Glossary of Terms

Acronym	Names	Description	More info
	Accelerated Compputing	Computing that uses specialized hardware (like GPUs) to perform certain types of calculations more efficiently than general-purpose CPUs.
	AI Ready Data Centre	Facilities specifically designed with the power, cooling, and connectivity infrastructure to support GPU-accelerated AI workloads.
	Aisle Containment	Data centre cooling configuration that separates hot (exhaust) and cold (supply) air to improve cooling efficiency for high-density GPU deployments.	Hexatronic
AHU	Air Handling Unit	An Air Handling Unit (AHU) is used to re-condition and circulate air as part of a heating, ventilating and air-conditioning system. The basic function of the AHU is to take in outside air, re-condition it and supply it as fresh air to a building. This controls the humidity and positive pressure in the data halls to maintain optimal operating conditions for IT equipment.	Airtecnics
ASHRAE	American Society of Heating, Refrigerating and Air-Conditioning Engineers	ASHRAE is a global technical society that establishes standards and guidelines for HVAC systems in various environments including data centres. Their publications define the acceptable environmental ranges for IT equipment operation, addressing temperature, humidity, and air quality parameters. These standards are widely adopted across the industry to ensure optimal equipment performance and energy efficiency.	ASHRAE
ATL	Authority to Let	Formal approval to lease or rent out a property or space. In data centres, this could refer to permission to lease rack space or dedicated areas to clients.
ATS	Automatic Transfer Switch	Device that automatically switches power supply from primary to backup source during an outage. Essential for maintaining uninterrupted power to critical data centre infrastructure.	TechTarget
AV	Audio Visual	Equipment and systems related to sound and visual displays used in conference rooms, security monitoring, and control centres within data facilities.	Data Projections
	Blackwell	NVIDIA's latest GPU architecture succeeding Hopper, designed specifically for next-generation AI and HPC applications, featuring significantly improved performance and efficiency.
BATNA	Best Alternative to Negotiate Agreement	Fallback option if negotiations fail. Helps parties make informed decisions about when to accept or reject proposed terms.	Corporate Finance Institute
BCA	Building Code of Australia	Set of technical provisions for building design and construction. Establishes minimum requirements for safety, health, and sustainability that data centres must meet.	ABLIS
BIM	Building Information Modelling	Digital representation of a facility's physical and functional characteristics. Enables collaborative design and identifies conflicts before construction begins.	Trimble
BMS	Building Management System	Centralised computer system controlling and monitoring building mechanical and electrical equipment. Essential for efficient data centre operations.	CIM
BoQ	Bill of Quantities	Document listing all materials, parts, and labour required for construction. Used for tendering and cost control in data centre projects.	Project Manager
	Container	A standardized software package that includes all dependencies needed to run an application, commonly used to deploy AI and HPC workloads.
	Conversational AI	AI systems designed to engage in human-like dialogue, requiring significant computational resources for real-time inference.
CBS	Cost Breakdown Structure	Hierarchical organisation of project costs by categories. Helps track and manage financial aspects of data centre construction.	InEight
CDU	Coolant Distribution Unit	A device used in liquid cooling systems, primarily in data centers, to distribute and control the flow of coolant (usually water or a specialized fluid) to various components that need cooling. Help manage heat loads, optimize temperature, and improve the overall efficiency and reliability of the cooling system.	BOYD
CFA	Construction Floor Area	CFA = Sum of Unenclosed covered areas (UCA) + Fully enclosed covered areas (FECA). Used for construction planning and calculating material requirements, primarily to estimate construction costs.	All Things Estimating
CMP	Construction Management Plan	Document detailing how construction activities will be managed. Addresses safety, environmental considerations, logistics, and quality control during building.	Procore
COD	Capacity on Demand	Service model allowing clients to quickly scale computing or power resources up or down as needed. Provides flexibility without requiring permanent infrastructure investments.	DEVX
CPO	Contract Price Offer	Formal price proposal provided by a contractor. Establishes the financial baseline for data centre construction contracts.	Linkedin
CRAC	Computer Room Air Conditioning	CRAC units are specialised cooling systems designed for enterprise data halls and network equipment rooms, which are tailored to provide precise control over temperature, humidity, and air filtration 24/7/365. They typically operate with direct expansion cooling technology and are positioned to create efficient hot/cold air separation in server environments.	TechTarget
CRAH	Computer Room Air Handling	Similar to CRAC units but uses chilled water systems rather than refrigerant. Often preferred in larger facilities for higher efficiency and better control.	Sunbird
CSA	Consulting Services Agreement	Legal contract between client and professional consultants. Defines scope, deliverables, payment and terms for design and engineering services.	Contracts Counsel
CSMS	Critical Services Monitoring System	Platform that tracks essential data centre services and alerts operators to potential issues. Usually displayed on dedicated monitoring panels. the health, performance, and availability of critical IT infrastructure and applications, ensuring minimal downtime and optimal functionality.	Linkedin
CUDA	Compute Unified Device Architecture	NVIDIA's parallel computing platform and programming model that enables developers to use NVIDIA GPUs for general-purpose processing.
