Skip to content
Knowledge Hub/Article
Article

What Is a Data Center? Tier Levels and Colocation

A data center houses server and network hardware with power, cooling and security. We explain Tier levels, colocation and how to choose a facility.

Oğuzhan Gerçek··8 min read
What Is a Data Center? Tier Levels and Colocation

Short answer: A data center is a facility that houses server, storage and network hardware and is designed to provide power, cooling, physical security and connectivity without interruption. The common language for describing one is Tier levels: in the Uptime Institute classification, Tier III means a facility that can be maintained without shutting systems down, and Tier IV one that keeps running despite a single failure. Colocation, in turn, means housing your own hardware in space you rent at such a facility.

What is a data center?

What defines a data center is less the servers inside it than the infrastructure that keeps them running:

  • Power: The utility feed, uninterruptible power supplies (UPS) and generators that take over when the grid fails. In a good facility, power arrives on two independent feeds.
  • Cooling: Systems that remove the heat hardware generates, and airflow managed through hot and cold aisle separation.
  • Physical security: Authorized access to the building and racks, entry logs, cameras and fire suppression.
  • Connectivity: Links to more than one internet and telecom provider (carrier diversity), and the option of direct connections between customers.

This infrastructure is known to use a lot of energy, and consumption is rising fast. According to the International Energy Agency's April 2026 assessment, data center electricity demand grew 17% in 2025 to around 485 TWh and is expected to reach around 950 TWh in 2030, roughly 3% of global demand.

Efficiency is measured by PUE (Power Usage Effectiveness): the ratio of the facility's total energy use to the energy used by the IT equipment. A value of 1.0 means no overhead at all. In Uptime Institute's 2026 survey the average annual PUE was 1.52; the year before it was 1.54, and the figure has barely moved for six years.

What are Tier levels?

The Tier classification belongs to the Uptime Institute. According to its own definitions:

  • Tier I (basic capacity): Protected against human error but without redundancy against an unexpected failure. The facility may need to shut down for maintenance.
  • Tier II (redundant components): Redundant components for power and cooling, but a single distribution path.
  • Tier III (concurrently maintainable): Any component can be maintained or replaced without shutting systems down.
  • Tier IV (fault tolerant): On top of Tier III, independent and physically isolated systems let it withstand a single failure.

One widely repeated piece of information needs correcting: availability figures such as 99.671% or 99.982% mapped to Tier levels are not part of Uptime's standard. According to Uptime's own explanation, the current standard does not assign availability predictions to Tier levels, and references to expected annual downtime were removed from the standard in 2009.

Besides Uptime, another common framework is ANSI/TIA-942. The standard, whose revision C was published in May 2024, covers architecture, telecommunications, power, cooling, fire protection and physical security, and rates facilities on four levels from Rated-1 to Rated-4. The logic of the levels runs in the same direction as the Tier classification: basic capacity, redundant components, concurrent maintainability and fault tolerance.

What is colocation?

With colocation, the server, storage and network hardware is yours; you house it in a rack or space you rent at a data center. The European Commission Joint Research Centre's data center energy efficiency guidelines define a colocation provider as one that sells space, power and cooling capacity to customers who install and manage their own IT hardware. Physical security and connectivity are usually part of the service too.

The units you look at in a colocation contract differ from those on a cloud bill:

  • Space: A full rack, a half rack or a number of units (U) within a rack.
  • Power: The power allocated to the rack (kW) and how anything above it is charged. Power and cooling for dense GPU racks need to be calculated separately; we cover this in our GPU article.
  • Connectivity: Which carriers are reachable, and the option of direct connections to other customers or cloud providers.
  • Remote hands: How tasks such as swapping a cable, installing hardware or rebooting are handled by the facility team.

Why do data center outages happen?

According to Uptime Institute's 2025 global survey, 50% of operators had at least one impactful outage in the past three years, down from 53% the year before. One in five of those who had a significant outage said it cost more than $1 million. According to its 2026 outage analysis, power remains the leading cause of outages.

The human factor is large as well. In the 2025 outage analysis, nearly 40% of organizations said they had a major outage caused by human error in the past three years; in 85% of those, the cause was either staff not following procedures or flawed procedures. That is why redundancy should be measured in drills rather than on paper; we describe the method in our failover testing guide.

What to look for when choosing a data center

  • Tier or Rated level, and its basis: Does the level rest on a certification or on a design claim?
  • Power: Does power really arrive on two independent feeds, and has generator start-up time been verified in a drill?
  • Cooling and power density: Is the power available per rack enough for the hardware you plan? Uptime's 2026 survey shows high-density racks spreading fast.
  • Carrier diversity and latency: How many independent carriers are there, and what is the latency to your users and your other systems?
  • Physical security and logs: Are the access logs an auditor will ask for kept, and can they be handed to you?
  • Exit clauses: Some facilities are easy to enter and hard to leave; the time and cost of moving out should be written down from the start.

Data centers and data residency in Türkiye

In Türkiye, choosing a data center often starts with a regulatory question. Presidential Circular 2019/12 on information and communication security requires critical data such as population, health and communication records, and genetic and biometric data, to be stored securely within Türkiye (in Turkish). The Information and Communication Security Guide that accompanies the circular has been published by the Cyber Security Presidency since its version 1.1 in March 2026, and includes separate security measures for server rooms and data centers. Banks must keep their primary and secondary systems inside the country; see our BDDK article for details.

Your own server room, colocation or cloud?

  • Your own server room: Only makes sense if you can genuinely operate power, cooling and physical security. For most organizations, that expertise is far from their core business.
  • Colocation: The hardware is yours, the facility is professional. Usually the most economical option for predictable, long-lived workloads; the hardware lifecycle still sits with you.
  • Cloud: The provider runs the hardware as well. Flexible for variable workloads; for steady workloads the total cost can climb. Our articles on IaaS and private cloud cover this option in detail.

For most organizations the answer is a combination: steady workloads in colocation or a private cloud, variable ones in public cloud. We explain how to run that setup in our hybrid cloud article.

Frequently asked questions

What does data center mean? A facility that houses server, storage and network hardware together with power, cooling, physical security and connectivity.

What is a Tier 3 data center? In the Uptime Institute classification, a concurrently maintainable facility: any power or cooling component can be maintained without shutting systems down.

What does colocation mean? Housing your own server and network hardware in a rack or space you rent at a data center. The hardware is yours; the space, power, cooling and physical security belong to the facility.

What is PUE? Power Usage Effectiveness: the ratio of a data center's total energy use to the energy used by its IT equipment. The closer to 1.0, the more efficient the facility; the average in Uptime Institute's 2026 survey was 1.52.

What is the difference between a data center and the cloud? A data center is the physical facility; cloud is the model of delivering the resources in those facilities as a service. Every cloud service runs in one or more data centers.

Sources