The Data Center Glossary: 30 Terms You Need to Know Before Your First Interview

June 26, 2026

If you're coming from outside the industry, data center interviews have a way of feeling like a foreign language test. Hiring managers don't always explain acronyms. Engineers drop terms casually and watch how you react. Candidates who know the vocabulary read as insiders; candidates who don't read as beginners — even when their underlying skills are exactly what the role needs.

This glossary won't make you a data center engineer overnight. What it will do is make sure you don't lose points on terminology when your actual experience is strong. Learn these 30 terms before you walk in the door.

Power

Kilowatt (kW) / Megawatt (MW): The basic units of power capacity in data centers. A single server rack might draw 10–30 kW. A hyperscale campus might be measured in hundreds of MW. When someone asks about the size of a facility, they often mean its power capacity, not its square footage.

Power Usage Effectiveness (PUE): The ratio of total facility power to the power used by the IT equipment itself. A PUE of 1.0 would be perfect — all power goes to compute. Most modern facilities run between 1.2 and 1.5; hyperscalers push closer to 1.1. Lower is better. You'll hear this number cited constantly as an efficiency benchmark.

UPS (Uninterruptible Power Supply): Battery-backed systems that bridge the gap between a utility outage and generator startup — typically 10 to 30 seconds. Critical for maintaining uptime during power transitions. Knowing the basics of UPS topology (double-conversion, line-interactive) is useful for any facilities or electrical role.

Generator / Genset: Diesel or natural gas generators that provide backup power during extended outages. Understanding runtime capacity (fuel tank size relative to load) and transfer switch logic is expected knowledge for critical facilities roles.

PDU (Power Distribution Unit): The rack-level device that distributes power from the facility's electrical system to individual servers. Comes in basic strip and intelligent (monitored/metered) variants. You'll hear "PDU" constantly in operations roles.

Busway / Busbar: An overhead or underfloor power distribution system that delivers electricity from the UPS to PDUs without traditional cable runs. Common in large deployments for flexibility and density.

N+1 / 2N Redundancy: How resilient a power system is. N+1 means one extra unit beyond what's needed to run at capacity — if one fails, the system keeps running. 2N means full redundancy: two complete, independent systems. You'll see these shorthand everywhere in specs and job descriptions.

Cooling

CRAC / CRAH (Computer Room Air Conditioning / Handler): The primary cooling units on a data center floor. CRACs use a refrigerant-based compressor (self-contained); CRAHs use chilled water from a central plant. Knowing the difference signals real familiarity with mechanical systems.

Hot Aisle / Cold Aisle Containment: The standard rack layout strategy where server intakes face cold aisles and exhausts face hot aisles, preventing hot and cold air from mixing. Containment systems (physical curtains or ceilings) improve this further. Fundamental to any conversation about cooling efficiency.

Chiller: Large mechanical systems that produce chilled water used to cool the facility. Often located outside the building. A central part of the cooling plant for most medium and large facilities.

Cooling Tower: Works in conjunction with chillers to reject heat to the atmosphere. If you've worked in commercial HVAC or industrial plants, you know these. In data centers, they're scaled up significantly.

PUE (again): Worth noting here too — PUE is as much a cooling metric as a power metric. The biggest lever in improving PUE is almost always cooling efficiency.

Liquid Cooling / Direct Liquid Cooling (DLC): Increasingly common for high-density AI and GPU workloads, where air cooling can't move enough heat. Includes cold plate systems (liquid directly on chips) and immersion cooling (servers submerged in dielectric fluid). Expect this to come up more and more in 2026 interviews.

Infrastructure & Operations

Tier Classification (I–IV): The Uptime Institute's framework for data center reliability, from Tier I (basic, ~99.67% uptime) to Tier IV (fault-tolerant, ~99.995% uptime). Most enterprise and colocation facilities aim for Tier III or IV. Knowing the definitions signals you understand the reliability business.

Colocation (Colo): A facility that leases space, power, and connectivity to multiple tenants who bring their own servers. Contrasts with hyperscale (built and operated by a single company for its own use). Many operator roles are at colo providers like Equinix, Digital Realty, or Iron Mountain.

