Data center

Data center, a room, building, or group of buildings used to host, manage, and run information technology (IT) infrastructure that enables the delivery of applications, data, and services. A data center’s key IT components include servers, storage systems, and network equipment, such as routers, switches, and firewalls. Because large data centers are essential for the processing, storage, and distribution of high volumes of data, they have become indispensable to the development and growth of generative artificial intelligence (AI). They have also become politically controversial because of concerns about their energy usage and environmental impact.

Facility types

There are several types of data centers, which can be differentiated by their design, the location of data and storage management, and the kinds of workloads they support. Some of the more common types are listed below.

• Enterprise data centers are privately owned or leased facilities that support a single organization, either on the organization’s premises or elsewhere. The in-house IT department manages equipment and infrastructure, and controls information security and data regulation compliance.
• Colocation data centers, also known as multi-tenant data centers, are third-party facilities that lease shared space to businesses seeking to outsource their IT operations or have their computer hardware and servers hosted offsite. Because businesses share the space based on individual needs, operational costs are often lower than those of on-premises data centers.
• Hyperscale data centers are massive cloud computing facilities owned and operated by companies such as Google Cloud Platform, Amazon Web Services, and Microsoft Azure. The facilities make IT infrastructure available to millions of customers through an Internet connection. Built to handle large-scale computing and storage needs, a typical hyperscale data center has at least 10,000 square feet (930 square meters) of floor space and houses at least 5,000 servers.
• Edge data centers are smaller cloud computing facilities built close to businesses and their customers. By processing data close to end users, edge data centers provide near-instantaneous connectivity, which allows for real-time data processing, analysis, and action.
• AI data centers are designed for training and running large language models (LLMs) and machine-learning (ML) models. These facilities use graphics processing units (GPUs) instead of central processing units (CPUs) because they handle the intensive computations that AI models require. GPUs generate more heat than CPUs and thus rely on liquid-cooled systems.

History

Data centers date to the early 1940s when the U.S. Army contracted University of Pennsylvania scholars to build the first all-electronic, general-purpose programmable computer, the Electronic Numerical Integrator and Computer (ENIAC). Work on the ENIAC began in 1943 and included the building of a large cabinet to house its central processing unit (CPU). When the ENIAC debuted in February 1946, the cabinet was called a “mainframe” in reference to its design.

Discover how ordinary objects work, and expand your understanding of the familiar.

SIGN UP

" height="20" alt="HEW Logo">

Thank you for signing up!"

Back to homepage

By clicking "Sign Up", you consent to the Britannica Privacy Policy and Terms of Use.

More From Britannica

cloud computing: Data centres and privacy

Over the next several decades, computers became smaller, requiring less physical space for their operations. With the increasing use of microcomputers in the early 1990s further reducing the space needed for computer operations, organizations began using their mainframe computer rooms to house microcomputers. Microcomputers then became known as “servers,” and mainframe rooms gradually came to be known as “data centers.”

With the growth of cloud computing in the 2000s and of generative AI in the 2020s, the term data center has come to refer to an entire building or group of buildings. There are approximately 12,000 data centers worldwide, including more than 5,400 in the United States.

Data centers are described in terms of the electric power they can consume at their peak operation. The largest individual data centers, those of Google in New Albany, Ohio, and Omaha, Nebraska, and that of Meta Platforms in Columbus, Ohio, require more than 500 megawatts (MW). Facilities as large as 11 gigawatts (GW) are in development.

The global power demand for data centers doubled from about 60 GW in 2020 to 122 GW in 2026. Worldwide power consumption for data centers is expected to double again by 2030. In 2025 AI was about one-third of data center power demand, and that number is expected to grow to one-half by 2030. The energy research firm BloombergNEF has forecast that data centers in the United States could reach a demand of 194 GW, or one-fifth of all American electricity generation, by 2035.

Explore Britannica Premium!

The trusted destination for professionals, college students, and lifelong learners.

SUBSCRIBE

Electricity, cooling, and controversy

In the United States the increased demand for electricity has outpaced the amount of power available. Goldman Sachs Research estimated a 9.3 GW shortfall in 2026, and that number would increase to 45 GW, enough to power 169 million homes, by 2028. PJM Interconnection, the largest power grid operator in the United States, has proposed cutting off grid-dependent data centers in the event of possible blackouts.

In response, some data centers have moved “behind the meter”; that is, they generate their own power and do not rely on the electrical grid. Some facilities have turned to renewable energy. Amazon, Google, and Microsoft have funded nuclear reactor projects to power their data centers. However, many facilities are using on-site natural gas turbines.

Data center operations generate a significant amount of heat. Therefore, their infrastructures are equipped with specialized liquid cooling systems to ensure uninterrupted operations and maintain optimal performance. Systems that use evaporative cooling use a lot of water because the water absorbs heat and is then allowed to evaporate and carry the heat away. Closed-loop systems, in which water circulates through pipes and is never exposed to air, use much less water.

Large data center projects have become controversial for how their noise pollution and their electricity and water usage affect nearby communities. In a May 2026 Gallup poll, about 70 percent of Americans would oppose nearby construction of an AI data center. Legislation for moratoriums on data center construction has been introduced in 15 states, and in 2026 New York announced a one-year moratorium on hyperscale data centers to allow time for the introduction of more comprehensive regulation on such facilities and their impacts.