Arc flash studies for data center campuses, AC and DC.

Which specification your US site is being held to

Three instruments decide what your study has to produce, and they do different jobs. One sets the duty on the employer, one sets the calculation method, and one decides what an inspector reads off the label at the panel.

What changes on our side is the deliverable, not the argument: the DC systems are modelled and reported separately with the method named, and the whole package carries a Professional Engineer seal rather than an unsigned calculation pack.

Where we work in the United States, and how quickly we get there

Campus work runs on outage windows and permit gates, not on calendars. What follows is how the field team reaches your site and what it needs in place before it arrives.

Four bases, four ways the work reaches your campus

Campus electrical topology is broadly similar wherever the site sits. What changes is how fast a crew gets to it and what the local inspection regime expects to see on the label.

Atlanta and Georgia

The largest data center construction market in North America. Atlanta held 2,882 MW under construction at the end of H1 2026, ahead of Northern Virginia at 2,420.2 MW, after a 52.3 percent rise year on year (CBRE, North America Data Center Trends H1 2026, 27 August 2026). Campuses here energize in phases, and the incident energy at a board changes between one commissioning stage and the next, so a study signed at stage one is already describing a different plant by stage three.

Indianapolis and the Midwest

Greenfield campuses arriving at speed across Indiana and Ohio, with no incumbent engineering relationship and no legacy model to inherit. The first study is the one that sets the baseline every later phase is measured against.

Houston and the Texas Gulf Coast

The VB US base, and the deepest concentration of heavy electrical engineering talent in the country. Campus work across the South is served from here, alongside the refining and petrochemical estate the practice has worked in for a decade.

New Jersey and the Northeast metro

The densest operating capacity in the Northeast and the strictest inspection culture in the country. Qualified staff work 480 V bus duct, large-format UPS, transfer switches and medium voltage distribution around the clock, so the energized-work justification is a weekly question rather than an annual one.

What to put in the scope before you go to tender

Two quotes for the same campus can differ by a multiple, and the reason is almost never margin. It is what each one counted, and whether the DC systems were counted at all.

1. What the study must coverEvery bus, switchboard, PDU, remote power panel and UPS output in scope, counted. State whether battery strings and battery cabinets are inside the count or outside it.
2. What we will need from your plantSingle-line diagrams, nameplate data, cable schedules, protective device settings and utility fault levels. Whatever is missing, the walkdown captures.
3. How to tell a good answer from a cheap oneNormal, alternate feed, generator and tie-closed. Each operating scenario is a separate model run and a separate set of results, and dual-fed boards are the ones that multiply.

What to put in writing

Name the standard and the edition, not the family. State the method for the AC distribution and the method for the DC systems separately. Require the labels to be printed, installed and photographed, not merely supplied. Require a conformance matrix mapping every clause in your specification to the page of the report that satisfies it.

How to read what comes back

Ask each bidder what bus count their price was built on, whether the battery systems are in or out, how many operating scenarios are modelled, and who signs. A quote that cannot answer those four has priced a different job.

How site data is collected and verified at your plant

Where the AC study stops and the DC scope begins

This is the part most campus studies leave open, and it is the part a data center has most of. It is also where a colocation scope has to be explicit, because the label on a shared board is read by the operator's staff and the tenant's.

The standard stops before your battery rooms do.

IEEE 1584-2018 is validated for AC systems from 208 V to 15 kV. Your UPS DC bus, battery strings and battery cabinets sit outside that range, and no consensus DC standard covers them. A report that prints one incident energy figure across the whole site has not told you how the DC number was produced. Ask that question before you sign.

UPS, battery strings and battery cabinets

IEEE 1584-2018 is validated for AC systems from 208 V to 15 kV. A UPS DC bus, a battery string and a battery cabinet are none of those. There is no consensus DC standard, which means the method has to be stated rather than assumed. VB states it, prices that scope separately, and signs it.

800 VDC distribution for AI racks

AI halls drawing 400 kW and more per rack are moving to 800 VDC distribution, and that is a different question again from a 540 V UPS bus. A manufacturer study published in August 2026 modelled rack-level and centralised 800 VDC architectures and reported incident energy below 1.2 cal/cm² at rack level even without dedicated protection, with the centralised case falling to AC-comparable levels once standard protective devices were applied. The same study states plainly that no industry-wide framework yet addresses converter-fed 800 VDC systems, and that conventional calculation methods can overstate DC incident energy because they do not fully account for transient fault behaviour (ARC Advisory Group, 800 VDC arc flash analysis, 4 August 2026). Two things follow for your specification. A DC number is only meaningful if the method that produced it is named, and a conservative number is not automatically the safe one, because over-stated incident energy buys PPE you do not need and can push work into shutdowns that were avoidable. VB names the method, states the assumptions beside the figure, and signs it.

Dual-fed boards and transfer switches

Concurrent maintainability means equipment is worked on while the load stays up, by design. Dual-fed and multi-source topology raises the bus count, and bus count is what the study is built on. Each source configuration is a separate scenario, not a footnote to one.

Eighty-seven buses, five blocks, one campus

A large multi-building industrial campus. Five process blocks on one site, shared medium voltage distribution, and no single-line diagram anyone still trusted. Eighty-seven buses were modelled in ETAP from a walkdown rather than from the drawing, because the drawing and the plant had stopped agreeing years earlier.

Thirty-one buses came back above 25 cal/cm² in the as-found state, concentrated at the drive transformers. The reflex answer is heavier PPE. Relay settings alone resolved twenty-two of those thirty-one, a 71% reduction, before a single piece of equipment was bought. Nine buses needed engineering work and those were scheduled into the next planned outage rather than forcing an unplanned one. Two hundred and forty staff were trained against the new labels. The group EHS audit closed clean.

Read it as a campus and the shape is familiar. Five process blocks behave like five data halls. Shared medium voltage distribution behaves like a shared utility service. And the settings-first result is the reason a study pays for itself before the PPE order is raised.

Questions we get from US sites

What else your US site needs around the study