INDUSTRY PRACTICE GUIDE | DIGITAL INFRASTRUCTURE
A practical guide to developing data centres in Australia when energy, not land, is the binding constraint
Version 1.0 · September 2026
Executive summary
The constraint on Australian data centre development has moved. Land, latency and capital were the historic gating items; firm, timely, low-emissions electricity now sits ahead of all of them. This is no longer a forecasting debate. AEMO has separated data centres into their own load category, the connection queue is being reported publicly, and Commonwealth and state governments are legislating obligations that attach energy conditions directly to a connection.
That shift changes what a competent development process looks like. A feasibility study that treats grid supply as a procurement line item rather than a design constraint is now producing an unbankable answer. The projects clearing financial close are the ones that arrive with a credible energy proposition already engineered: additional generation, contracted firming, demonstrated flexibility, and a defensible position on water and heat.
This guide restates that problem in Australian terms and gives you the decision tools to act on it. It is deliberately practical: what to test at site selection, what to price into capex, what to commit to in a connection agreement, and what to measure once you are operating.
Seven takeaways
- Power availability is now a site-selection gate, not a criterion. Score candidate sites on deliverable energy first; anything that fails the energy gate is a stranded asset regardless of its land economics.
- Assume you will be required to underwrite new renewable generation and firming. Australian policy has converged on this position across the Commonwealth and multiple states. Price it into base capex, not into a sustainability contingency.
- Flexibility is the cheapest concession you can offer. Peer-reviewed reviews find real, automatable flexibility in both IT workloads and cooling plant — but the magnitude and response speed vary sharply by mechanism, so it must be designed in, not assumed.
- Design flexibility explicitly, then contract it. Uncoordinated load behaviour by large programmable loads is itself a system risk identified in the literature; the commercial value sits in flexibility that a network operator can rely on.
- Treat water as a first-order approval risk. Cooling architecture determines water usage effectiveness far more than operating discipline does and closed-loop liquid cooling can approach zero site water use.
- Waste heat is a siting decision, not a retrofit. The peer-reviewed case for heat recovery is strong where an off-taker exists. Australia lacks legacy district heating, so the off-take has to be found at site selection or it will never exist.
- Digital twins are an operating asset, not a marketing artefact. Their value is in continuous optimisation of a facility against its own generation and constraints — which is precisely what a flexibility obligation will require you to evidence.
1. What has changed in Australia
Three things happened in 2026 that together reset the development calculus.
1.1 Data centres became a named driver of national electricity planning
AEMO’s 2026 Integrated System Plan, released 25 June 2026, treats data centre load as a distinct, separately modelled driver of national consumption for the first time — previously it sat inside general business demand. The August 2026 Electricity Statement of Opportunities put the Step Change projection at around 34 TWh of data centre consumption by 2035–36, against roughly 5 TWh in 2025–26, lifting the sector’s share of NEM operational demand from about 3% to about 13%. AEMO considered 225 projects in that forecast.
Two details matter more than the headline. First, AEMO notes data centres run relatively flat across the day and seasons, which presses on the system in exactly the hours it runs thinnest. Second, more than a third of the data centre projects AEMO listed a year earlier have since been cancelled — a strong signal that the pipeline contains far more optionality than deliverable capacity.
1.2 Energy obligations are being attached to the connection
On 5 August 2026 three developments landed together: the AEMC published its advice to Energy Ministers on requiring data centres to fully offset demand through new renewable generation and firming; the Commonwealth Energy Minister lodged rule change requests to ensure data centres pay the network costs they cause or accelerate; and NSW introduced legislation giving the state minister power over grid access and network cost allocation.
The AEMC’s recommendations are structural: mandate offsetting via certificates linked to new renewable generation, introduce a contracting obligation for firming capacity, introduce market registration requirements for large inverter-based loads to improve system visibility, and support demand flexibility and co-location with generation through connection agreements.
The direction is consistent with what Ireland has already implemented. Ireland’s CRU decision of 12 December 2025 requires new data centres above 10 MVA to provide dispatchable generation or storage matching their maximum import capacity, participate in the wholesale market, and meet at least 80% of annual demand with additional Irish renewable generation on a six-year glide path. Ireland reached that point after data centre demand grew from 5% of national electricity use in 2015 to 22% in 2024.
1.3 The states are competing on approval certainty, not incentives
Victoria, South Australia, Tasmania and New South Wales all now hold dedicated data centre policy instruments. South Australia’s June 2026 strategy pairs a coordinated planning pathway with a stated principle of “new energy for new demand”, and requires SA Water advice on water sufficiency and a Technical Regulator certificate on power system compliance as part of the application. New South Wales published Data Centre Guidelines on 17 August 2026.
