The same handful of causes sit behind many data-centre delay claims, and none of them should be a surprise by the time they land. They are the failure modes we know from energy, grid-connection and M&E-integrated work — the systems a data centre is built from. Here are the ten worth watching, and acting on before they turn into a dispute.
1. The grid connection
Power energisation is the master key; the rest of the job hangs off it. Connection dates slip by years, and when they move, the whole fit-out and commissioning sequence moves with them. In October 2026 a hyperscaler reportedly withdrew from anchoring a UK AI campus after the connection date slipped into the 2030s. Who carries that risk depends on the contract, and it is worth knowing before the date moves. It is a programme risk, a cost risk and, often, a termination.
2. Long-lead equipment
Switchgear, transformers, generators, UPS and chillers now sit on 12–18 month leads, sometimes longer. One late or re-sequenced delivery re-plans the job around it. If the order dates and the programme were never reconciled, the slippage is baked in before a spade goes in the ground. Check who owns the kit as well. If the employer supplies it (OFCI), late or damaged delivery is usually the employer’s risk, and the contractor’s time and cost entitlement depends on the programme showing the delivery windows.
3. Commissioning Is Often The Critical Path
Level 1–5 commissioning and integrated systems testing usually decide the go-live date, not practical completion of the build. Many programmes under-model it — a nominal bar at the end rather than the real, logic-linked sequence. When the IST runs long, the claim is already weak because the baseline never showed the path.
4. Client change and GPU churn
Power density, cooling and rack layouts move mid-build as the end user’s hardware changes. Each change ripples through the M&E and back into commissioning. Priced and recorded properly, these are compensable change (compensation events under NEC); left loose, they become the employer’s argument that the contractor caused its own delay.
5. Trade Stacking And Disruption
Services-dense halls, too many trades in too little space, work taken out of sequence to hold a date. Disruption rarely shows on the critical path but quietly drains productivity. Proving it later needs the measured-mile records nobody kept — unless someone set them up early. Concurrency is a separate question: it arises where two or more delay events from different causes are effective at the same time, and it needs its own analysis.
6. A baseline nobody owns
An Accepted Programme that was accepted once and never revised works as wall decoration, not as a baseline. Without contemporaneous updates and a proper revision history, the delay analysis at the end is built on sand, whichever side is running it.
7. The records gap
As-built dates, progress data, site records and delay notices kept to the standard a claim needs — or not. The contract usually spells the obligation out; the pressure of the job usually buries it. The cheapest hour in the whole dispute is the one spent getting the record-keeping right in month two.
8. Planning to the wrong milestone
The date that triggers damages is set by the contract. On hyperscale work it is often ready-for-service or customer handover, not the Completion Date the programme drives to. Teams plan to the date on the front of the contract and miss the one that actually bites. The two can be weeks or months apart.
9. Supply-chain failure
A specialist going under mid-scope strands the work it was holding. The 2024 collapse of a major UK contractor with a data-centre order book left live projects to be finished and a trail of final-account and delay disputes behind them. Employers need completion-delay and damages analysis; subcontractors need their loss-and-expense properly built.
10. Leaving it too late
The go-live the sales team promised was often never buildable, and everyone works to it until the gap is undeniable. By then the honest conversation is overdue, and so is the delay case. EOT entitlement can erode with every un-notified event and every un-updated programme, depending on the contract; under NEC the notification time bar is strict. The firms that recover time and money treat the delay case as a live document from day one; leave it until the account turns sour and half the entitlement has already gone.
Reference: Microsoft / Nscale UK power-connection delay, as reported 18/10/26.
Most of these are visible before they become a claim. A short independent delay-risk review on a live programme is the quickest way to find out where you stand — see data-centre programme, delay and dispute support.