Working in marketing in the data centre and deep tech space means you’re wading in technical language from day one. This jargon glossary exists to give you enough context of the most technical terms, enabling you to brief content accurately, hold a credible conversation with engineering and stop Googling things under the table during client calls.

Please bear in mind that it’s a working refrence for the non-technical professional who needs to understand what’s bein said wlel enough to communicate it clearly to people are whoa ren’t engineers. It’s updated regulattly as new terminology emerg.es. If there’s a term you’d like to add, drop me a message through the contact form.

The Glossary

A

AI Infrastructure – physical and digital systems that support artificial intelligence workloads. From servers, networking, storage and the data centres that house them.

Availability Zone – a separate location within a data centre region designed to be isolated from failures in other zones. It’s like a backup neighbourhood; one area goes down, the others keep running.

B

Bandwidth – The amount of data that can be transmitted over a network connection in any one given time. More bandwidth = faster data transfer, so in data centre marketing, this term is often used in relation to connectivity and network performance.

Brownfield – An existing site or facility being upgraded rather than being built from scratch. It’s the opposite of Greenfield.

C

Cage – A secure, fenced-off area in a colocation facility where a customer has their own equipment. It still sits in the facility, but it gives more privacy than a standard rack.

CapEx (Capital Expenditure) – Money spent on buying phsyical assets such as building a data centre, hardware, servers, etc. It is not OpEx, which is ongoing spending to keep operational. It’s important to know the difference when pitching to CFOs or writing about investment.

Cloud – Computing resources – storage, processing power, networking – delivered over the internet, not from on-site hardware. Public cloud means shared resources from a provider like Amazon Web Services (AWS). Private cloud means dedicated resources. Hybrid cloud is a mix of both.

Colocation (Colo) – A facility where businesses rent space, power and connectivity for their own servers and equipment, rather than building their own data centre.

Connectivity – This is how data travels in and out of a data centre. It’s the physical and digital links to networks, internet exchanges and other facilities.

CPU (Central Processing Unit) – The primary processor in a computer; think of it like the brain that executes instructions. Different to a GPU, which handles graphics and AI workloads.

D

Data Centre – A facility housing computing infrastructure such as servers, storage, and networking equipment, used to store, process and distribute data. It starts from small server rooms to campuses with hundreds of thousands of square metres.

Data Sovereignty – The principle that data is subject to the laws of the country where it is stored or processed. It’s incredibly important in marketing and PR conversations as regulations tighten globally.

Dedicated Server – A physical server used only for one customer, and not shared with others. This is more expensive than virtual or shared hosting but it has better security and performance.

E

Edge Computing – Processing data closer to where it’s generated, rather than sending all the data to a centralised data centre. It reduces latency and bandwidth use (please see separate definitions!). It’s becoming important as Internet of Things (IoT0 devices, autonomous vehicles and real-time applications multiply.

Energy Intensity – How much energy is consumed relative to the computing work that has been done. If there’s lower energy intensity means operations are more efficient.

F

Fibre – Optical fibre cables transmit data as light rather than electrical signals because it’s more capable and faster than carrying more data over long distances.

Footprint – The physical space a customer takes up in a data centre. It’s measured in rack units, cabinets, or square metres. The term can also be used to describe a company’s total data centre presence across multiple sites.

G

GPU (Graphics Processing Unit) – Renders graphics and is the primary hardware for AI and machine learning workloads because of its ability to process many calculations simultaneously.

Greenfield – A new site built from scratch rather than on an existing developed site. It has a high upfront constraint but less infrastructure constraints.

Grid Connection – The physical connection between a data centre and the electricity grid. It’s increasingly a bottleneck for new data centre development as power demand outstrips grid capacity in many markets.

H

HPC (High Performance Computing) – Extremely powerful computing systems used for complex calculations (weather modelling, scientific research, AI training, etc). Data centres built for HPC have different power and cooling needs compared to standard facilities.

Hybrid Cloud – A computing environment that combines private cloud or on-premises infrastructure with public cloud services.

Hyperscale – Data centres that are built and operated at a huge scale, typically by big cloud providers such as Microsoft Azure, Google Cloud and Amazon Web Services.

Hyperscaler– The companies that build and operate hyperscale facilities – so Meta, Google Cloud, Apple, Amazon Web Services, etc.

I

Interconnection – linking of clouds, networks and IT infrastructure within or between data centres. It’s a key selling point for colocation providers, so the more interconnected a facility is, the more valuable it is to customers.

IP Address – A unique numerical label assigned to every device connected to a network (the internet’s postal address, if you will).

IXP (Internet Exchange Point) – A physical location where different networks connect and exchange traffic. Being close to or housed within an IXP is a significant connectivity advantage for a data centre.

L

Latency – The time delay between sending and receiving data. Having lower latency means a faster response time, which is critical for financial trading, gaming and real-time AI inference. The proximity to end users and major network hubs reduces latency.

