Critical Path Hiring logoCritical PathHiring
Book a hiring diagnostic

Reference

Data center glossary

Every term people hear on a data center job or project, defined in plain English, with what each one means for the work and the people who do it.

New to the industry? Start with the data center guide.

Glossary

Types of data centers

Hyperscale data center
A very large facility, often tens to hundreds of megawatts, built by one company to run its own cloud or AI workloads.
On the job: The biggest builds and the longest projects; commissioning and construction crews follow them.
Hyperscaler
One of the largest cloud operators, such as AWS, Microsoft, Google, Meta and Oracle.
On the job: Their build schedules set demand for trades and Cx talent in every major market.
Colocation (colo)Colo
A data center that rents space, power and cooling to many tenants, who bring their own servers.
On the job: Operations roles here serve many customers at once, so customer-facing skills count.
Wholesale vs. retail colocation
Wholesale leases large blocks, often a full hall or megawatts at a time. Retail leases racks or cages.
On the job: Wholesale sites look like hyperscale builds; retail sites need more customer support staff.
Enterprise data center
A facility a single company owns for its own IT, such as a bank or hospital system.
On the job: Smaller teams and broader roles; one engineer may cover several systems.
Edge data center
A small facility placed close to users to cut delay, often in modular units.
On the job: Field technicians often cover several sites.
Neocloud (GPU cloud)GPU cloud
A newer cloud provider built mainly to rent GPU capacity for AI.
On the job: Fast hiring for GPU cluster, network and facilities roles, often in startup conditions.
Powered shell
A building delivered with power and fiber but no interior fit-out; the tenant builds the rest.
On the job: Fit-out work brings its own wave of MEP and commissioning hiring.
Campus
Several data center buildings on one site sharing power and infrastructure.
On the job: Multi-year work for the same crews as each building comes online.

Glossary

Power

How power reaches a serverNormal path: utility grid, substation and medium-voltage switchgear, transformer, automatic transfer switch, UPS with batteries, PDU or busway, rack. Outage path: a standby generator feeds into the automatic transfer switch.NORMAL PATHUTILITY GRIDSUBSTATION / MVSWITCHGEARTRANSFORMERATSUPSPDU OR BUSWAYRACKSTANDBYGENERATOROUTAGE PATH
Normal power flows from the utility through medium-voltage switchgear and transformers, then through the UPS to PDUs and racks. If utility power fails, the UPS batteries carry the load for seconds to minutes while the generator starts and the automatic transfer switch moves the load onto it.
Critical IT load (MW)MW
The power available to run the computers themselves, measured in megawatts. It's the standard way to size a data center.
On the job: Bigger MW means bigger teams. AI sites are pushing single campuses past 1 gigawatt.
Utility interconnection
The process and equipment that connect a site to the power grid, usually through a new or upgraded substation.
On the job: Waiting for grid power is the top cause of project delays, and delays move hiring dates.
Medium-voltage switchgearMV switchgear
Equipment that receives power from the utility and distributes it around the site at medium voltage.
On the job: Electrical Cx and NETA testing focus heavily here; a top skill for electrical commissioning.
Transformer
Equipment that steps voltage down from utility levels to levels the building can use.
On the job: A common long-lead item; shortages have delayed builds.
UPS (uninterruptible power supply)UPS
Batteries and power electronics that keep IT running for seconds to minutes when utility power drops, until generators start.
On the job: Core equipment for critical facilities engineers; UPS experience is screened for directly.
Standby generator
An on-site engine, usually diesel, that powers the site during a utility outage.
On the job: Generator testing is part of IST; operations teams run regular load tests.
ATS (automatic transfer switch)ATS
A switch that moves the load from utility power to generator power and back automatically.
On the job: Transfer testing is one of the highest-risk parts of commissioning.
PDU / RPPPDU / RPP
Power distribution units and remote power panels that deliver power to rows of racks.
On the job: Day-to-day work for data center technicians.
Busway
An overhead power distribution system that racks plug into, instead of individual cables.
On the job: Common in new builds; installed by electricians, then verified in Cx.
BESS (battery energy storage system)BESS
A large battery installation that stores power, increasingly used alongside or in place of generators.
On the job: A growing skill set that crosses over from utility-scale solar and storage.
Behind-the-meter powerBTM
Power generated on or next to the site, such as gas turbines or fuel cells, instead of drawn from the grid.
On the job: Brings power-plant skills into data center projects.