Hyperscale: Data centers built and operated by the major cloud providers (AWS, Azure, Google, Microsoft) at massive scale — typically 100 MW+. A distinct hiring ecosystem from colo; different culture, different ops model.

Edge Data Center: Smaller, distributed facilities located closer to end users to reduce latency. Growing fast with AI inference and IoT demand. An increasingly common term in site selection and construction conversations.

DCIM (Data Center Infrastructure Management): Software platforms that monitor and manage power, cooling, space, and assets across the facility. Think of it as the operating system for the physical infrastructure. Vendors include Schneider Electric (EcoStruxure), Vertiv, and Nlyte. Familiarity with any DCIM platform is a resume plus.

BMS / BAS (Building Management System / Building Automation System): The broader facility control system that integrates HVAC, power, fire suppression, access control, and more. DCIM is often layered on top of or alongside a BMS. Electricians, HVAC techs, and controls engineers will know BMS from other industries — translate that directly.

COLO vs. Hyperscale vs. Edge These three describe the type of facility, not the role. Worth being fluent in the distinctions because the culture and operational demands differ significantly across each.

Networking

Cross-Connect: A physical cable connection between two parties within a colocation facility — for example, a tenant connecting to a network carrier. A basic revenue unit for colo providers and a term that comes up constantly in sales and operations conversations.

Meet-Me Room (MMR): The area within a colo where carriers and networks interconnect. Knowing this term signals familiarity with how connectivity works inside a facility.

Dark Fiber: Fiber optic cable that's been laid but not yet lit with active networking equipment. Data centers and carriers buy and lease dark fiber to build private network connections. Comes up in network and real estate conversations.

Latency: The delay in data transmission. Low-latency requirements drive edge facility placement and campus interconnect strategy. If you're in site selection or development, understanding how latency maps to geography is increasingly important.

Real Estate & Development

Critical Load: The actual IT power draw of a facility, as opposed to total building power. When a developer or broker says a campus is "200 MW," they often mean 200 MW of critical load. Getting this distinction right matters in real estate conversations.

Time-to-Power: How long it takes to get a new facility — or new capacity within a facility — energized and ready for tenants. One of the most important competitive factors in the development market right now, often cited in months.

Wholesale vs. Retail Colo: Wholesale means leasing large blocks of capacity (typically 1 MW+) to a single tenant. Retail means leasing individual cages or cabinets to smaller customers. Different business models, different sales cycles, different operations profiles.

Power Purchase Agreement (PPA): A contract for buying electricity — often renewable — directly from a generator rather than the utility. Increasingly important for large operators with sustainability commitments. Comes up in finance, development, and ESG conversations.

Safety & Compliance

NFPA 70E: The OSHA-aligned standard for electrical safety in the workplace — arc flash protection, lockout-tagout, PPE requirements. Expected knowledge for any electrical or facilities technician role. If you're coming from electrical work in other industries, you likely have this already.

Lockout/Tagout (LOTO): The procedure for isolating and de-energizing equipment before maintenance to prevent accidental startup. A fundamental safety protocol in data center operations — zero tolerance for shortcuts here.

Change Management: The formal process for planning, approving, and executing any change to a live system. In mission-critical environments, undocumented changes are how incidents happen. If you're interviewing for an operations role, expect to be asked how you've handled change management in your past work.

How to Use This in an Interview

Knowing the vocabulary is the floor, not the ceiling. The goal isn't to recite definitions — it's to use these terms naturally when describing your experience. "I've worked on BMS integration" lands better than "I've worked on building automation systems." "We ran N+1 on the generator side" reads as someone who's been in the room.

If a term comes up that you don't know, ask. Interviewers in this industry respect candidates who ask precise, curious questions over those who bluff. "Can you walk me through how your facility handles that?" is a perfectly good response.

The data center industry is hiring aggressively across every function right now. The gap between you and a job offer is often smaller than you think — and it starts with being able to speak the language.

UptimeJobs.io lists data center and AI infrastructure roles across operations, power & cooling, network engineering, AI/ML, and construction, updated regularly across every major U.S. market. Browse current openings to find roles that fit your background.