	CUDA Cores	Processing cores within NVIDIA GPUs that handle parallel computational tasks. More CUDA cores generally mean more processing power.
CUE	Carbon Usage Effectiveness	Metric that measures the carbon emissions associated with data centre operations. Lower values indicate more environmentally friendly operations.	Nlyte Software
CW	Chilled Water (riser)	System of pipes that distribute cooled water throughout a building to remove heat. In data centres, chilled water systems are often used for cooling IT equipment efficiently.	EndoCool
CxL	Commissioning Level	Classification of commissioning depth and rigour. Higher levels involve more systematic and comprehensive testing of data centre systems.	CxPlanner
	Deep Learning	A subset of machine learning that uses neural networks with multiple layers to analyze various factors of data.
DA	Development Application	Formal request submitted to local government for permission to build or modify a property. Required before construction of new data centres or expansions can begin.	BuildCert
DCI	Data Centre Interconnection	Network links that connect multiple data centres, allowing them to share resources and providing redundancy. Critical for disaster recovery and distributed computing architectures.	NextDC Intranet
DCP	Development Control Plan	Local planning instrument that provides detailed guidelines for development. Influences data centre siting, appearance, and operation.	NSW.Gov
DGX	Deep GPU Xceleration	NVIDIA's line of purpose-built AI supercomputers that combine GPUs, networking, and software in an integrated system.
DLC	Direct Liquid Cooling	Cooling solution where liquid coolant directly contacts the GPU components, allowing for higher power densities than traditional air cooling.
DMP	Design Management Plan	Strategy for coordinating design activities across multiple disciplines. Ensures integration between architectural, structural, mechanical, and electrical elements.	Designing Buildings
DOCA	Data Center on a Chip Architecture	NVIDIA's software framework for data centre infrastructure, enabling developers to build applications that leverage NVIDIA's networking, security, and storage capabilities.
DPHI	Department of Planning, Housing and Infrastructure	Government body that oversees planning approvals. Reviews and approves large-scale data centre developments.	DPHI
DPU	Data Processing Unit	A specialized processor that combines networking, security, and storage virtualization capabilities to offload these functions from the CPU.
DRP	Disaster Recovery Plan	A Disaster Recovery Plan is a comprehensive document that details the procedures and resources needed to recover IT systems after a catastrophic event. It includes recovery strategies, emergency response protocols, backup solutions, and step-by-step instructions for restoring critical infrastructure. The DRP typically works alongside business continuity plans to minimize downtime and data loss during major incidents such as natural disasters, cyberattacks, or infrastructure failures.	Cloudian
DTC	Direct to Chip	Direct to Chip cooling is an advanced thermal management approach that places liquid cooling mechanisms directly on or very near to the CPU and other high-heat components. Instead of cooling the entire air volume in a data hall, DTC targets heat generation at its source, allowing for much higher power densities (25kW+ per rack) while improving energy efficiency. This approach is increasingly used in high-performance computing environments where traditional air cooling becomes inadequate or cost-prohibitive.	Data Center Knowledge
	East-West Traffic	Network communication between servers within the same data centre or cluster. In AI Factory and HPC environments, East-West traffic is typically very high volume due to inter-GPU and inter-node communication during distributed training and parallel computing. Optimizing this traffic path is critical for AI workload performance.