The practical consequence is that jurisdiction selection is now partly a regulatory-velocity decision. A site with a slower planning path and no water pathway can lose to a technically inferior site with a coordinated approval route and a settled energy position.
2. Six evidence-based insights and what they mean here
Each insight below states what the peer-reviewed literature supports, then what a developer should do about it in the Australian context. Numbered references correspond to the evidence base in Section 6.
2.1 Flexibility is the highest-value, lowest-capex concession
Crozier and Liska’s 2025 review in Current Sustainable/Renewable Energy Reports assesses the state of research on data centre flexibility and finds several distinct mechanisms by which data centres can provide it . The important qualification, echoed in the broader review literature, is that the magnitude and response time of that flexibility vary significantly by mechanism and by facility type — IT workload shifting, cooling plant thermal inertia and on-site storage are not interchangeable
Al Kez and Foley’s review of programmable load risks makes the counterpoint that should shape your engineering: data centres possess the technical capability for high-speed, precision-aligned response, but their integration into demand response frameworks remains constrained by regulatory inertia, visibility gaps and operational risk aversion — and uncoordinated load behaviour by programmable loads poses systemic risk in its own right.
2.2 Cooling architecture, not operating discipline, sets your water risk
Mytton’s Perspective in npj (Nature Partner Journals) Clean Water established the core transparency problem: data centres consume water directly for cooling and indirectly through the water requirements of electricity generation, and fewer than a third of operators measure their water consumption at all. Privette, Barros and Cai reinforce in AGU (American Geophysical Union
) advances that while aggregate national water use may appear modest, localised impacts in water-stressed regions can be significant, and the transparency gap undermines regulation, innovation and community planning .
The engineering picture is now well quantified. Cooper and Nguyen’s review in Fluids surveyed reported water usage effectiveness across 83 data centre entries and found dry facilities using closed-loop liquid cooling require essentially no water, while evaporative systems typically report WUE values up to 2.5 L/kWh; 23 of the surveyed facilities exceeded the 0.4 L/kWh target set by the Climate Neutral Data Centre Pact for new facilities in water-stressed regions using potable water . Lei and colleagues, writing in Resources, Conservation and Recycling, found workload-level water use varies more than 10,000-fold, and ranked the determinants: server efficiency first, then grid water consumption factors, server utilisation, cooling system type, infrastructure efficiency and climate zone .
2.3 Waste heat recovery is real — but it is a siting decision
The evidence base for recovering data centre waste heat is mature. Huang and colleagues’ Applied Energy review framed data centres as prosumers within district energy systems, integrating renewable energy supply and reusing waste heat for district heating. Yuan and colleagues’ 2025 review in Renewable and Sustainable Energy Reviews evaluates the integration technically, energetically, economically, and environmentally, and identifies the binding constraints honestly: complex technical issues, economic limits, policy gaps, and missing infrastructure. Zhang and colleagues’ systematic review in Sustainability assesses recovery pathways — heat pumps, thermal storage, absorption cooling, and generation via the Organic Rankine Cycle, Kalina Cycle, and thermoelectric generators — and concludes that recovery can both lower PUE and deliver economic and emissions benefits.
The enabling technology is also proven. Wang and colleagues modelled high-temperature heat pump configurations capable of lifting data centre waste heat to supply temperatures up to 120 °C for district heating, using low-global-warming-potential refrigerants.
The honest reading for Australia is that the technology is not the gap. The off taker is. Australia has no equivalent of the Nordic legacy district heating networks that make these economics work, and the reviews are explicit that infrastructure gaps are among the primary constraints.
2.4 Digital twins earn their keep in operations, not in the pitch deck
Athavale and colleagues, writing in Computer, describe digital twins as living digital models of physical systems that enable data-driven analysis and the application of AI to manage selected aspects of the data centre and to drive efficiency for sustainability. In the microgrid domain, digital twin approaches are being developed for real-time control and energy management of distributed resources.
The link to everything else in this guide is direct. If you are obliged to demonstrate flexibility, to operate against a firming contract, and to report measured efficiency and water metrics, you need a calibrated model of the facility and its energy system to do any of it credibly.
2.5 Behind-the-meter assets are a network-aware design problem
The review literature on grid-aware operation of behind-the-meter assets sets out the objectives and constraints that govern how on-site flexibility resources behave in a distribution network, and identifies persistent gaps in prosumer-centric grid consideration and control strategy . The relevance is that on-site generation and storage do not simply subtract from your import; they interact with local network constraints, and their value depends on being operated in a way the network operator can plan around.