Lease – In the data centre industry, it’s the agreement between a customer and colocation, or wholesale provider covering the power, space and connectivity they’re renting.

Load Balancing – Disrupting computing workloads across multiple servers or resources to prevent any single point from being overwhelmed. It keeps systems running smoothly under high demand.

M

Meet-Me Room (MMR) – The area in a data centre where different networks and carriers connect physically.

Modular Data Centre – A data centre that’s built from prefabricated modules rather than traditional construction. It’s scalable, faster to deploy and increasingly popular for edge and remote deployments.

N

N+1 Redundancy – Having one more unit of a critical component than is strictly needed to operate. So if you need four cooling units, N+1 means you have five units, which allows one to fail without affecting operations.

Network Fabric – The underlying infrastructure of switches, routers and connections that form a data centre’s internal network. It’s how data moves around inside the facility.

O

OpEx (Operations Expenditure) – This is the ongoing costs of running a business, including power, staff, connectivity, and maintenance. It’s different from CapEx, which covers physical asset purchases.

Overhead Cabling – Instead of under a raised floor, power and data cables are routed above the server racks. It’s a design choice that allows for more airflow management and maintenance access.

P

PDU (Power Distribution Unit) – A device that distributes electrical power to servers in a rack; it comes up in power density and infrastructure conversations.

Peering – When two networks agree to exchange traffic directly with each other rather than routing through a third party, this reduces costs and latency.

POP (Point of Presence) – A physical location where a network or service provider has equipment enabling connectivity. Having multiple POPs means better coverage and lower latency for end users.

Power Density – This is how much power is consumed per unit of floor space or per rack and it’s measured in kilowatts per rack. AI workloads are driving power density requirements dramatically higher than traditional IT workloads, which is a challenge for older facilities.

PUE (Power Usage Effectiveness) – This is currently a widely used metric for data centre energy efficiency. It’s calculated by dividing total facility power by the power used by IT equipment. A PUE of 1.0 is perfect efficiency because all power goes to computing (though some claim it’s a myth). A PUE of 1.5 means that for every watt used by IT equipment, 0.5 watts is used by cooling, lighting, and other overhead. Lower is better.

R

Rack – A metal frame that houses servers, networking equipment, and other hardware in a data centre. It’s measured in rack units (U), and 1 U is 1.75 inches of vertical space.

Redundancy – Having backup systems in place so if something fails, another takes over without any interruption to ensure it remains operational. It’s measured in tiers, with Tier 1 being basic and Tier IV being fully fault-tolerant.

Resilience – The ability for systems and facilities to continue operating through disruptions, disasters, or failures. Resilience is a much broader concept than just redundancy as it encompasses power systems, disaster recovery and physical security.

S

SLA (Service Level Agreement) – A contract defining level of servies a provider commits to delivering, this could look like uptime guarantees, response times, support commitments, etc. A 99.999% uptime SLA means less than six minutes of downtime per year.

Software Defined Networking (SDN) — A networking approach that separates the control of network traffic from the physical hardware, making networks more flexible and programmable. Increasingly common in modern data centre environments.

Stranded Power – Power capacity that has been contracted or allocated but isn’t being used; it’s actually a significant issue for operators managing large facilities with uneven utilisation.

T

Tier Rating – A classification system developed by Uptime Institute that rates data centres from Tier I to Tier IV based on resilience and redundancy; Tier IV is fully tolerant of faults with no points of failure, and Tier I is the least tolerant.

TOR (Top of Rack) – A networking switch positioned at the top of a server rack that aggregates the connections from all servers in the rack to the wider network.

U

Uptime – The percentage of time a system is operational and running; 99.999% uptime, known as five nines, means less than six minutes of unplanned downtime each year. This is what the industry strives for. Uptime guarantees are a core part of data centre and colocation sales conversations.

UPS (Uninterruptible Power Supply) – These are battery backup systems that provide immediate power if the main supply fails, which bridges the gap until the generators kick in.

V

Virtualisation – Technology that creates virtual versions of physical computing resources, such as servers, storage and networks, allowing multiple virtual machines to run on a single physical server.

VLAN (Virtual Local Area Network) – A logical grouping of network devices that behave as if they’re on the same network regardless of their actual physical location. It’s used to segment traffic and improve security in data centre environments.

W

Wholesale Colocation – Leasing large amounts of space in a data centre, typically an entire hall or more, rather than individual racks. It’s chosen by large enterprises or cloud providers who need significant capacity at scale.

Workload – The computing tasks being processed by a system. Different workloads have different infrastructure requirements: AI training workloads need high power density and GPU capacity, while the more traditional enterprise workloads might prioritise storage and network performance.

This glossary is maintained by Lauren Raybould, Founder of Freshly Brewed Intelligence.

Last updated: July 2026.

Something missing? Get in touch: lauren@freshlybrewedintelligence.com