	Edge AI	AI processing that occurs near the data source rather than in a centralized data centre, requiring specialized infrastructure considerations.
	Executive Summary	This glossary provides essential terminology used by NVIDIA and its customers in AI Factory and High-Performance Computing (HPC) environments. Understanding these terms will help sales teams effectively engage with technical stakeholders, qualify opportunities, and align data centre infrastructure solutions with customer requirements.
ECI	Early Contractor Involvement	Process of engaging construction contractors during the design phase. Allows their expertise to inform design decisions, improving buildability and reducing delays.	Mastt
EDR/HDR/NDR	Enhanced / High / Next Data Rate	Different generations of InfiniBand technology, offering progressively higher bandwidth for HPC and AI applications.
EGX	Edge AI Acceleration Platform	NVIDIA's platform for edge AI computing, combining hardware and software to enable AI processing at the edge of the network.
EHS	Environmental Health and Safety	Policies and procedures to ensure workplace health and safety. Critical for managing risks during data centre construction and operation.	IBM
EIS	Environmental Impact Statement	Document assessing a project's potential environmental effects. Required for large data centres that may impact surrounding ecosystems.	Investopedia
EOT	Extension of Time	Formal approval for additional project time due to delays. Provides relief to contractors when delays are outside their control.	Designing Buildings
EWR	Early Warning Register	Log of potential issues that might impact a project. Helps project teams proactively address risks before they cause significant problems.	Construo
FECA	Fully Enclosed Covered Area	All interior spaces that are completely enclosed. Constitutes the majority of functional space in data centres, including equipment halls and support areas.	GBCA
FER	Fire Engineering Report	Technical assessment of fire safety systems and strategies. Critical for ensuring data centres meet fire safety regulations.	Origin Fire Consultants
FIE	Free Issue Equipment	Materials or equipment provided by the client to contractors. Common in data centres where specialised equipment is client-specified but contractor-installed.	Alsyed Construction
FIP	Fire Indicator Panel	Central control unit for fire detection and alarm systems. Monitors sensors throughout the data centre and coordinates emergency responses.	Fire Safe
FLOPS	Floating Point Operations Per Second	A measurement of computer performance, particularly in scientific calculations and AI workloads. Often expressed as TFLOPS (trillion) or PFLOPS (quadrillion).
FM	Facility Management	Administration and maintenance of buildings and grounds, including mechanical systems, security, and cleaning. Ensures data centre operates reliably and efficiently.	IBM
FOH	Front of House	Public-facing or client-accessible areas of a facility. In data centres, this includes reception, meeting rooms, and client work areas separate from secure equipment zones.	Torrens
FP	Floating-Point Precision	Floating-point precision formats (FP16: 16-bit, FP32: 32-bit,FP64: 64-bit) used in computational workloads. Lower precision (FP16) enables faster computation but with potential loss of accuracy.
FSR	Floor Space Ratio	Ratio of total floor area to the size of the land parcel. Subject to planning regulations that may limit data centre size relative to its site.	Archistar
FWU	Fan Wall Unit	Fan Wall Units are arrays of multiple smaller fans working in parallel to cool hyperscale data halls. They replace traditional single large CRAC units by creating a more uniform airflow pattern across the entire server space. This modular approach improves energy efficiency, provides N+1 redundancy (if one fan fails, others continue operating), and allows for more precise airflow management and temperature control in high-density computing environments.	Canatec
	Grace	NVIDIA's Arm-based data centre CPU architecture designed for high-performance computing and AI applications.
GB200	Grace Blackwell superchip	NVIDIA's latest integrated CPU-GPU superchip combining Grace CPU and Blackwell GPU architectures, designed for advanced AI and HPC workloads. Successor to the GH200, featuring significant improvements in performance and energy efficiency for large language models and generative AI applications.
GBA	Gross Building Area	Total area of all enclosed spaces, measured from the exterior walls. Used for overall project sizing and zoning compliance.	Room Sketcher
GFA	Gross Floor Area	Total floor area inside the building envelope. Used for planning purposes and calculating building costs and capacity.	Room Sketcher
GH200	Grace Hopper superchip	NVIDIA's integrated CPU-GPU superchip combining Grace CPU and Hopper GPU architectures, designed for giant-scale AI and HPC applications.
