5 Common Power Cable Mistakes in Data Centers

1. Undersizing the Cable (Ignoring Derating and Continuous Loads)

This is the classic. Someone looks at a simple ampacity table, picks a wire size, and calls it good. In a real data center the load is continuous, ambient temperatures near hot aisles or dense cable trays run higher than 30 °C, and multiple current-carrying conductors share the same tray or conduit.

NEC requires 125 % of continuous load for most feeders and branch circuits. Add temperature correction and bundling derating and that “perfectly sized” cable is suddenly operating near or past its limit. Result: elevated conductor temperatures, insulation degradation, higher voltage drop, and in extreme cases, thermal runaway. High-density AI racks make this especially dangerous because the load stays near peak for long stretches.

Fix: Size for continuous load + derating factors from the start. When in doubt, go up one size. Model the actual thermal environment, not the textbook one.

2. Mixing Power and Data Without Proper Separation

Running power feeders parallel to data cables (or stuffing both into the same tray) is still common, especially in tight spaces or during rushed installs. Power cables radiate electromagnetic interference. In high-current environments that noise couples into copper data links and can even affect sensitive monitoring signals on smart PDUs.

Standards such as TIA-942 call for meaningful separation (often 12 inches or more for higher-voltage power). Crossing at 90 degrees is better than long parallel runs, but physical separation or metal barriers is the real solution.

Fix: Dedicated pathways. Power on one side (or one tray), data on the other. Color-code and enforce it.

3. Poor Labeling (or No Labeling at All)

When something goes wrong at 2 a.m., the last thing you want is a detective mission through unlabeled power cords. No labeling ranks among the most frequent operational mistakes. Technicians pull the wrong feed, waste time tracing circuits, or create new problems while trying to fix the original one.

Fix: Label both ends of every power cable with durable, machine-printed labels that survive heat and handling. Include circuit ID, source PDU/panel, destination, and A/B feed designation. Maintain a simple power path map.

4. Daisy-Chaining or Sharing Power Improperly

Daisy-chaining PDUs or power strips, or casually sharing circuits between racks, looks convenient until a single failure takes out more equipment than necessary. It also hides true loading, makes capacity planning impossible, and increases the chance of overloaded circuits.

Consumer-grade strips have no place in a data center. Even some “industrial” strips get used without metering.

Fix: Dedicated, properly rated metered rack PDUs. Maintain clear A/B redundancy paths. Never chain power strips. Know the actual load on every circuit.

5. Using Non-Compliant or Low-Quality Cables and Ignoring Management Basics

Cheap power cords with undersized conductors, poor copper, weak connectors, or no proper listings (UL, etc.) show up more often than they should. Add excessive coiling on the floor, tight zip ties that crush the jacket, or blocking airflow, and you create both fire risk and cooling inefficiency.

Power cables should be the correct gauge, properly rated for the environment (consider LSZH in many facilities), and managed so they do not obstruct airflow or create trip/fire hazards.

Fix: Spec certified cables that meet the actual current, temperature, and fire requirements of the space. Use Velcro instead of permanent zip ties where possible. Keep excess length minimal and neatly managed. Route to support—not fight—the cooling design.

Summary

Power cables are foundational infrastructure. Treat it with the same rigor as the computer and network layers and you avoid a lot of expensive surprises. Most of these mistakes are preventable with a bit of planning, the right materials, and consistent standards.