Systems · Lesson · Intermediate · about 18 minutes
Data Center Electrical Infrastructure
A data center is an electrical system first. Power arrives from the utility, passes through a transformer and switchgear, is conditioned by a UPS and distributed to rows of racks, and a generator stands by for the day the utility fails. Walk the chain below and watch it ride through a loss.
You will learn:
- What each piece of a data center power chain does
- How a UPS, a generator and a transfer switch share a utility loss
- Working out battery runtime from energy and load
- What N, N+1 and 2N redundancy mean for a single failure
- Where the work is and who is qualified to do it
What this model assumes
- One power path (A) is modeled in detail; on a 2N design the second path (B) is healthy and identical, and the racks draw from either cord.
- The UPS switches to its battery without a break, and its static bypass closes within a few milliseconds when the UPS fails.
- Battery runtime is energy divided by load (kWh ÷ kW × 60 minutes) with no derating for age, temperature or inverter losses. Real batteries are sized with margin and tested.
- The generator reaches voltage at a fixed start time, the transfer switch waits a fixed delay, and the UPS waits a fixed time on a restored bus before it returns to line. Real settings are coordinated for each site.
The power chain, from the street to the server
A data center exists to keep thousands of servers running without a break, and everything about its electrical design follows from that one job. The servers are the critical load. Measured in kilowatts for a small room and in megawatts for a large hall, that load has to be fed through a chain of equipment that can lose a piece and keep going.
- Utility service. Power arrives at medium voltage, often 12.47 kV or higher, through one or more feeds.
- Transformer. Steps the medium voltage down to the building's distribution voltage, usually 480 V three-phase in North America.
- Main switchgear. Heavy breakers and bus that split the power into the feeders for the building and protect each one.
- Generator and automatic transfer switch (ATS). A standby engine-generator starts when the utility fails, and the transfer switch moves the switchgear onto it.
- UPS (uninterruptible power supply). Converts the incoming power, filters it and backs it with a battery so the racks never see the gap between a utility loss and a running generator.
- PDU (power distribution unit). Steps the UPS output down to rack voltage, often 208Y/120 V or 415Y/240 V, and splits it into branch circuits.
- Remote power panels or busway. Carry those branch circuits down the rows to each rack.
- Racks and power cords. Each rack's own power strips feed the servers. Critical racks have two cords, one to each power path.
Everything upstream of the UPS is ordinary heavy commercial and industrial work: medium-voltage cable, transformers, switchgear and feeders. Everything downstream is about distribution density, with hundreds of branch circuits in a room that can never go dark. The UPS and the generator are the two pieces that make the room special, and most of this lesson is about how they share the job.
Walk the one-line
Engineers draw a power system as a one-line diagram: one line stands for all three phases, and each piece of equipment is a labeled box. The one below shows a single power path from the utility to a rack. Pull it apart to see each piece on its own, then tap any piece, or its name in the list, to learn what it does and where electricians meet it.
Interactive: Inside a data center power path
What happens when the utility fails
The utility drops. The switchgear bus is dead in the same instant, and so is every motor and light fed from it. The racks are not, because the UPS was already converting power the whole time. Its inverter keeps producing output and simply draws from the battery instead of the bus, with no gap the servers can detect.
The same loss of voltage tells the generator controller to start the engine. A standby set takes about ten seconds to crank, reach speed and build voltage. When the transfer switch sees good voltage from the generator, it waits a short settling delay and transfers the switchgear bus to it. The bus is back, the cooling restarts, and the UPS sees a stable supply again. After its own delay it returns to line, stops discharging and begins to recharge its battery, a process that takes hours.
How long does the battery last?
Given: 500 kWh of battery; 1,000 kW of critical load
- Runtime = 500 ÷ 1,000 hours = 0.5 hours
- 0.5 × 60
Answer: 30 minutes
Thirty minutes sounds like a long time against a ten-second generator start, and that is the point. The battery is sized for the day the generator does not start, so that operators have time to try again, bring in a rental or shut systems down in order. Many sites carry far less, five to fifteen minutes, and accept that the generator has to work. The sim lets you compare.
A UPS can also fail, and for that it carries a static bypass: a solid-state switch that connects its input straight to its output within milliseconds. The racks stay powered, but nothing conditions the supply now, and the next utility dip reaches them directly. Running on bypass is an alarm condition, and the fix is on the clock.
N, N+1 and 2N
Designers describe how much spare capacity a system carries with a short vocabulary. N is exactly what the load needs: one UPS sized for the load, one path, one of everything. It rides through a utility loss, because that is what the UPS and the generator are for, but any failure inside the path itself reaches the racks.
N+1 adds one spare unit to the pieces most likely to fail or need maintenance. A UPS built as N+1 has one more module than the load needs, so one module can fail or be taken out for service while the others carry the full load, protected. The single path stays single, though: a PDU or a switchgear breaker is still one failure away from the racks.