GLP	General Light and Power Board	Electrical distribution board for non-critical building services. Supplies power to office areas, corridors, and auxiliary services.	Riello Ups
GMP	Guaranteed Maximum Price	Contract type where the contractor commits to a ceiling price. Provides cost certainty for data centre projects while allowing some flexibility.	NetSuite
GPU	Graphics Processing Unit	A specialized electronic circuit designed to accelerate computer graphics rendering and image processing tasks, excel at performing parallel computations in AI and HPC applications, making them ideal for tasks involving large datasets and complex calculations, such as gaming, video editing, and machine learning.	AWS
GPUaaS	GPU-as-a-Service	Cloud service model providing access to GPU computing resources on a pay-as-you-go basis.
	GPU Density	The number of GPUs that can be installed per rack unit or per rack, a critical metric for data centre capacity planning.
	GPU Direct	NVIDIA technology that enables direct access to GPU memory from third-party devices, reducing latency in multi-GPU systems.
GTC	GPU Technology Conference	NVIDIA's annual conference focused on GPU computing, AI, and related technologies.
	High Density Rack	Server racks specifically designed to accommodate the power and cooling requirements of GPU-intensive workloads, typically supporting 15-30kW per rack or more.
	Hopper	NVIDIA's GPU architecture that powers the H100, designed specifically for AI and HPC applications with improved performance over Ampere.
H100	Hopper architecture	NVIDIA's Hopper architecture GPU designed for AI and HPC workloads, succeeding the A100 with significant performance improvements.
	Hybrid AI Infrastructure	Combination of on-premises GPU computing resources and cloud-based resources to optimize performance, cost, and flexibility.
HGX	Hyperscale Graphics Extension	NVIDIA's GPU platform for AI cloud computing that combines multiple GPUs with high-speed NVLink interconnects.
HPC	High Performance Computing	The use of supercomputers and parallel processing techniques to solve complex computational problems.
HV	High Voltage	Electrical systems operating at higher voltages (typically above 1000V). Data centres require HV infrastructure to distribute power efficiently throughout the facility.	Seaward
	Immersion Cooling	Cooling method where electronic components are submerged in a thermally conductive dielectric liquid, enabling extreme power densities.
	Inference	The process of using a trained AI model to make predictions on new data, typically requiring less computational power than training but often with stricter latency requirements.
IB	InfiniBand	A high-performance, low-latency networking technology commonly used in HPC and AI clusters to connect servers and storage systems.
IDF	Intermediate Distribution Frame	Secondary wiring connection point between the main distribution frame and endpoint devices. Organises and manages network cabling within sections of a building.	FS Community
IPSB	Input Switchboard	Electrical panel that receives incoming power supply. First point of distribution in a data centre's power infrastructure.
IR	Interconnect Room	Specialised area where various network providers can connect their infrastructure. Allows clients to choose from multiple connectivity options.
IRU	Indefeasible Right of Use	Long-term lease agreement (usually 10, 15, or 20 years) where the buyer purchases the right to use specific capacity or fibre pairs on a network	123.Net
IST	Integrated Systems Testing	Comprehensive testing of all interrelated systems working together. Final commissioning stage ensuring all data centre systems operate correctly as an integrated whole.	CxPlanner
ITX	Isolation Transformer	Device that transfers electrical power while isolating circuits. Helps protect sensitive equipment from power disturbances.	Ortea Next
IXP	Internet Exchange Point	A physical location where different Internet Service Providers (ISPs) connect to exchange traffic between their networks	CoreSite
K8s	Kubernetes	An open-source platform for managing containerized workloads and services that facilitates both declarative configuration and automation.
	Language Model Reasoning	The capability of advanced AI models to perform step-by-step logical thinking, often implemented through techniques like chain-of-thought prompting or specialized model architectures that require significant GPU computing resources.
	Liquid Cooling	Advanced cooling technology used for high-density GPU deployments that uses liquid instead of air to dissipate heat, enabling higher power densities.
