Data Centre Design Checklist: What Electrical Contractors Need to Know Before an AI-Ready Retrofit

A contractor who's been bidding data centre jobs for over a decade told us something that stuck: the retrofits that go wrong almost never fail on the big-ticket items — the switchgear, the UPS, the cooling plant. They fail on the parts everyone assumed would just carry over from the original build. Tray capacity. Cable routing. The assumption that a rack pulling 6kW and a rack pulling 30kW can share the same containment plan.

That assumption is exactly what's breaking down across India right now, as facilities built for a pre-AI world get retrofitted to handle GPU-dense compute. For electrical contractors and EPC firms bidding this work, the difference between a profitable job and a change-order nightmare usually comes down to what got checked before the bid went in, not what got fixed after.

Why AI-Ready Retrofits Aren't Just an IT Upgrade

It's tempting to treat a data centre retrofit as primarily a compute and cooling problem, with electrical work as a downstream consequence. That framing causes more bid errors than almost anything else. Rack densities that used to sit comfortably at 4–8kW are now routinely specified at 25–40kW for AI training and inference clusters, and every kilowatt increase cascades backward through the electrical chain — larger conductors, more parallel runs, higher fault currents, and containment systems that were never sized for this kind of cable volume.

A contractor walking into a retrofit site with a legacy single-line diagram in hand is often working from numbers that no longer describe the building they're standing in. The first real task isn't installation — it's re-verifying what the existing infrastructure can actually support.

Load Calculation Changes Contractors Must Plan For

Data centre load calculation for an AI-ready retrofit isn't a simple multiplication of the old figures. Power draw at the rack level has to be reconciled against upstream transformer capacity, UPS sizing, and — critically — short-circuit current at every point downstream, since higher available fault current changes what protective devices, cable cleats, and containment hardware are actually rated to handle safely.

This is where a lot of retrofit bids go sideways. A contractor pricing the job purely on cable footage and labour hours, without re-running the fault-current study for the new load profile, risks specifying restraint and protection hardware that's technically inadequate for the environment it's going into. That's not a paperwork problem — it's a safety and liability one, and it tends to surface only after commissioning, when it's the most expensive time to fix.

Cable Tray and Containment Upgrades

Legacy trays sized for a handful of Cat6 runs and modest power feeds rarely have the fill capacity or clearance for the bundled, high-current runs an AI retrofit demands. Contractors evaluating an existing site need to look past whether a tray is "still there and usable" and ask whether its fill ratio, support spacing, and clearance from adjacent circuits still meet code once the new cable count and diameter are factored in.

Ladder trays generally hold up better than solid-bottom trays in these upgrades, since they allow better heat dissipation for higher-current runs, but even ladder tray systems often need additional supports, wider sections, or entirely new runs parallel to the old ones rather than a simple swap-in. This is usually where a retrofit budget either holds or blows past its original estimate — and it's worth pricing conservatively rather than assuming the existing containment infrastructure has more headroom than it does.

Cable Management and Standards at Higher Densities

Cable volume and bundling requirements change significantly at higher densities — we cover this in detail in our guide on high-density data centre cable management, but the short version for contractors bidding retrofit work is this: the restraint, routing, and labelling hardware that worked fine at legacy densities usually can't be reused wholesale. Cable management planning for an AI-ready retrofit needs to account for larger bundle diameters, tighter bend-radius constraints on higher-gauge power cables, and short-circuit restraint hardware rated for the actual fault current the retrofit will carry, not the fault current the original design assumed.

Installation speed also becomes a real cost factor once cable counts climb. Tools like the Panduit GTS-E cable tie gun exist specifically because manual tensioning and cutting on a high-volume retrofit adds up fast in labour hours — a tensioning tool that delivers consistent, repeatable strap tension across hundreds or thousands of ties isn't a convenience item on a job this size, it's a scheduling one. Contractors pricing labour hours without accounting for install-speed tooling on high-cable-count jobs tend to underbid the timeline, which shows up later as pressure on the crew rather than a line item anyone budgeted for.

Sourcing matters here too. Working with an authorised Panduit distributor rather than a grey-market reseller means the cable ties, cleats, and tools actually carry the fault-current and mechanical ratings printed on the datasheet — which matters directly when a retrofit's short-circuit study is the basis for what hardware gets specified in the first place.

A Practical Pre-Bid Checklist for EPC Firms

Before submitting a bid on an AI-ready retrofit, it's worth confirming a short set of things directly on-site rather than relying on the facility's existing drawings. Re-verify the current fault-current rating at each distribution point against the proposed new load, rather than trusting the original design study. Physically inspect tray fill ratios and support spacing under the new cable count and diameter, not just the tray's rated capacity on paper. Confirm containment and restraint hardware — cleats, ties, raceways — is rated for the retrofit's actual short-circuit conditions, and sourced from an authorised distributor whose ratings can be verified. And price installation tooling and labour hours based on the real cable volume for the new density, not the volume the original facility was built around.

Skipping any one of these is usually how a competitively priced bid turns into a loss-making job by the time commissioning rolls around.

Frequently Asked Questions

What changes electrically when a data centre is retrofitted for AI workloads?

Rack power density increases significantly, which raises conductor sizing, cable bundle volume, and available fault current at every downstream distribution point — all of which affect load calculations, tray capacity, and restraint hardware ratings.

Why do old cable trays often fail in AI-ready retrofits?

Legacy trays are frequently undersized for the fill ratio, clearance, and heat dissipation needs of higher-current, higher-volume cable runs, even when they appear to have physical space remaining.

How does cable management differ at higher data centre densities?

Higher densities require larger bundle diameters, tighter bend-radius planning, and short-circuit restraint hardware rated for the retrofit's actual fault current — not hardware carried over from the original lower-density design.

Why does it matter if cable ties and cleats come from an authorised Panduit distributor?

Genuine, authorised-distributor hardware carries verified fault-current and mechanical ratings, which matters directly when a retrofit's electrical safety case depends on those ratings being accurate.

For a deeper technical breakdown of why rising rack densities are driving this shift, see Why the AI Boom in India Is Driving Demand for High-Density Data Centre Cable Management.