Seven Variables That Determine Whether Your Australian Data Centre or Supercomputer Site Will Actually Work

Australia is in the middle of a once-in-a-generation infrastructure build. Hyperscale operators, sovereign wealth vehicles, defence contractors, and university research consortia are all moving simultaneously, competing for the same finite pool of viable sites across a continent that is simultaneously large in landmass and deeply constrained in the variables that actually matter for critical infrastructure.

The constraint is not land. Land is abundant. The constraint is the convergence of seven interdependent variables at a single location: power, water, fibre, zoning, climate, talent, and regulatory speed. When all seven align, you have a site. When even one is misaligned, you have a liability dressed as an opportunity.

At Divine Lab Worx, I have built our entire practice around this reality. The firms and investors who come to us are not struggling to find land. They are struggling to identify which land actually works, and to understand why a site that looks compelling on three variables fails catastrophically on a fourth.

This article defines each of the seven variables, explains the failure mode associated with underweighting it, and describes what rigorous site intelligence looks like in practice. It is the framework we apply before we advise any client to commit capital to a location.

Why single-variable optimisation fails

• $100B+: Committed or announced data centre and sovereign AI infrastructure investment across Australia through 2030.
• 12%: Estimated proportion of shortlisted sites that achieve simultaneous alignment across all seven viability variables.
• $40M: Estimated cost per month of delay from a single-variable site failure after capital commitment and construction commencement.

The site that fails is almost never the one that looked worst on the short list. It is the one that looked best on the primary selection criterion and was never interrogated on the other six.

Dainu Devis, Commercial Architect, Divine Lab Worx

The Seven Variables

These are not independent factors. They interact. A site with abundant water access may sit inside a climate zone with extreme temperature variance that negates the cooling advantage. A site with fast zoning clearance may lack fibre carrier diversity that no amount of speed can compensate for. The framework must be applied as a system, not as a checklist.

Variable 01: Power Proximity and Grid Connection Capacity

Risk: Critical. A hyperscale data centre or sovereign AI compute cluster draws between 50 MW and 500 MW at steady state. The first question is not whether power is available in the region. It is whether the transmission infrastructure exists to deliver it at the required voltage, with redundancy, at the specific parcel boundary.

The connection queue at AEMO for large industrial loads is currently measured in years, not months. A site that requires a new zone substation or high-voltage transmission extension can add three to seven years to commissioning timelines before a single rack is installed.

We evaluate: proximity to existing 66kV or 132kV infrastructure, available headroom on the local feeder, AEMO connection queue position, renewable energy firming potential, and backup generation feasibility. Power is the single most frequent reason a site collapses at technical due diligence.

Variable 02: Cooling Water Access and Thermal Management Capacity

Risk: Critical. Modern high-density compute infrastructure generates thermal loads that air cooling alone cannot manage at economic scale. Liquid cooling and immersion systems require water: clean, cold, and available in volume.

Australia is the driest inhabited continent on Earth. Water entitlements in most inland regions are heavily regulated, contested, and increasingly subject to climate-driven restriction orders. A data centre drawing two to five megalitres per day for evaporative cooling requires a secured entitlement of that volume before construction begins, with no guarantee of continuity under future climate legislation.

This variable is increasingly the constraint that eliminates inland sites that pass the power test. We evaluate it before any site proceeds to detailed engineering.

Variable 03: Fibre Density and Carrier Diversity

Risk: Critical. Critical infrastructure requires network connectivity that is not only fast but structurally redundant. A data centre served by a single fibre provider on a single physical route has no meaningful network resilience. A single backhoe strike or carrier outage takes the entire facility offline.

Genuine carrier diversity means at least two independent physical routes, entering from different directions, served by different operators. In Australia, this exists in a relatively small number of locations, primarily metropolitan corridors and designated carrier exchange points.

A site without dual-carrier physical diversity does not progress to a capital recommendation from us, regardless of its performance on other variables.

Variable 04: Zoning and Environmental Clearance Timelines

Risk: High. Australia's planning and environmental approval system operates at state level with significant local government overlay. The combination of zoning classification, local environmental plan, and state environmental impact assessment regime can create approval pathways of anywhere from six months to six years.

Data centres are large industrial buildings with significant power infrastructure, cooling systems, backup generation, and noise output. They attract scrutiny from local residents, environmental groups, and heritage authorities. Sites that require rezoning before development approval add a full cycle of political risk to an already complex technical risk profile.