LAA	Least Acceptable Agreement	Minimum terms a party is willing to accept in a negotiation. Used in contract negotiations for data centre projects and services.	Procuropedia
LLE	Long Lead Equipment	Items requiring extended procurement timeframes. May include generators, chillers, and switchgear that need early ordering for data centre projects.	DCDesign.Africa
LLMs	Large Language Models	AI models with billions of parameters that can understand and generate human-like text, requiring massive computational resources to train and run.
LOD	Level of Detail	Specification for the completeness and accuracy of building drawings. Higher LOD numbers indicate more precise, construction-ready documentation.	Designing Buildings
LTSA	Long-Term Supply Agreement	Contract securing ongoing provision of essential supplies or services. Ensures consistent availability of critical components for data centre operations.	Contracts Counsel
	Mellanox	NVIDIA-acquired company specializing in high-performance networking solutions critical for AI and HPC infrastructure.
	Model Parallelism	A technique for training large AI models by splitting the model across multiple GPUs when it's too large to fit in the memory of a single GPU.
	Multi-modal AI	AI systems that can process and understand multiple types of input data (text, images, audio, video) simultaneously, requiring specialized GPU configurations to handle diverse data formats.
	Multi-modal Reasoning	AI systems that can reason across different types of data (text, images, video, etc.), requiring specialized GPU configurations to process various data types simultaneously while performing complex logical operations.
MCC	Motor Control Centre	A motor control center (MCC) is an assembly to control some or all electric motors in a central location. One of the basic functions of a motor control center is to distribute power, which is done via a horizontal main bus and a vertical bus to each compartment.	Exertherm
MCU	Motor Control Unit	System that manages and protects electric motors. Controls cooling fans, pumps, and other mechanical equipment in data centres.
MCX	Mission Critical Offices	Office spaces designed with high reliability features for tenants who need to maintain operations during emergencies. Often located within data centre facilities.	Alaska Structures
MDB	Main Distribution Board	Central electrical panel that distributes power to subsidiary boards throughout a facility. Critical for power management and safety in data centres.	BMS Controls
MDO	Most Desirable Outcome	Best possible scenario or terms in a negotiation. Represents the ideal result a party hopes to achieve through the negotiation process.	Procuropedia
MEP	Mechanical, Engineering and Plumbing	Design disciplines covering building systems essential to data centre operation. Includes cooling, power distribution, and water management systems.	APA Engineering
MGX	Modular GPU Experience	NVIDIA's modular reference architecture that standardizes system designs to accelerate deployment of AI infrastructure.
MIG	Multi-Instance GPU	NVIDIA technology that allows a single GPU to be partitioned into multiple instances, each with dedicated compute and memory resources.# NVIDIA AI Factory & HPC Terminology Glossary
MLOps	Machine Learning Operations	Practices that aim to deploy and maintain machine learning models in production reliably and efficiently.
MLOPs Platform	Machine Learning Platform	End-to-end platform for managing the machine learning lifecycle, from development to deployment and monitoring.
MMR	Meet Me Room	Central area where telecommunications carriers and clients establish physical connections. Similar to interconnect rooms, they are essential for network diversity.	Sunbird
MOP	Method of Procedure	Detailed step-by-step instructions for completing technical tasks. Ensures consistent and safe execution of operations in data centres.	Roadmunk
MOQ	Minimum Order Quantity	Smallest amount of a product a supplier will sell. Relevant when procuring equipment and materials for data centre construction or upgrades.	ShipBob
MPC	Memo/Method of Proposed Change	Document detailing suggested modifications to plans or contracts. Formalises the change management process during construction.
MPX	Multiplex	Construction company name, often abbreviated in project documents. Frequently involved in large-scale data centre construction projects.	Multiplex
MSA	Multi-Solution Approach	Strategy of implementing various technologies or vendors to solve complex problems. Helps avoid single points of failure in data centre design.	Linkedin
MSB	Main Switch Board	Primary electrical distribution panel that receives power from the utility and distributes it throughout the facility. Critical infrastructure for data centre operations.	Zit21
	North-South Traffic	Network communication between the data centre and external networks (including the internet). In AI Factory environments, this includes data ingestion, model serving, and user interactions with deployed AI systems. While typically lower volume than East-West traffic in AI training clusters, North-South capacity is critical for inference workloads and data preparation pipelines.
