Ninety planned sites means ninety separate consent processes, and the money behind them divides more evenly than the trouble does. Spread the committed capital across the pipeline and it comes to roughly A$1.7bn a site [19]; the pipeline itself is about a 56 percent lift on the count already operating [18]. Those are the figures that travel well in an investment committee. The figures that decide whether one particular building gets poured are in metres, and the BBC supplies two of them.
The Lane Cove hum arrives from 3.5 times the distance the Marsden Park houses are dealing with [20], and it still reaches a child's bedroom [2]. A continuous low sound at bedtime is not a function of how many megawatts the site draws; it is a function of how far the plant sits from the nearest sleeping person, and of whether anyone specified the fans with that person in mind.
Teams tend to frame the objection as being about emissions, water and grid capacity, arriving as a written submission during public exhibition. What residents actually notice first, in the BBC's own account of it, is the size, and the sound, described as either a jet plane about to take off or a high-pitched whine [4]. Water and power turn up later, in the arguments people reach for once they already want the thing gone.
The pitch and the build are answering different questions. Belinda Dennett of Data Centres Australia makes a national argument, that without local infrastructure Australia keeps none of the value chain and becomes an importer of tools [7], and a technical one, that proximity buys latency for things like air traffic control and robotic surgery [8]. Both points hold up on their own terms, but they are not the question a council is actually weighing. A council hears one building, one street, one set of noise conditions.
Dennett's latency point is also the forcing function, if you take it literally. Two axes for every site in that pipeline of 90. First: does this workload's value actually depend on sitting close to its users, or is that a habit inherited from the era before the load got this big? Second: is the nearest dwelling inside a few hundred metres? A site that is both latency-bound and close in is worth buying the consent up front, in acoustic treatment and in what you tell the neighbours before they hear it. A site that is close in but not latency-bound is paying a consent premium for nothing and should be looking further out. A site that is latency-bound but far out is the one you want. And a site that is neither is the default, and probably where most of the pipeline belongs.
The source does not break the 90 down by workload, and that breakdown is exactly what a planner would need. What it does show is that the constraint is already binding: the one-gigawatt western Sydney project, the biggest single energy user in the country if it happens, is described as pending approval [5]. The Australian Energy Market Operator's forecast that data centre demand could triple by 2030 [13] is a national number that no individual consent authority is deciding. Jon Whittle, formerly of CSIRO's digital research centre, says the government has no real choice but to keep investing [12]. That argument is about the country as a whole; it does not settle what happens at the site in Lane Cove, where four more are queued at the same industrial park [3].