Glossary

Reliability and redundancy

Uptime Institute TIER IBasic capacity. No redundant parts, one path. Maintenance needs a shutdown.TIER ISRCITBASIC CAPACITYNo redundant parts, one path.Maintenance needs a shutdown.
Uptime Institute TIER IIRedundant capacity. Redundant capacity parts, one path. Some maintenance still needs a shutdown.TIER IISRCITREDUNDANT CAPACITYRedundant capacity parts, onepath. Some maintenance still needsa shutdown.
Uptime Institute TIER IIIConcurrently maintainable. Redundant parts and multiple paths. Any part can be serviced without a shutdown.TIER IIISRCITCONCURRENTLY MAINTAINABLERedundant parts and multiplepaths. Any part can be servicedwithout a shutdown.
Uptime Institute TIER IVFault tolerant. Keeps running through any single unplanned failure.TIER IVSRCITFAULT TOLERANTKeeps running through any singleunplanned failure.
Uptime Institute rates data center designs from Tier I to Tier IV. Each tier adds redundancy. Tier III, the level most job posts ask for, means any part can be maintained without shutting down IT. Tier IV adds tolerance for any single failure.
N, N+1, 2N
N is the equipment needed to carry the load. N+1 adds one spare. 2N duplicates the whole system.
On the job: Shapes how maintenance is done and how many MOPs a team writes.
Concurrent maintainability
Any single component can be taken offline for maintenance without shutting down IT.
On the job: The standard most operations teams work to; it's why MOPs matter.
Fault tolerance
The site keeps running through any single unplanned failure.
On the job: Higher-tier sites with more complex systems need more experienced engineers.
Uptime Institute Tiers (I to IV)
A rating system for data center design from the Uptime Institute: Tier I is basic, Tier II adds redundant parts, Tier III is concurrently maintainable, Tier IV is fault tolerant.
On the job: Most job posts say "Tier III or higher experience"; it's a common screening question.

Glossary

Cooling

Air coolingCooling tower rejects heat outdoors and exchanges with the chiller. Chiller, chilled water loop, CRAH, cold air into the cold aisle, through the rack, hot air out the hot aisle, and back to the CRAH.AIR COOLINGCOOLING TOWERREJECTS HEATOUTDOORSCHILLERCHILLED WATER LOOPCRAHCOLD AISLECOLD AIR INRACKHOT AISLEHOT AIR OUTBACK TO CRAH
Direct-to-chip liquid coolingFacility water loop from the chiller or dry cooler, CDU which separates the building loop from the rack loop, technology cooling loop, cold plates on the GPUs and CPUs inside the rack, and warm liquid back to the CDU.LIQUID COOLING (DIRECT-TO-CHIP)FACILITY WATER LOOPFROM CHILLER ORDRY COOLERCDUSEPARATES THEBUILDING LOOPFROM THE RACKLOOPTECHNOLOGY COOLING LOOPCOLD PLATES ON GPUS /CPUSINSIDE THE RACKWARM LIQUID
Air-cooled halls chill water in a central plant, use CRAH units to cool air, and push that air through the racks. Liquid-cooled AI racks pipe coolant directly to cold plates on the chips, with a CDU separating the building's water loop from the loop that runs into the racks. AI racks drawing 40 kW or more usually need liquid.
CRACCRAC
Computer room air conditioner: a unit that cools air with its own refrigerant compressor.
On the job: Common in older and smaller sites.
CRAHCRAH
Computer room air handler: a unit that cools air using chilled water from a central plant.
On the job: Standard in large sites; maintained by critical facilities teams.
Chiller
Central equipment that produces chilled water for cooling.
On the job: Chiller plant experience is a core mechanical Cx and operations skill.
Cooling tower / evaporative cooling
Equipment that rejects heat outdoors, often by evaporating water.
On the job: Water use is now a permitting issue in dry markets like Phoenix.
Hot aisle / cold aisle containment
Arranging racks so cold supply air and hot exhaust air don't mix.
On the job: Basic knowledge for any technician role.
Direct-to-chip liquid cooling
Liquid piped to cold plates attached directly to the processors. Now standard for high-density AI racks.
On the job: The fastest-growing skill gap in facilities and commissioning.
CDU (coolant distribution unit)CDU
Equipment that separates the building water loop from the liquid loop going to the racks, and controls flow and temperature.
On the job: New equipment many Cx agents haven't tested yet, which makes CDU experience valuable.
Immersion cooling
Servers submerged in a non-conductive fluid.
On the job: Niche but growing; mostly specialist vendors.
Rear-door heat exchangerRDHx
A water-cooled coil mounted on the back of a rack.
On the job: Often used to bring existing halls up to higher densities.