2.6 Efficiency is a supply-side strategy
Two of the reviews converge on a point that is easy to lose. Lei and colleagues rank server efficiency as the single largest determinant of workload-level water use — ahead of cooling system type and climate zone . Zhang and colleagues find waste heat recovery can significantly reduce PUE (Power Usage Effectiveness) while delivering economic and emissions returns . Efficiency measures reduce the size of the energy problem you have to solve, and in a constrained-connection market every megawatt you do not need is a megawatt you do not have to underwrite.
3. The energy-first development framework
The framework below reorders a conventional development sequence so that energy resolves before land is committed.
3.1 Site selection: the energy gate
Apply this as a pass/fail screen before any commercial land position is taken. A site that fails any Tier 1 test should not proceed to due diligence regardless of its other merits.
Tier | Test | What good looks like |
|---|---|---|
1 | Deliverable connection capacity and date | Written network advice on available capacity and an indicative energisation date consistent with your build programme. AEMO has named an approximate two-year target from application to energisation for transmission-connected projects. |
1 | Path to additional renewable supply | An identified, contractable source of new generation in the state of operation, plus a firming counterparty. Assume certificates must be linked to new build, not existing supported generation. |
1 | Water sufficiency | A water authority position, and a cooling architecture whose water demand matches it. Closed-loop liquid cooling removes the constraint almost entirely . |
2 | Planning pathway velocity | A dedicated or coordinated assessment route in the jurisdiction, with a known determination timeframe. |
2 | Thermal off-take adjacency | A credible industrial, agricultural or institutional heat consumer within economic pipe distance . |
2 | Co-location potential | Proximity to existing or committed generation, enabling a connection agreement that reflects it. |
3 | Community and social licence | A defensible local benefit position and a disclosure commitment on energy and water metrics . |
3 | Conventional criteria | Fibre, latency, land cost, seismic and flood, workforce access. Necessary, but no longer differentiating. |
3.2 What moves into base capex
The single most common costing error is treating the energy position as a sustainability overlay. Under an offset-and-firming obligation it is a condition of connection, which makes it base scope.
Item | Old treatment | Treatment under an offset obligation |
|---|---|---|
New renewable generation (PPA or equity) | Optional ESG procurement | Base scope; volume set by contracted load net of efficiency gains |
Firming capacity contract | Not carried | Base scope; a demonstrable contracting obligation |
Network augmentation contribution | Negotiated, partly socialised | Full cost of augmentation you cause or accelerate |
On-site storage sized for flexibility | Sized for ride-through only | Sized for ride-through plus contracted flexibility delivery [1, 13] |
Metering, telemetry and digital twin | Optional optimisation tool | Evidence infrastructure for flexibility and reporting obligations [11] |
Low-water cooling architecture | Cost-optimised against climate | Approval-critical; drives WUE more than operations do [5, 6] |
3.3 Sequencing
A workable order of operations for a greenfield Australian campus:
- Establish the energy proposition first — target load, efficiency assumptions, and the resulting offset and firming volume. This defines the project.
- Screen sites against the Tier 1 energy gate. Do not shortlist on land merit.
- Secure network advice and lodge the connection application early; treat the energisation date as the critical path anchor for the entire programme.
- Negotiate generation and firming in parallel with the connection process, not after it.
- Fix the cooling architecture against the water position and any thermal off-take, before detailed design.
- Design and instrument the flexibility envelope, and reflect it in the connection agreement.
- Commission the digital twin against as-built data and hand it over as an operating asset.
4.
4.1 Metrics that should appear in board reporting
Metric | Why it matters |
|---|---|
Measured PUE | Determines how much of your contracted energy does useful work; heat recovery improves it |
Measured WUE (L/kWh) | Approval and social licence exposure; benchmark against the 0.4 L/kWh threshold for water-stressed regions |
Contracted flexible MW and delivery rate | The evidence base for any flexibility concession in your connection agreement |
Offset coverage (%) | Share of annual consumption matched by additional new renewable generation |
Firmed capacity ratio | Contracted firming against maximum import capacity |
Connection milestone variance | Energisation date is the true critical path; variance here moves revenue |
5. What we deliberately do not claim
Two limitations are worth stating plainly, because the industry literature on this topic is unusually prone to overreach.
- Waste heat economics do not transfer from the Nordics to Australia. The peer-reviewed case rests on integration into existing district heating infrastructure, and the reviews identify infrastructure gaps as a primary constraint . Australian applications must be justified on a specific, contracted off-take, not on European system-level savings figures.
