Clarke Energy v Alinta: What a 2026 Delay Judgment Means for Your Programme
A $78 million power station, squeezed onto a small site between two live ones in remote WA. The principal promised to move the underground services before the contractor started. It did not. Later, a superintendent signalled he was minded to certify practical completion and was promptly replaced. When Clarke Energy (Australia) Pty Ltd v Alinta Energy Transmission (Roy Hill) Pty Ltd [2026] NSWSC 1124 reached the NSW Supreme Court, the contractor came away with its extensions of time and its delay costs. It won the way these cases are reliably won: on the programme and the contemporaneous record, backed by the contract's wording and the parties' conduct. Five things carry across to any job you run.
One note before the lessons. In this considered judgment, handed down by her Honour Justice Rees on 21 September 2026, the Supreme Court of New South Wales worked through the delay analysis in detail. As a first-instance decision it is not yet appellate authority and may be subject to appeal, so it is best read as a well-reasoned guide rather than settled law.
1. Know whose float it is, and what it actually buys
The Court held that, under this contract, Clarke owned the float in its programme and could spend it for its own benefit (at [130]-[134]). Then it drew the line contractors keep tripping over. Where an activity's float is eaten up only because a delay event pushed that activity onto the critical path, you get nothing extra for the lost float. Only delay to practical completion counts, for time (at [524]-[526]) and for delay costs (at [528]). So keep two questions apart: who has the benefit of the float, and whether the event has actually moved practical completion. Record the float you consume, but build the EOT on the demonstrated movement of the finish date, not on the float itself. Float ownership here was a finding on this contract's words, not a universal rule, so do not assume the next contract hands you the same.
2. A genuine later revision is evidence about your baseline
This is the sharpest lesson in the case, and it cuts against instinct. Clarke's September 2020 baseline ran powerhouse electrical installation alongside cladding, on a Start to Start link. In the next programme the logic changed to Finish to Start, and it stayed that way for six to twelve months. The delay experts disagreed on what that meant. Clarke's expert, Mr Griffith, treated the later, persistent change as evidence that the September programme's original sequence was not reasonable, achievable and appropriate, the very test the contract set at clause 12.10(g). The Court accepted that approach (at [832]).
Here is the part that matters for how you run your programme. The change carried weight precisely because it was genuine. It came out of three Incident/Injury Free workshops with Ahrens and Alinta, made to reduce injury risk, and it was not a switch made after a delay event to inflate the claim (at [833]). And the fact the works were ultimately done differently again did not undermine it, because time impact analysis looks at the effect of the event on the intended forward plan, not the as-built (at [834]).
So the point is not that any tidy-up sinks your baseline. It is that a programme's version history is evidence, and a substantive, persistent change can be used to test whether your starting logic was ever sound. Get the baseline logic right and achievable on day one, justify every link, and record why every later change was made, who approved it and what site information drove it. A contemporaneously explained revision is far easier to defend than an unexplained one.
3. Time impact analysis stands or falls on its inputs
The Court worked with time impact analysis (at [407]-[408]), the SCL Protocol method described in Alstom v Yokogawa: take a programme progressed to the point of the delay, insert the event, and measure how far forecast completion moves. The method was never the fight. The fight was the inputs. A baseline that reflects real progress to the delay, and an event modelled the way it actually bit, is what makes the number hold. It does not stop at the movement in the finish date either, the analysis has to test concurrency and mitigation, deducting concurrent delay and accounting for the contractor's duty to mitigate (at [417]). A persuasive analysis is one another scheduler can rerun, so keep the native XER, the data date, calendars, constraints, progress evidence, the fragnet logic and the pre and post impact outputs, and state every assumption and tie it to the contemporaneous record. This is forensic delay analysis done so it survives a second pair of hands.
4. Serve the notice right, and do not count on being rescued
Clarke served its EOT claim on the principal and did not copy the superintendent, though the clause required both. That was a potentially serious condition-precedent problem. It did not sink the claim, for two reasons. Clarke moved quickly once it spotted the omission, sending the missing claim and the earlier notices to the superintendent. And Alinta pressed the defect, then chose to have a later superintendent determine the claim substantively, so the Court found that in choosing between inconsistent rights Alinta had waived reliance on the condition precedent (at [761]-[766]).
Do not build a claim around that. The clause also showed that not every step in the EOT machinery was a condition precedent, only some were, which is exactly why you have to know which is which. Treat notice as a controlled workflow: map every recipient, deadline, content requirement and update obligation the contract names, keep proof of service, cure any slip the moment you find it, and log any conduct by the other side that might amount to waiver or election. Then you are not relying on a court to pull you out. This is the discipline behind every extension of time claim that stands up.
5. Write every record for the eye that will test it
The contract required the superintendent to act honestly, fairly and reasonably. The judgment was blunt about what happened when that slipped: private communications between a superintendent and one party undermine the role, and removing a superintendent to stop a determination you do not want "effectively destroys the role". A superintendent who indicated he was minded to certify practical completion was promptly replaced, and the Court noticed. Whichever side of the table you sit, work to one assumption: every programme, every update, every progress note and every email may be read one day by a superintendent, and after that by a judge. The strongest claim file is not a polished document built after the event. It is the chain of notices, programme versions, progress reports and reasoned correspondence created while the event is still unfolding.
What a programme built this way gives you
A baseline whose logic holds up, updates that match what happened on site, and a record that carries the extension of time when it is challenged. That is the Primavera P6 discipline behind the construction scheduling we run every day. This is a scheduler's read of a recent decision, not legal advice, and on a live claim the contract and the facts will drive the answer.
Frequently asked questions
What is time impact analysis?
A forward-looking delay method: take a programme progressed to the point of the delay, insert the delay event, and measure how far it pushes forecast completion, in the light of progress actually achieved at the time of the event.
Who owns the float in a construction programme?
It depends on the contract and the programme regime. In Clarke Energy the Court held that, on this contract, the contractor owned the float and could use it for its benefit.
Can I claim delay costs for lost activity float?
On the contract in Clarke Energy, no. Under this contract, recoverable delay costs had to flow from a compensable delay to practical completion, and lost activity float was not a separate basis for them.
Do I have to copy the superintendent on an EOT notice?
Follow the clause exactly. Here the contractor served the principal but not the superintendent, cured the omission once it was found, and the claim survived because the Court found Alinta had waived reliance on the defect. Do not rely on that.
