Data Centre Glossary
Essential terminology for understanding the data centre industry, from infrastructure fundamentals to transaction structures.
AI Infrastructure
Specialised data centre configurations designed to support artificial intelligence and machine learning workloads, typically featuring high-density GPU clusters (50-100+ kW per rack), liquid cooling systems, high-bandwidth networking (400G+), and dedicated power distribution. AI infrastructure demands are driving unprecedented growth in data centre capacity and reshaping facility design standards.
Availability Zone
A physically isolated section within a cloud provider's data centre region, designed to be independent from other zones in terms of power, cooling, and networking. Deploying across multiple availability zones provides resilience against localised failures. Major cloud providers typically offer 2-3 availability zones per region.
Build-to-Suit (BTS)
A data centre developed to a tenant's specific requirements, including power density, cooling, security, and connectivity specifications. BTS facilities are typically pre-leased on long-term agreements before construction begins, with the developer retaining ownership of the asset.
Carrier Hotel
A data centre that serves as a major telecommunications hub where multiple network carriers and internet service providers converge. Carrier hotels are characterised by extremely high fibre density and are critical points of interconnection. Examples include 60 Hudson Street and 111 8th Avenue in New York.
Carrier Neutral
A data centre facility that allows tenants to choose from multiple telecommunications providers rather than being locked into a single carrier. Carrier-neutral facilities typically host multiple network operators in their meet-me rooms, enabling competitive pricing and redundant connectivity.
Cloud On-Ramp
A direct, private connection from a data centre to a public cloud provider (AWS, Azure, Google Cloud), bypassing the public internet. On-ramps provide lower latency, higher bandwidth, and improved security compared to internet-based cloud connectivity. Major colocation providers offer on-ramp services as a premium interconnection product.
Colocation
A service model where businesses rent space, power, and cooling in a third-party data centre to house their own servers and networking equipment. Colocation allows organisations to benefit from enterprise-grade infrastructure without the capital expenditure of building their own facility.
Cross-Connect
A direct physical cable connection between two tenants or between a tenant and a network provider within the same data centre facility. Cross-connects enable low-latency, high-bandwidth communication without traversing the public internet.
Dark Fiber
Unused or unlit optical fiber that has been installed but is not currently carrying data. Organisations lease dark fiber to create private, high-capacity network connections between facilities, providing dedicated bandwidth and enhanced security compared to shared network services.
Demand Response
Programmes where data centre operators agree to reduce or shift power consumption during periods of grid stress in exchange for financial incentives from utility providers. Demand response is becoming increasingly important as data centres represent a larger share of grid load, and some jurisdictions require participation as a condition of grid connection.
Edge Computing
A distributed computing model that processes data closer to the end user or data source, rather than in a centralised data centre. Edge facilities are typically smaller (1-5 MW) and located in metropolitan areas to reduce latency for applications such as autonomous vehicles, IoT, and content delivery.
Free Cooling
Cooling techniques that utilise outside air or naturally cold water temperatures to remove heat from data centre equipment, reducing or eliminating the need for mechanical refrigeration. Free cooling is most effective in cool climates (Nordic countries, Pacific Northwest) and can reduce cooling energy consumption by 50-80%.
Green Data Centre
A facility designed and operated to minimise environmental impact through energy efficiency, renewable energy procurement, water conservation, and sustainable building practices. Certifications include LEED, BREEAM, and Energy Star. Key metrics include PUE below 1.3, high renewable energy percentage, and water-free cooling where possible.
Hot Aisle / Cold Aisle
A rack layout strategy that separates hot exhaust air from cold intake air to improve cooling efficiency. Server racks are arranged so that cold aisles face the air conditioning output and hot aisles face the return air intakes, preventing warm air from mixing with cooled supply air.
Hyperscale
A class of data centre typically exceeding 20 MW of critical IT capacity, designed to serve the massive compute and storage requirements of cloud providers, social media platforms, and large enterprises. Hyperscale facilities are characterised by highly standardised, modular designs optimised for efficiency at scale.
Interconnection
The practice of directly connecting networks, cloud providers, and enterprises within a data centre or across data centres. Interconnection reduces latency, improves performance, and lowers bandwidth costs compared to routing traffic over the public internet.
Latency
The time delay between a request being sent and a response being received, typically measured in milliseconds. In data centre contexts, latency is influenced by physical distance, network hops, and processing time. Low latency is critical for financial trading, gaming, and real-time AI applications.
Liquid Cooling
A cooling method that uses liquid (water, dielectric fluid, or refrigerant) to remove heat directly from IT equipment, as opposed to traditional air cooling. Liquid cooling is increasingly essential for high-density AI and HPC workloads where air cooling alone cannot adequately dissipate heat at rack densities above 30-40 kW.
