A data centre site is chosen by the electricity network, not by the planning scheme. That is the single largest difference between this asset class and everything else a developer might build on industrial land. A warehouse needs a road, a slab and a permit. A data centre needs tens of megawatts of firm grid capacity in a specific part of a specific network, and if that capacity is not there, no amount of zoning, land price or design skill fixes it. Developers coming from industrial, residential or commercial backgrounds tend to run the checks in the order they are used to, and find out about the grid constraint after they are committed.
The second difference follows from the first. Almost every regulatory change happening in this sector right now is about who pays for the energy and water that a data centre consumes. New South Wales released a data centre policy framework on 17 August 2026. The Australian Government published national expectations for data centre developers in March 2026. The Australian Energy Market Commission gave advice to energy ministers in July 2026 recommending that data centres bring their own clean firmed energy and register as market participants. South Australia amended its planning regulations in September 2025 to require water and power certificates with a data centre application. The direction of travel is consistent across all of them, and it changes what a feasibility has to carry.
The figures, thresholds and dates in this guide were current when it was written in August 2026, and several of them are mid-reform. Where a number matters to your deal, the linked primary source is where to confirm it still stands. A great deal of what follows is draft policy, a bill before parliament, or a consultation that has not closed, and that is flagged as it comes up.
What makes a data centre different from any other industrial building?
The building is the cheap part. A data centre is a shell wrapped around an electrical and mechanical plant, and the plant is where the money and the risk sit. Substations, transformers, switchgear, uninterruptible power supplies, standby generators, fuel storage, chillers or evaporative cooling, and the redundancy that makes the whole thing survive a fault, all of it before a single server arrives. The floor area is almost incidental. Two buildings of the same 10,000 square metres, one designed for 5 megawatts of information technology load and the other for 20 megawatts, are not comparable projects. They have roughly the same concrete and roughly four times the electrical and mechanical content.
That is why the industry measures itself in megawatts rather than square metres, and why a cost per square metre benchmark carried across from a warehouse feasibility is close to meaningless. It is also why the constraint that decides whether a site works is a network constraint rather than a planning one.
The asset behaves like an income-producing property rather than a build-to-sell one. Revenue is generally expressed as a rate per kilowatt of contracted capacity per month, not a rate per square metre of lettable area, and the end value tends to come from capitalising that income rather than from comparable sales. If you have modelled a childcare centre or a logistics facility on a lease-backed basis, the shape is familiar. The unit of account is what changes.
Scale is worth calibrating. The Australian Energy Market Operator reports around 162 operational data centres nationally, concentrated in Sydney and Melbourne, accounting for roughly 2 per cent of grid-supplied electricity today. Its 2025 inputs and scenarios work projected consumption rising to around 12 terawatt hours, or about 6 per cent of grid-supplied electricity, by 2030, and around 34 terawatt hours, or about 12 per cent, by 2050. In New South Wales alone, the ministerial release accompanying the state’s new framework states there are 19 data centre projects valued at $50.3 billion in the state significant development pipeline, on top of more than 60 facilities already operating or under construction.
Why does grid capacity decide the site before the planning scheme does?
Because connection capacity cannot be bought off the shelf, cannot generally be reserved, and takes years to obtain. The Australian Energy Market Operator’s own reporting is the clearest statement of the timeframe. In its update on data centre integration, the market operator says that “current experience indicates that large data centre connections are targeting an approximate two-year timeframe from application to energisation, though timelines vary depending on project readiness and system conditions”. Two years from application, on top of whatever due diligence and design precedes the application, and that is the target rather than the outcome.
At the end of the March 2026 quarter, the market operator recorded 11 large-scale projects above 5 megawatts, representing 5.4 gigawatts of maximum demand, progressing through the transmission connection process, with around 60 per cent of that capacity in New South Wales and 40 per cent in Victoria. Most were in early stages. That concentration is the reason the constraint bites: a large share of the national pipeline is chasing capacity in two networks.
