Home  ›  Blog  ›  Bus Duct for Data Centers

Bus Duct for Data Centers: What Should You Specify?

By DHYN Engineering Team  ·  August 19, 2026

Quick answer: For a data center, specify overhead plug-in busway with dual A and B feeds per row, copper conductors, tap-off boxes with breakers and per-circuit metering, joint temperature monitoring feeding your DCIM, and compliance with IEC 61439-6 or UL 857 depending on the market. Size each feed for the full row load at realistic rack density, and leave spare tap-off capacity, because rack power only goes up. (Answer time: August 19, 2026)

Fifteen years ago, most data centers distributed power in cable trays under a raised floor. Today, when we quote a new build, the default question is which busway to run above the racks, not whether to run one. The shift happened for practical reasons, and the specification details matter more than most buyers realize.

Why busway beat cable in the white space

Cables under a raised floor block the very airflow the floor exists to deliver. Every added circuit made the cooling problem worse, and tracing one failed feed in a tray of four hundred look-alike cables is nobody's idea of fun. Overhead busway removes both problems. Air moves freely, and each rack's feed is a visible, labeled tap-off box you can reach from a ladder.

The bigger driver, though, is change. Racks get reconfigured constantly. With cable, a new 32 A rack feed means pulling a new cable from the PDU. With plug-in busway, it means clipping a tap-off box onto the duct, often live, in under an hour. Operators who have done both do not go back.

There is an efficiency argument as well. A busway's compact, closely coupled conductors have lower impedance than an equivalent cable run, which trims distribution losses. It is a small percentage, but in a facility where every wasted kilowatt worsens the PUE, small percentages get budget.

The specification checklist

When we help a client write a data center busway spec, these are the lines that decide whether the installation is loved or cursed for the next fifteen years.

Item	What to specify	Why
Architecture	Two independent busways per row (A and B feed), physically separated routes	Concurrent maintenance without dropping the row; a fire or fault on one path leaves the other live
Rating per feed	Sized for full row load, typically 250 to 630 A today, derated for ambient temperature	Either feed must carry the whole row alone during A-side maintenance
Conductor	Copper, tinned at joints	Smaller duct above crowded racks, better joint stability; see tin vs silver plating
Neutral	100 percent neutral minimum, 200 percent where IT harmonic loads dominate	Third harmonics from server power supplies add up in the neutral instead of canceling
Tap-off boxes	Breaker-protected, per-circuit metering, hot-swappable design, spare positions reserved	Per-rack energy data and live reconfiguration without an outage window
Monitoring	Joint temperature sensors plus metering over Modbus TCP or SNMP into the DCIM	Hot joints are the failure mode; see joint overheating warning signs
Standard	IEC 61439-6 with test reports, or UL 857 listing for North American sites	Inspectors and insurers want the mark that matches the jurisdiction
IP rating	IP40 to IP54 is normal for conditioned white space	Higher ratings cost money and trap heat you do not need to trap indoors

The density question you cannot skip

A row that drew 5 kW per rack in 2015 may be planned for 20 or 30 kW today, and AI hardware pushes individual racks well past that. The busway you hang this year should carry the row at its end-of-life density, not its day-one density, because replacing an energized busway above live racks is a project nobody wants. We usually tell clients to size the duct for the ten-year density forecast and let the tap-off boxes handle today's smaller per-rack loads. The sizing method itself is in our ampacity calculation guide.

Monitoring is where the value compounds

A plain busway is already better than cable for most data centers. An instrumented one is better still. Per-tap-off metering gives you per-rack energy figures, which is the raw material for PUE analysis and tenant billing. Joint temperature sensing catches the one failure mode that matters, a loosening joint, weeks before it becomes an incident. We covered the full picture in what is an intelligent busway system, including when the monitoring premium pays for itself and when a standard duct is honestly enough.

Mistakes that show up in year two

The classic error is running the A and B busways down the same cable route side by side. One falling ladder or one fire takes out both, and the redundancy was decoration. Keep them on opposite sides of the row, ideally with a physical barrier between. Second: no spare tap-off positions. Rows grow denser, and a duct with every outlet used forces ugly workarounds. Third, under-specifying the neutral for harmonic-heavy IT loads, which cooks the neutral bar at loads that look safe on a per-phase basis. And fourth, accepting a datasheet without the IEC 61439-6 temperature rise and short-circuit test reports. A busway over a hall of live servers is the last place to take a rating on faith.

Get the architecture and the paperwork right, and the busway becomes the part of the data center you never think about. That is the goal.