- Flexibility potential is not uniform. The reviews are explicit that magnitude and response time vary significantly by mechanism and facility type. Any single headline figure for “data centre flexibility” should be treated as unreliable until it is decomposed by mechanism for your specific facility.
6. Evidence base: peer-reviewed literature
[1] Crozier, C. & Liska, M. (2025). The Potential of Data Center Energy Demand To Provide Grid Flexibility. Current Sustainable/Renewable Energy Reports. https://doi.org/10.1007/s40518-025-00258-9
[2] Al Kez, D. & Foley, A. (2025). Programmable Load Risks and System Flexibility: Rethinking Data Center Participation in Modern Power Systems. https://doi.org/10.2139/ssrn.5395002
[3] Mytton, D. (2021). Data centre water consumption. npj Clean Water, 4, 11. https://doi.org/10.1038/s41545-021-00101-w
[4] Privette, A. P., Barros, A. & Cai, X. (2026). Data Centers Water Footprint: The Need for More Transparency. AGU Advances, 7, e2025AV002140. https://doi.org/10.1029/2025AV002140
[5] Cooper, A. & Nguyen, T. B. T. (2026). A Review of Thermal Management in Modern Data Centres: Water Usage Effectiveness and Heat Transfer Coefficients. Fluids, 11(8), 201. https://doi.org/10.3390/fluids11080201
[6] Lei, N., Lu, J., Shehabi, A. & Masanet, E. (2025). The water use of data center workloads: A review and assessment of key determinants. Resources, Conservation and Recycling, 219, 108310. https://doi.org/10.1016/j.resconrec.2025.108310
[7] Huang, P., Copertaro, B., Zhang, X., Shen, J., Löfgren, I., Rönnelid, M., Fahlén, J., Andersson, D. & Svanfeldt, M. (2020). A review of data centers as prosumers in district energy systems: Renewable energy integration and waste heat reuse for district heating. Applied Energy, 258, 114109. https://doi.org/10.1016/j.apenergy.2019.114109
[8] Yuan, X., Liu, J., Sun, S., Lin, X., Fan, X., Zhao, W. & Kosonen, R. (2025). Data center waste heat for district heating networks: A review. Renewable and Sustainable Energy Reviews, 219, 115863.
[9] Zhang, L., Zhao, Z., Chen, B., Zhao, M. & Chen, Y. (2025). Zero-Carbon Development in Data Centers Using Waste Heat Recovery Technology: A Systematic Review. Sustainability, 17(22), 10101. https://doi.org/10.3390/su172210101
[10] Wang, P., Kowalski, S., Gao, Z., Sun, J., Yang, C.-M., Grant, D., Boudreaux, P., Huff, S. & Nawaz, K. (2024). District heating utilizing waste heat of a data center: High-temperature heat pumps. Energy and Buildings, 315, 114327. https://doi.org/10.1016/j.enbuild.2024.114327
[11] Athavale, J., Bash, C., Brewer, W., Maiterth, M., Milojicic, D., Petty, H. & Sarkar, S. (2024). Digital Twins for Data Centers. Computer, 57(10), 151–158. https://doi.org/10.1109/MC.2024.3436945
[12] Digital Twin for Microgrid Control and Energy Management (2025). IEEE. Overview of digital twin application in microgrid control and management.
[13] Getie, E. M., Fani, H., Hashmi, M. U., Mbuwir, B. V. & Deconinck, G. (2025). Grid-Aware Flexibility Operation of Behind-the-Meter Assets: A Review of Objectives and Constraints. IEEE ISGT Europe 2025.
7. Primary policy and market sources
The following establish the Australian and Irish regulatory and market facts used in Sections 1 and 3. They are primary institutional documents, not peer-reviewed research.
AEMO (2026). 2026 Integrated System Plan, published 25 June 2026.
AEMO (2026). 2026 Electricity Statement of Opportunities, published August 2026.
AEMO (2026). Digital demand surge: Preparing Australia’s power systems for the rise of data centres — transmission connection pipeline disclosure, March 2026 quarter.
Oxford Economics Australia (2025 and 2026). Australia’s Data Centre Energy Demand, prepared for AEMO.
AEMC (2026). Data centre energy obligation — Advice to Energy and Climate Change Ministers, July 2026.
Commission for Regulation of Utilities, Ireland (2025). Large Energy Users Connection Policy, Decision Paper CRU/2025236, 12 December 2025.
Government of South Australia (2026). Data Centre and AI Infrastructure Strategy, June 2026.
NSW Government (2026). Data Centre Guidelines and Policy Framework, 17 August 2026; Electricity Infrastructure Investment Amendment Bill 2026.