As computational demand increased, so did the scale and complexity of these facilities.
Modern are designed to support large numbers of computing systems operating continuously while providing the electrical power, environmental controls, networking, and physical infrastructure necessary for reliable operation.
Although individual facilities differ in design and purpose, they all exist to provide the physical environment in which can occur.
Like the evolution of computing itself, the evolution of data centers has not occurred through replacement but through expansion. As new computational architectures emerged, data centers evolved to support them while continuing to rely upon the infrastructure already in place.
Different organizations perform different kinds of computation. As those computational needs have expanded, different types of data centers have emerged to support them.
Some data centers primarily provide cloud computing, allowing individuals and organizations to store information, access applications, and utilize computing resources over the internet.
Others are built as enterprise data centers, supporting the internal operations of a single business, government agency, university, or research organization.
Additional facilities specialize in particular services. Telecommunications data centers support internet and communications networks. Content delivery facilities distribute streaming media and web content to large numbers of users.
Financial data centers process banking, payment, and market transactions where speed and reliability are critical. Scientific computing facilities support research involving large-scale simulations, engineering analysis, and scientific modeling.
More recently, data centers have been designed to support the enormous computational demands associated with training and operating modern AI systems.
Many modern facilities perform several of these functions simultaneously. Consequently, the term data center describes a broad category of computational infrastructure rather than a single standardized type of building.
The services performed within a facility largely determine its computing hardware, electrical demand, networking requirements, storage systems, cooling design, and physical layout.
Modern data centers rarely rely upon a single type of processor. Instead, they combine multiple processor architectures, each designed to perform different kinds of computational work.
Like the evolution of data centers themselves, the evolution of processors reflects successive expansions of computational capability rather than replacement.
The , is the general-purpose processor that forms the foundation of modern computing.
CPUs execute operating systems, manage software applications, coordinate communication between hardware components, process user requests, perform calculations, and direct the overall operation of computing systems.
CPUs are designed for flexibility. They excel at tasks requiring frequent decision-making, logical operations, database management, web services, financial transactions, and general-purpose computation.
Nearly every computing system—from personal computers to the world's largest data centers—relies upon CPUs to coordinate overall system operation.
The , was originally developed to accelerate computer graphics by performing many similar mathematical operations simultaneously.
Modern graphics require millions of pixels to be processed at the same time, making this architecture highly effective for .
Researchers later recognized that this same architecture could efficiently solve many non-graphical problems involving large numbers of repeated mathematical operations.
Artificial intelligence is one example. Training modern AI models requires enormous numbers of similar calculations across vast quantities of data. GPUs are designed to perform these highly parallel mathematical computations much more efficiently than general-purpose CPUs.
Within AI-focused data centers, CPUs and GPUs work together. CPUs continue managing operating systems, networking, storage, scheduling, security, and overall system coordination, while GPUs perform the large-scale mathematical computations required by artificial intelligence workloads.
GPUs therefore expand computational capability without replacing the role of CPUs.
As computational requirements have continued to diversify, manufacturers have developed processors optimized for specific computational tasks.
These processors are commonly referred to as because they accelerate particular workloads rather than replace general-purpose computing.
Dedicated encryption processors help secure internet communications. Video encoding and decoding processors support video streaming and video conferencing. Networking processors accelerate the movement of information through communication systems.
Signal processors are widely used in wireless communications, audio processing, radar, and navigation systems. Storage controllers optimize the movement and protection of digital information within storage systems.
Although each accelerator is designed differently, they all follow the same architectural principle: performing specialized computations while relying upon CPUs to coordinate the overall operation of the computing system.
Rather than expecting one processor to perform every task equally well, modern computing systems increasingly combine multiple processor types, each contributing its strengths to a larger computational system.
represents the next stage in this architectural progression.
Rather than replacing existing processors, current research envisions quantum processors becoming another specialized computational resource within .
In a hybrid environment, CPUs would continue managing operating systems, networking, storage, communications, security, and overall system coordination.
GPUs and other accelerators would continue performing specialized classical computations. When a computational task contains a portion that may benefit from quantum methods, that portion could be processed by a quantum processor before the results are returned to the classical computing environment.
Whether quantum computing ultimately fulfills current expectations remains an active area of research.
Regardless of the outcome, its proposed role illustrates the continuing evolution of computational architecture: expanding computational capability by integrating new processor types alongside existing ones rather than replacing the systems already in operation.
Processors perform computation, but they cannot operate independently.
Every modern data center depends upon a collection of supporting systems that supply power, remove heat, store information, transport data, protect equipment, and maintain continuous operation.
These systems do not perform computation themselves. Instead, they provide the physical conditions that allow computation to occur reliably.
Like the processors they support, these engineering systems have expanded over time as computational capability has increased.
Modern data centers are therefore best understood as integrated engineering systems in which multiple forms of infrastructure work together to support continuous computation.
Every computational operation consumes electrical energy. As processors perform calculations, they convert electrical energy into computation while simultaneously producing heat.
