Why Data Centres End Up in the Same Few Suburbs

Data centres are not distributed by where the demand is. If they were, they would follow population fairly evenly.

Instead they concentrate in a handful of areas, and the same few suburbs appear again and again. That is not coincidence or herd behaviour. It is what happens when three separate constraints have to be satisfied at once and the overlap between them is small.

Constraint One: Transmission With Headroom

The binding constraint is almost always electricity, and specifically where the grid can already deliver it.

A large facility needs a connection at a scale that most of the network cannot supply without reinforcement. Building that reinforcement — new transmission, upgraded substations — takes years and requires regulatory approval, so what actually determines siting is where spare capacity exists on infrastructure that has already been built.

Connection queues follow. When several large loads want to connect in the same area, they join a process that assesses each one’s effect on the network, and that process can take a long time. A site with an existing connection agreement or a clear path to one is worth a great deal more than an identical site without one, which is why land near substations with headroom carries a premium that has nothing to do with the land.

Constraint Two: Fibre Already in the Ground

The second constraint is connectivity, and the important word is existing.

Trenching new long-haul fibre is expensive and slow, and it needs its own approvals and easements. So operators site facilities on or near routes that already exist — and those routes were laid decades ago along corridors that were convenient then: rail lines, highways, existing utility easements.

Redundancy compounds it. A serious facility wants diverse physical paths, so that a single backhoe cannot isolate it. Genuinely diverse routes are rarer than single ones, and they intersect in fewer places.

Constraint Three: Land a Council Will Approve

The third constraint is planning, and it is the one that has changed most.

A data centre is an industrial use with an unusual profile. It occupies a large footprint, runs continuously, generates noise from cooling plant, may have significant water requirements depending on the cooling design, and — the politically important part — employs very few people relative to its size.

That last point is the crux. A council weighing an industrial development usually trades amenity impact against employment. A facility with a large footprint and a small permanent headcount offers a weaker version of that trade than a factory or a distribution centre would, so the political case rests on rates revenue and investment rather than jobs.

The result is that facilities go where the land is already zoned for it and the local planning environment is known to be workable — which, again, means going where others have already gone.

Why Latency Matters Less Than It Used To

There is a fourth consideration, and it is worth understanding because it is changing the pattern.

Some workloads are genuinely latency-sensitive. Interactive services, trading, real-time applications — these want to be physically close to their users, which pulls facilities toward population centres regardless of what the other constraints say.

Large-scale training workloads are not like that. A training run does not care whether it is fifty kilometres from a city or five hundred. What it cares about is power availability and cost.

That difference is pulling siting in two directions: latency-sensitive capacity staying close to cities and paying for it, and power-hungry capacity willing to go wherever the electricity is. Where those two used to be the same decision, they increasingly are not.

The Australian Version

The general constraints apply everywhere. Australia sharpens them in a specific way.

Population is concentrated in a small number of coastal cities, so the latency-sensitive requirement points at a handful of metropolitan areas. International connectivity arrives via a limited number of subsea cable landing points, which concentrates the network geography further. And grid conditions vary considerably by region, so transmission headroom is unevenly distributed.

Stack those together and the set of viable sites is genuinely small — which is why the same industrial corridors keep appearing, and why land in them is valued on the basis of its grid connection rather than its area.

Why It Matters Beyond the Industry

For a business with no interest in where servers live, two consequences are worth knowing.

Concentration is a resilience question. Capacity clustered in a small number of corridors shares exposure to regional events — grid disturbance, flooding, a fibre cut affecting multiple routes. Availability zones in the same metropolitan area are less independent than the marketing implies, and if continuity genuinely matters, the question to ask a provider is about physical separation rather than logical separation.

And the siting constraints are why capacity does not simply appear where demand grows. Grid connection queues and planning timelines are measured in years, which means regional pricing and availability can stay tight for a long time after demand has moved.