2N builds two complete, independent paths, A and B, each able to carry the entire load, and gives every critical rack two power cords, one to each path. A failure anywhere in path A, from the utility feed to the PDU, leaves the racks on path B without a flicker. Each path can be shut down for work while the other runs the room, which is what concurrent maintenance means in practice.
| Component that fails | N | N+1 | 2N |
|---|---|---|---|
| Utility feed | No, generator and UPS | No, generator and UPS | No, generator and UPS |
| Main switchgear | Yes, after the battery | Yes, after the battery | No, path B |
| UPS | No, on bypass, unprotected | No, the spare module | No, path B |
| PDU | Yes | Yes | No, path B |
Read the table down the UPS row and across the PDU row and the logic shows itself. N+1 protects against the failure of the thing it duplicates and nothing else. Only a second path protects against a failure in the distribution. That is why 2N costs roughly twice as much copper, switchgear and floor space, and why large operators pay it.
Run a utility loss
The simulator steps through each scenario one event at a time. Pick a scenario and a redundancy level, then play the sequence or step through it. Each box turns on or off as the event reaches it, the arrows show which source is feeding the racks, and the readout shows how much battery is left. Change the load and the battery to see how runtime moves.
Interactive: Data center power chain, utility loss and recovery
Two things to try. Run the generator failure at 1,000 kW with the 500 kWh battery and watch the runtime count down to the outage, then cut the load in half and see the runtime double. Then set 2N, fail the UPS, and notice that path B carries the racks while path A sits on bypass.
Trace the power
Three challenges on the same one-line. Each one checks your answer at once against the same model the simulator runs.
Challenge: Find the equipment
The lesson names a piece of the power chain, such as the automatic transfer switch or the static bypass, and you find it on the one-line or pick it from the list.
Challenge: Who is feeding the racks?
Four moments in different scenarios: five seconds into a utility loss, a minute in, a UPS failure on an N design, forty minutes into a generator failure. Say which source feeds the racks and through what.
Challenge: Work out the runtime
Three battery and load pairs. Work out the minutes of runtime from energy divided by load, then check your figure.
Where the work is
Data centers are among the largest electrical construction projects being built, and they keep electricians busy long after the ribbon is cut. Building one means medium-voltage terminations, switchgear and generator installation, thousands of feet of busway, PDU and remote power panel installation, and branch circuits to every rack position. Operating one means testing the generators and transfer switches on a schedule, maintaining batteries and UPS modules, and adding and moving circuits as the floor fills in.
The skills transfer in both directions. An industrial electrician who knows switchgear, motor control and medium voltage is most of the way to data center work. A commercial electrician who has run feeders and panels in a high-rise knows the distribution side. Low-voltage trades handle the structured cabling and the monitoring systems alongside them.
If this is the work you want, the openings below are the industrial and commercial roles where it lives, and the superintendent guide covers how large projects like these are run.
Check your understanding
1. The utility fails. For the ten seconds before the generator is at voltage, what feeds the racks?
Answer: The UPS, from its battery. The UPS inverter keeps running and draws from the battery the instant the bus drops. The generator is still starting.
2. A 500 kWh battery carries a 1,000 kW critical load. About how long can it run?
Answer: 30 minutes. Runtime = 500 ÷ 1,000 hours = 0.5 hours, which is 30 minutes.
3. On an N design the UPS fails and its static bypass closes. What is the state of the racks?
Answer: They stay powered but unprotected until the UPS is repaired. The bypass connects the switchgear bus straight to the PDU. The racks stay up, but nothing conditions the supply and the next utility dip reaches them.
4. A PDU fails. Which design keeps the racks powered?
Answer: 2N. N+1 only duplicates the UPS modules. Only a second complete path, 2N, covers a failure in the distribution, through the racks' second power cord.
5. When does the automatic transfer switch move the bus to the generator?
Answer: After the generator reaches voltage and the transfer delay passes. The switch waits for good generator voltage, then a short settling delay, then transfers. The UPS covers the gap.
Where this knowledge gets used
- Industrial openings: Switchgear, generators, medium voltage and motor control are the industrial side of data center work.
- Commercial openings: Feeders, panels, busway and branch circuits by the thousand are the distribution side.
- Running a large project: What electrical superintendents do on projects of this size and how electricians grow into the role.
Sources and further reading
Uptime Institute
Tier Classification SystemThe tier vocabulary behind N, N+1 and 2N.
Global. Current description.
National Fire Protection Association
NFPA 110, Standard for Emergency and Standby Power SystemsGenerator and transfer switch classification and testing.
United States. 2025 edition current.
National Fire Protection Association
NFPA 70, National Electrical Code (NEC)Articles 645, 700, 701 and 702.
United States; adopted by jurisdiction. 2026 edition issued August 2025; the 2023, 2020 and 2017 editions remain in force in many states; each state adopts its own edition and amendments.
National Fire Protection Association
NFPA 855, Standard for the Installation of Stationary Energy Storage SystemsStationary battery installation.
United States. 2026 edition current.
Lessons are free study material. Finishing one does not certify a skill, count toward a license or replace hands-on training under qualified supervision.