Meet-Me Room (MMR)
A designated area within a data centre where telecommunications carriers and tenants can establish physical network connections. Meet-me rooms serve as the central point of interconnection, housing patch panels, fiber cross-connects, and network demarcation equipment.
Megawatt (MW)
The standard unit of measurement for data centre capacity, referring to the amount of electrical power available for IT equipment. A single megawatt can typically support between 200-400 standard server racks, depending on power density and cooling configuration.
Network-Neutral
A data centre that does not favour any particular network provider, allowing tenants to choose from multiple carriers and internet service providers. Also known as carrier-neutral. Network-neutral facilities typically offer meet-me rooms where carriers can establish points of presence and customers can cross-connect to any available network.
Power Density
The amount of electrical power consumed per unit of floor space, typically measured in kilowatts per rack (kW/rack). Traditional enterprise deployments average 5-8 kW per rack, while modern AI training clusters can exceed 100 kW per rack, requiring specialised cooling and power distribution.
Power Usage Effectiveness (PUE)
A metric that measures data centre energy efficiency by dividing total facility power by IT equipment power. A PUE of 1.0 would mean all energy goes to computing, while a PUE of 2.0 means the facility uses as much energy for cooling and overhead as for IT. Modern facilities target PUE ratings of 1.2-1.4.
Raised Floor
An elevated floor system that creates an underfloor plenum for distributing cooled air to server racks and routing power and data cables. While traditional in data centre design, raised floors are increasingly being replaced by overhead cooling distribution in modern hyperscale facilities.
Redundancy (N+1, 2N)
The level of backup infrastructure built into a data centre's critical systems. N+1 means one additional component beyond the minimum required (e.g., 3 generators when 2 are needed). 2N means a fully duplicated system, providing complete redundancy with no single point of failure.
Retail Colocation
Colocation services sold in small increments, typically individual racks or cages, with shared power and cooling infrastructure. Retail colocation serves small to mid-size enterprises and typically commands higher per-kW pricing than wholesale, with greater flexibility in contract terms.
Sale-Leaseback
A transaction in which a company sells its data centre facility to an investor and simultaneously enters into a long-term lease to continue operating from the same location. Sale-leasebacks allow organisations to unlock capital from owned real estate while maintaining operational continuity.
Stranded Capacity
Data centre capacity that cannot be utilised due to constraints in power delivery, cooling infrastructure, or network connectivity - even though physical space is available. Stranded capacity is an increasing concern in markets with grid connection delays, where facilities may be built but unable to energise for months or years.
Submarine Cable
Fibre optic cables laid on the ocean floor connecting continents and regions, carrying over 95% of international data traffic. Proximity to submarine cable landing stations is a key factor in data centre site selection, particularly for content delivery, cloud services, and international connectivity. Major hubs include Marseille, Singapore, and Virginia Beach.
UPS (Uninterruptible Power Supply)
A power protection system that provides emergency electrical power when the main utility supply fails. UPS systems use batteries, flywheels, or fuel cells to bridge the gap between a power outage and backup generator activation, typically providing 10-30 minutes of runtime.
Uptime Tiers (I-IV)
A classification system developed by the Uptime Institute that rates data centre infrastructure reliability. Tier I provides basic capacity (99.671% uptime), Tier II adds redundant components (99.741%), Tier III allows concurrent maintenance (99.982%), and Tier IV is fault tolerant (99.995% uptime).
Water Usage Effectiveness (WUE)
A sustainability metric that measures how efficiently a data centre uses water, calculated by dividing annual water usage by IT equipment energy consumption. Lower WUE values indicate more efficient water use. Water-free cooling designs are increasingly sought after in water-scarce regions.
White Space
The usable floor area within a data centre where IT equipment (servers, storage, networking) can be deployed. White space excludes support areas such as mechanical rooms, electrical rooms, loading docks, and offices. It is a key metric for determining a facility's deployable capacity and is typically measured in square metres or square feet.
Wholesale Colocation
Colocation services sold in large increments, typically entire data halls or buildings, with dedicated power and cooling infrastructure. Wholesale colocation serves hyperscale cloud providers and large enterprises, with lower per-kW pricing but longer lease terms and larger minimum commitments.
Zero Carbon
A data centre that produces no net carbon emissions from its operations, achieved through a combination of energy efficiency, 100% renewable energy procurement (either directly generated or via power purchase agreements), and carbon offsetting for any residual emissions. Several major operators have committed to achieving zero-carbon operations by 2030.