	NVIDIA AI Enterprise	NVIDIA's end-to-end, cloud-native suite of AI software, certified to deploy anywhere.
	NVLink	NVIDIA's high-bandwidth, energy-efficient interconnect that allows multiple GPUs to communicate directly with each other.
	NVSwitch	NVIDIA's networking switch that enables multiple GPUs to communicate at high bandwidth within a server.
NaaS	Network as a Service	A model where network infrastructure and services are provided to organizations as a subscription-based service, rather than requiring them to purchase and manage their own hardware and software	Fortinet
NCC	National Construction Code	Comprehensive set of technical building requirements for Australia. Includes the BCA and plumbing regulations that impact data centre design.	NCC
NeMo	Neural Modules	NVIDIA's framework for building, training, and fine-tuning GPU-accelerated language models, including those with advanced reasoning capabilities.	Scaleway
NeMo Guardrails	Neural Modules Guardrails	NVIDIA's framework for implementing guardrails in language models to ensure responsible AI outcomes, particularly important for reasoning-focused applications.
NER	National Engineering Register	Official register of qualified engineers in Australia. Relevant for verifying credentials of engineers involved in data centre design.	Engineer Australia
NLA	Net Lettable Area	Floor space that can be leased to tenants. In data centres, this might include white space (IT areas) and office areas available for client use.	Go Legal Ai
NTE	Not to Exceed	Similar to GMP, a budget limit that cannot be surpassed. Establishes financial boundaries for project components or overall costs.	Field Complete
NTT	Nippon Telegraph and Telephone	Japanese telecommunications company that operates data centres globally. Often a tenant or partner in data centre facilities.	NTT
NVME	Non-Volatile Memory Express	A specification for accessing solid-state storage media attached via the PCIe bus, commonly used in high-performance AI storage systems.
	Omniverse	NVIDIA's platform for connecting 3D worlds in a shared virtual universe, used for simulation and collaboration.
	Optical Networking	High-speed networking technology using light for data transmission, critical for connecting GPU clusters in large AI factories.
OAP/OU	Outside Air Pressure / Outside Unit Pressure	System that manages air pressure, temperature, and airflow in server aisles. Helps prevent hot spots and ensures efficient cooling throughout the data hall.
OPSB	Output Switchboard	Electrical panel that distributes power from UPS systems to critical loads. Ensures clean, reliable power reaches IT equipment.
ORBD	Over-Rack Busduct	Power distribution system installed above server racks. Provides flexible power connectivity for changing IT equipment layouts.	Sunbird
ORTOB	Over-Rack Tap-Off Box	Connection point from the overhead busduct to server racks. Allows power to be safely distributed to equipment.
PCIe	Peripheral Component Interconnect Express	A high-speed serial computer expansion bus standard used to connect GPUs to the rest of the computer system.
PCSA	Pre-Construction Services Agreement	Contract that defines services provided before main construction begins. Includes design development, cost estimation, and planning activities.	Archor
PDS	Product Data Sheet	Technical document that details the specifications and performance characteristics of equipment. Used for selecting and comparing data centre infrastructure components.	Bluestone PIM
PDU	Power Distribution Unit	PDUs are specialised power strips designed for data centre environments that distribute electrical power to multiple servers and networking devices within a rack. Modern PDUs include advanced features such as remote power monitoring, outlet-level power cycling, environmental monitoring, and power usage metrics. They serve as the final link in the power chain from the main facility distribution to the IT equipment, often with redundant power paths for critical infrastructure.	Sunbird
PFLOPS	Petaflops	A measure of computer performance equal to 10^15 floating-point operations per second.	Merriam Webster
PID	Project Initiation Document	Formal document authorising the start of a project. Defines key parameters, stakeholders, and objectives for data centre development.	Indeed
PM	Project Manager	Professional responsible for planning, executing, and closing projects. Coordinates all aspects of data centre development from concept to completion.	CIO
PMP	Project Management Plan	Document outlining how a project will be executed, monitored, and controlled. Includes schedules, budgets, communication plans, and risk management strategies.	Simpli Learn
PO	Purchase Order	Commercial document authorising purchase of goods or services. Formalises procurement processes for data centre equipment and construction materials.	Square
POD	Performance Optimized Data Centre	Modular data centre units that can be rapidly deployed to scale infrastructure for AI and HPC workloads.
