup, video conferencing, streaming services, and artificial intelligence.

Centralizing computing resources in large facilities can also be more efficient than operating millions of small, underutilized servers in individual offices or organizations.

Large-scale operators can share infrastructure, maintain higher average server utilization, and invest in highly efficient cooling and electrical systems.

The disadvantage is that this efficiency comes with extreme concentration. A single hyperscale facility may require so much electricity that its construction becomes a major issue for regional power-grid planning.

The Central Question: How Much Value Do We Obtain per Unit of Environmental Cost?

The most useful question is therefore not simply whether data centers are “good” or “bad” for the environment.

A more meaningful question is:

How much useful digital activity do we obtain for every kilowatt-hour of electricity, liter of water, and tonne of carbon dioxide associated with the infrastructure?

A poorly utilized data center powered predominantly by fossil fuels and equipped with inefficient cooling systems can have a very large environmental footprint.

By contrast, a highly utilized and energy-efficient facility supplied with low-carbon electricity, located appropriately, designed to minimize water use, and capable of recovering waste heat can achieve a much lower environmental impact for the same amount of computing work.

The issue is becoming increasingly important as artificial intelligence expands.

Current energy projections indicate that global electricity consumption by data centers could rise sharply during the second half of the 2020s, with AI-related computing accounting for a substantial share of this growth.

The challenge for the coming years will therefore not be to eliminate data centers—they have become essential infrastructure—but to ensure that the rapid increase in computing capacity is accompanied by equally rapid improvements in energy efficiency, low-carbon electricity production, cooling technology, water management, and hardware sustainability.

In this sense, the future of data centers is closely connected to a broader question facing modern society: how can we continue to increase the benefits of digital technology without allowing its physical and environmental footprint to grow at the same rate?