Zoning certainty is infrastructure in its own right. Speed matters. A site that requires three years of planning approval delays a project that has time-to-market pressure it cannot absorb.

Variable 05: Climate and Natural Hazard Exposure

Risk: High. Data centres have design lives of twenty to thirty years. A site evaluated under 2026 climate conditions will operate under materially different conditions by 2036. Climate exposure is not a static variable. It is a trajectory that must be modelled forward.

The hazards relevant to Australian infrastructure sites include cyclone track probability in northern Queensland and WA, bushfire risk in coastal and inland interface zones, flood frequency in river corridors, extreme heat events that stress cooling systems and power grids, and sea level rise in coastal locations.

Our climate analysis uses CSIRO projection data in conjunction with state flood mapping, bushfire hazard overlays, and infrastructure resilience standards to assess both current exposure and thirty-year forward trajectory.

Variable 06: Talent Pool Proximity

Risk: High. This variable is underweighted in most site selection frameworks. The assumption is that facilities operations teams are small and can be recruited locally. That assumption is wrong for two reasons.

First, the organisations using the facility need to physically access the site for deployment, maintenance, and integration work. A facility four hours from the nearest major labour market creates chronic friction for every tenant. Second, data centre technicians, high-voltage engineers, cooling specialists, and network operations professionals are sourced from a limited pool. A site in a region with no viable talent pipeline will face ongoing operational staffing costs that erode the economic case year after year.

Variable 07: Regulatory Pathway Speed

Risk: Medium. Distinct from zoning clearance, regulatory pathway speed encompasses the full landscape of approvals that determine how quickly a development moves from site control to operations: AEMO connection agreements, water entitlement transfers, telecommunications licence conditions, biosecurity clearances, and, for facilities handling sovereign or classified workloads, Defence and security agency requirements.

The sites best positioned for sovereign infrastructure are those where the regulatory pathway is not only manageable but predictable. Where agencies have existing relationships with the jurisdiction. Where precedent exists for similar approvals. Where the operator can engage regulators before site commitment rather than discovering the full pathway after capital is deployed.

Regulatory unpredictability is a capital risk, not just an administrative inconvenience. We map the full regulatory matrix for each shortlisted site before issuing any recommendation.

Why Single-Variable Optimisation Always Fails

The history of failed data centre and compute facility projects in Australia is largely a history of single-variable optimisation. A development consortium identifies a site with cheap land and good access to a solar generation corridor. The power economics look compelling. Capital is committed. Construction commences.

Then the connection queue position becomes clear. Or the water entitlement review triggers a restriction order. Or the carrier diversity assessment reveals that the site can only be served by a single fibre route that adds fourteen milliseconds of latency and zero redundancy.

At that point the capital has moved. The development agreement is signed. The cost of pivoting exceeds the cost of proceeding. The project goes ahead and the facility operates at a structural disadvantage for its entire asset life, or fails to attract the anchor tenant that made the business case.

Site intelligence is not a report you commission after you have identified a preferred site. It is the process by which the preferred site is identified in the first place. Those are fundamentally different briefs.

Dainu Devis, Commercial Architect, Divine Lab Worx

What Divine Lab Worx Delivers

Our engagement model is built around three phases that correspond to three stages of capital commitment.

• Pre-Commitment Screening: We apply the seven-variable matrix to a long list of potential sites, using desktop data, GIS analysis, infrastructure mapping, and regulatory research to eliminate sites that cannot pass on one or more critical variables. This phase prevents capital from being committed to a fatally flawed site before due diligence.
• Technical Due Diligence: For shortlisted sites, we conduct detailed site-level assessment, engaging with network operators, reviewing AEMO connection data, analysing water entitlement positions, commissioning climate hazard modelling, and mapping the full regulatory pathway. This phase produces a ranked site recommendation with a risk-adjusted infrastructure cost model.
• Deployment Architecture: Once a site is selected, we design the deployment architecture, infrastructure sequencing, utility connection strategy, and regulatory engagement programme to optimise the path from site control to operational facility. This is where the Sharktech Global AI platform layer integrates, providing the operational intelligence stack that runs across the facility once it is live.

Each phase is scoped to the risk level of the capital decision it informs. Screening happens before site control. Due diligence happens before construction commitment. Deployment architecture happens before operational handover.

Seven variables. One integrated framework. No wasted capital.