POE	Post Occupancy Evaluation	Assessment of a building's performance after it has been occupied. Identifies lessons learned and potential improvements for future data centre designs.	RIBA
PPR	Principal Project Requirements	Document outlining the client's key requirements and specifications. Serves as the foundation for data centre design decisions.	PWC
PUE	Power Usage Effectiveness	PUE is a ratio that measures how efficiently a data centre uses energy by dividing total facility energy consumption by the energy consumed solely by IT equipment. A perfect PUE would be 1.0, meaning all energy powers computing equipment with no overhead. Modern efficient data centres typically achieve PUE values between 1.1 and 1.4, while older facilities may have PUE values of 2.0 or higher. This metric is the industry standard for assessing energy efficiency.	Digital Realty
	Quantum Computing	An emerging computing paradigm that uses quantum-mechanical phenomena to perform operations on data, sometimes complemented by GPU-accelerated simulations.	IBM
QS	Quantity Surveyor	Professional who estimates and monitors construction costs. Helps manage budgets and evaluates financial impacts of design changes in data centre projects.	RICS
RCI	Rack Cooling Index	Metric that evaluates how effectively IT equipment is being cooled within racks. Helps identify cooling issues before they cause equipment failures.	RF Code
RDHx	Rear Door Heat Exchanger	A cooling solution attached to the rear of a server rack that uses liquid to remove heat from the air exiting the rack.
RDMA	Remote Direct Memory Access	GPU-Direct RDMA is a technology enabling direct access to GPU memory from third-party devices using Remote Direct Memory Access, reducing latency in multi-GPU systems.	Weka
RFI	Request for Information	Formal process of gathering information from potential vendors or contractors. Used during planning stages of data centre projects to understand available solutions.
RFP	Request for Proposal	Document that solicits proposals from potential suppliers or contractors. Details project requirements and evaluation criteria for selecting vendors.
RFS	Ready For Service	Milestone indicating a data hall is fully equipped and ready for client connections. Marks the transition from construction to operational status.
RFX	Request for X	Generic term for various formal requests in procurement (X could be Proposal, Quotation, Information). Used throughout data centre development and operation.
RIBA	Royal Institute of British Architects	Professional organisation that establishes design process standards. Their project methodology influences many international construction practices.
RNI	Ready for Network Installation	Milestone indicating a data hall is prepared for installation of network infrastructure. Precedes the RFS stage in the commissioning process.
RPO	Recovery Point Objective	RPO defines the maximum tolerable amount of data that can be lost during a system failure, measured as time between the last valid backup and the point of failure. For example, an RPO of 4 hours means systems must be configured to ensure no more than 4 hours of data can be lost in any scenario. This metric drives backup frequency decisions and is typically defined by business requirements around the criticality of the data and cost of potential data loss.
RTO	Recovery Time Objective	RTO specifies the maximum acceptable time it takes to restore a system after a failure or disaster event. It represents how long the business can tolerate being without this system before experiencing serious adverse consequences. For critical systems, RTOs might be measured in minutes; for less essential systems, RTOs might be measured in hours or days. This metric drives investment decisions in redundancy, backup systems, and recovery processes.
RTS	Response to Submission	Document addressing questions or concerns raised by government authorities after submitting development applications. Critical for moving projects through approval processes.
RTX	Ray-tracing Texel eXtreme	NVIDIA's advanced real-time ray tracing technology, primarily used in graphics applications but increasingly finding use in scientific visualization for HPC.
RU	Rack Unit	A standardized unit of measurement used to define the vertical space in a server rack.	Rack Solutions
RUPS	Rotary Uninterruptible Power Supply (Unit)	UPS system that uses a motor-generator to provide reliable power during outages. Preferred in large data centres for higher reliability and capacity.
	Supercomputer	A computer with performance capabilities far above a general-purpose computer, typically measured in PFLOPS for modern systems.
SCADA	Supervisory Control and Data Acquisition	System for monitoring and controlling industrial processes and infrastructure. Used in data centres to manage mechanical and electrical systems in real-time.