The greater the computational workload, the greater the electrical demand.
Modern data centers therefore require capable of supplying continuous, stable power to large numbers of computing systems.
This infrastructure may include high-voltage utility connections, substations, transformers, switchgear, battery systems, and backup generators designed to maintain operation during interruptions to the primary electrical supply.
The amount of electrical capacity required varies considerably depending upon the size of the facility, the types of processors installed, and the computational workloads being performed.
Nearly all electrical energy consumed by computing equipment ultimately becomes heat.
If that heat is not removed, processors become less reliable and may automatically reduce performance or shut down to prevent damage.
exist to remove that heat and maintain operating temperatures within the design limits of the equipment. Different facilities accomplish this in different ways.
- Air cooling using large ventilation and heat-exchange systems.
- that transfers heat directly from computing equipment into circulating coolant.
- Closed-loop cooling systems that continually recirculate cooling fluids.
- Evaporative cooling systems that use water to improve heat removal under certain environmental conditions.
The choice of cooling technology depends upon engineering requirements, climate, processor density, operational goals, and local resource availability.
No single cooling design is used by every facility.
Modern computation rarely occurs on a single computer. Large computational tasks are frequently divided among many computing systems that exchange information continuously while working toward a common result.
allows processors, storage systems, and external communication networks to exchange information with very low delay.
provide communication both within the facility and with other data centers, internet providers, cloud services, businesses, research organizations, and end users.
As computational workloads become more , networking has become as important to modern computing as processing power itself.
Computation depends not only upon processors but also upon the ability to preserve information.
Modern data centers employ multiple forms of designed for different purposes.
Some systems provide extremely rapid access to actively used information, while others preserve enormous quantities of data for long-term retention, backup, regulatory compliance, or disaster recovery.
Storage systems are continually expanding in both capacity and performance as organizations generate increasing amounts of digital information.
While storage preserves information over long periods of time, provides temporary working space for processors while computations are actively taking place.
Processors constantly move information between memory, storage, and computation.
The speed with which this information can be transferred often has a significant influence on overall system performance.
As processor capabilities have increased, memory technologies have evolved to provide greater capacity and higher transfer rates capable of supporting modern computational workloads.
Because data centers support essential communications, financial systems, healthcare, government operations, research, and commercial services, they incorporate multiple layers of physical security.
Depending upon the facility, these measures may include controlled access, surveillance systems, environmental monitoring, fire suppression systems, redundant communications, and continuous operational monitoring.
The specific security measures employed vary according to the services provided and the operational requirements of the facility.
Many modern data centers are designed so that the failure of a single component does not immediately interrupt operation.
This engineering principle is known as .
Redundancy may include multiple electrical feeds, duplicate networking equipment, backup cooling systems, redundant storage, additional processors, and reserve power systems.
These components are intended to improve reliability by allowing computation to continue when maintenance or unexpected failures occur.
The degree of redundancy varies according to the purpose of the facility and the level of service it is designed to provide.
Although these supporting systems perform different functions, they are not independent.
Electrical infrastructure powers processors and cooling systems. Cooling systems maintain operating temperatures. Networking allows distributed computation. Memory and storage preserve information. Security protects equipment and operations. Redundancy improves reliability across the entire facility.
Taken together, these systems form the physical foundation upon which modern computation depends.
Public discussions concerning electricity, water, noise, land use, or infrastructure expansion are most meaningful when considered in relation to the engineering systems that produce those observable characteristics rather than the general term data center alone.
Because data centers are large physical infrastructure projects, they often become subjects of public discussion during planning, permitting, and construction.
Residents commonly ask practical questions concerning electricity consumption, water use, noise, land use, traffic, environmental effects, tax incentives, and long-term community impact.
These questions concern the observable characteristics of the facility rather than the computation occurring within it.
The purpose of this paper is not to evaluate whether the benefits of a particular data center outweigh its costs. Those determinations depend upon project-specific engineering, economics, regulation, and local conditions.
Instead, this paper provides the architectural foundation necessary for informed evaluation.
Readers interested in evaluating a specific proposal—including questions to ask, documents to request, primary sources to consult, and methods for evaluating public statements—should refer to the Community Evaluation Guide for Proposed Data Centers.
Before meaningful conclusions can be reached, readers should understand that many important questions cannot be answered from the term alone.
Different facilities vary substantially in their electrical demand, cooling systems, water requirements, noise characteristics, backup power systems, communications infrastructure, operational design, computing hardware, and planned phases of expansion.
Likewise, their effects on surrounding communities depend upon local utility agreements, water availability, zoning decisions, environmental regulations, tax arrangements, transportation infrastructure, and the cumulative impact of nearby development.
For this reason, discussions surrounding proposed data centers are most productive when they focus on project-specific information rather than assumptions based upon the general term .
The engineering, regulatory, and community questions necessary to evaluate a specific proposal are provided in the Community Guide.