Glossary

Efficiency and density

PUE (power usage effectiveness)PUE
Total facility power divided by IT power, a metric from The Green Grid. 1.0 is perfect. Industry averages run around 1.5; new hyperscale sites target about 1.2 or lower.
On the job: Operations teams are measured on it.
WUE (water usage effectiveness)WUE
Liters of water used per kilowatt-hour of IT energy, also from The Green Grid.
On the job: Increasingly reported and regulated.
Rack density (kW per rack)kW / rack
The power drawn by a single rack. Traditional racks run about 5 to 15 kW; AI racks now run 40 to over 100 kW.
On the job: High density drives the shift to liquid cooling and new skill needs.
White space / gray space
White space holds the IT racks. Gray space holds the power and cooling equipment that supports them.
On the job: Technicians mostly work in white space; critical facilities engineers mostly in gray.

Glossary

IT and AI infrastructure

Rack / cabinet
The standard frame that holds servers, switches and storage.
On the job: Racking and stacking is entry-level technician work.
Structured cabling
The organized copper and fiber network that connects racks and rooms.
On the job: A common entry path into data center careers.
Meet-me roomMMR
The room where network carriers and customers physically interconnect.
On the job: Network technicians and colo operations staff work here.
GPU cluster
Thousands of GPUs networked together to train or run AI models.
On the job: Brings in cluster engineers, network engineers and HPC operations roles.
InfiniBand / RoCEIB / RoCE
High-speed networks used to connect GPUs in AI clusters.
On the job: One of the scarcest AI infrastructure skill sets.

Glossary

Building, commissioning and operations

Design-build
One contract covers both design and construction.
On the job: Common on fast-track data center projects; brings preconstruction roles in early.
General contractor (GC) / EPCGC / EPC
The firm that builds the project. EPC (engineering, procurement and construction) also covers design and buying equipment.
On the job: Employs superintendents, project managers and MEP managers.
Owner's project manager (OPM)OPM
The owner-side manager who oversees the builder for the data center's owner.
On the job: A common next step for experienced GC project managers.
Long-lead equipment
Equipment with order times of a year or more, such as generators, switchgear and transformers.
On the job: Drives the schedule and creates procurement roles.
Prefabrication / modular
Building power or cooling systems as factory skids or modules, then setting them on site.
On the job: Moves some testing into the factory (Level 1 Cx).
MEPMEP
Mechanical, electrical and plumbing systems and the trades that install them.
Full definition in the commissioning glossary
Commissioning (Cx), Levels 1 to 5Cx
Staged testing, from the factory through integrated systems at full load.
Full definition in the commissioning glossary
IST (integrated systems test)IST
The Level 5 test with all systems running together under simulated full load.
Full definition in the commissioning glossary
CxA (commissioning agent)CxA
The independent party that plans, witnesses and documents commissioning.
Full definition in the commissioning glossary
Energization
When permanent utility power is first applied to the site.
Full definition in the commissioning glossary
MOP / SOP / EOPMOP / SOP / EOP
Method of procedure, standard operating procedure and emergency operating procedure: written scripts for work on live systems.
Full definition in the commissioning glossary
CMMSCMMS
Computerized maintenance management system: software that schedules and tracks maintenance work.
On the job: Daily tool for operations teams.
NFPA 70E / arc flashNFPA 70E
The US standard for electrical safety at work, including protection from arc flash.
On the job: Training is required or preferred on almost every electrical role.
NETA acceptance testingNETA
Electrical equipment testing to standards from the InterNational Electrical Testing Association (NETA).
On the job: NETA certification is a strong signal for electrical Cx and testing roles.

See how these pieces fit together in the data center guide.

Newsletter

Get new guides and market updates

No sequences. One issue when there is something worth saying.