SCEC	Security Construction and Equipment Committee	Government body that establishes security standards. Relevant for data centres handling classified or sensitive information.
SDN	Software Defined Networking	A network management approach that uses software to control and manage network devices, enabling dynamic and programmatically efficient network configuration.
SLA	Service Level Agreement	Contract between service provider and client that defines expected service quality, availability, and performance metrics. Establishes accountability and remedies for service failures.
SLI	Scalable Link Interface	NVIDIA technology that allows multiple GPUs to work together to render graphics, a concept now expanded to AI computing.
SME	Subject Matter Expert	Individual with specialised knowledge in a specific field. Provides technical guidance on complex aspects of data centre design or operation.
SOC	Security Operation Centre	Facility where security experts monitor and defend against cybersecurity threats. Often integrated with data centres to protect physical and digital assets.
SOQ	Statement of Qualification	Document detailing a company's capabilities, experience, and credentials. Used during vendor selection processes for data centre projects.
SPD	Surge Protective Device	Equipment that shields electrical systems from voltage spikes. Protects sensitive data centre equipment from damage during electrical events.
SSDA	State Significant Development Application	Development application submitted directly to state government rather than local council. Required for large-scale data centre projects deemed important to state interests.
	Tensor Core	Specialized cores within NVIDIA GPUs designed specifically to accelerate deep learning operations.
	Transformer	A type of neural network architecture that has become dominant in natural language processing and increasingly in computer vision, known for requiring substantial computational resources.
TCO	Total Cost of Ownership	Comprehensive assessment of the direct and indirect costs of owning and operating GPU infrastructure over its lifetime.
TFLOPS	Teraflops	Represents one trillion floating-point operations per second. It is a measure of a computer's ability to perform complex mathematical calculations involving decimal numbers with high precision.	Lenovo
TOB	Tap Off Box	Connection point to the over-rack busduct system. Allows power to be safely distributed to server racks from overhead busways.
TTS	Time-to-Solution	The total time required to solve a scientific or business problem using HPC or AI infrastructure, a key performance metric.
TZ Rack	TZ is the brand for rack lock	Secure locking system for server cabinets. Provides physical security and access control for data centre equipment.
	Unified Memory	NVIDIA CUDA feature that creates a single memory address space accessible from any processor in the system, simplifying memory management in GPU programming.
UAT₁	Universal Accessible Toilets	Bathroom facilities designed for use by people with disabilities. Required by building codes and accessibility standards in commercial facilities.
UAT₂	User Acceptance Testing	The final stage of software development where real users test the software to ensure it meets their needs and functions as expected in real-world scenarios
UCA	Unenclosed Covered Area	Areas that are roofed but not fully enclosed. May include loading docks, equipment yards, or covered walkways in a data centre complex.	GBCA
UPS	Uninterruptible Power Supply	UPS systems provide instantaneous backup power when the main power supply fails, primarily using rotary UPS technology in enterprise data centres. Unlike battery-only systems, rotary UPS combines a motor-generator with flywheels or batteries to deliver consistent, clean power during outages. These systems bridge the gap between utility failure and generator startup, ensuring zero interruption to critical IT loads while also conditioning power to filter out surges and other electrical anomalies.
UTI	Uptime Institute	Organisation that defines and certifies data centre reliability tiers. Their standards are widely used to classify facility redundancy and resilience.
VGPU	Virtual GPU	NVIDIA technology that enables the sharing of physical GPUs across multiple virtual machines.
VPA	Voluntary Planning Agreement	Contract between developer and planning authority for infrastructure contributions. May facilitate faster approvals for data centre projects.
VPN	Virtual Private Network	A secure, encrypted connection over a public network, like the internet, that allows users to access a private network as if they were directly connected.
WBS	Work Breakdown Structure	Hierarchical decomposition of project work into manageable components. Organises the total project scope for planning and execution.
WUE	Water Usage Effectiveness	Metric that measures water consumption in data centre operations. Important for sustainability assessments, especially in water-stressed regions.
ZOPA	Zone of Potential Agreement	Range where parties might reach agreement. Area where the buyer's maximum price overlaps with the seller's minimum acceptable price.