The practical consequence for site selection is that the grid question comes first. Before contract, before a planning consultant, arguably before a valuation. A site that is perfectly zoned, flat, serviced and cheap is worth very little as a data centre if the local network cannot deliver load, and the same site in a location where the transmission network already has headroom may be worth a multiple of its industrial land value. It is a highest and best use question answered by an electrical engineer rather than a planner.
How does the electricity connection process actually run?
Network service providers lead the connection process, and the market operator works alongside them on performance standards. The sequence generally runs from a connection enquiry, through a feasibility or capacity assessment by the network, into a formal connection application, then negotiated performance standards, a connection agreement, construction of the connection assets and commissioning. Each of those steps has its own cost and its own opportunity for the answer to change.
Location does more work than most developers expect. The market operator’s stated position is that coordinated planning to align new transmission investment with areas of growing demand, alongside developers considering locations in regions with abundant renewable energy resources, can help support lower-cost outcomes. Translated into deal terms: a site adjacent to spare transmission capacity, or near generation, may connect for a fraction of the cost and years sooner than a site that requires network augmentation. The augmentation cost is not a rounding error, and increasingly it lands on the proponent rather than being socialised.
The trap is treating the connection enquiry response as a commitment. It is generally an indicative assessment based on conditions at the time, and conditions in a network absorbing gigawatts of new load change.
What is changing in the national connection rules?
The Australian Energy Market Commission published a draft determination on 12 March 2026 proposing new technical standards for large data centres connecting to the National Electricity Market. Two elements matter to a developer. First, the draft rule would raise the threshold defining a large inverter-based load from 5 megawatts to 30 megawatts and embed that definition in the National Electricity Rules, so that the stricter technical requirements apply only to the connections most likely to affect system security. Second, facilities above the threshold would need to meet disturbance ride-through requirements, staying connected during specified voltage and frequency disturbances and recovering power within defined timeframes.
The reason is not abstract. The commission’s media release notes that “in July 2024, 60 data centres in the U.S state of Virginia pulled 1,500 MW off the grid simultaneously during a single fault”, causing cascading failures, and that similar incidents elsewhere prompted some jurisdictions to halt new data centre connections entirely. The commission’s stated intent is that the proposed standards would largely match those in use or proposed in Texas, Ireland and Finland, so that equipment and studies can be reused across markets.
This rule is not final. The commission’s project page for Improving the NEM access standards, Package 2 states that it has “extended the deadline to publish a final determination due to the complexity of the issues raised in response to the draft determination” and now plans to publish a final determination on 29 October 2026. A design being progressed on the assumption that the draft threshold and ride-through curves will be adopted unchanged carries real risk, and the market operator has said it is publishing interim guidelines to help proponents in the meantime.
Who pays for the network upgrades a data centre triggers?
The policy answer, in every jurisdiction that has published one, is the data centre. How that is enforced is still being built.
At national level, the Australian Energy Market Commission gave advice to energy ministers on 28 July 2026, published on 5 August 2026, recommending a package with four elements. Its advice project page sets them out as mandating that data centres offset consumption using certificates linked to new renewable generation under the Renewable Electricity Guarantee of Origin scheme, introducing a contract obligation to mandate contracting for firming capacity, introducing market registration requirements to improve visibility, and supporting demand flexibility and co-location with generation through connection agreements. The commission notes that “the recommendations are based on preliminary analysis and views, and further work is required to refine the options ahead of implementation”, and that the reforms could be staged and applied differently by size so that smaller facilities are not caught by the requirements designed for the largest connections. The Energy and Climate Change Ministerial Council considered the advice on 28 July 2026 and agreed to progress regulatory arrangements.
In New South Wales, the second pillar of the state’s framework is regulatory reform on cost recovery. The ministerial release states the government will make changes to “ensure that data centres pay for the electricity network upgrades they require, so that extra costs aren’t shifted to households and small businesses”, to support more coordinated planning of the network, and to support the guidelines. Legislation has been introduced and consultation on detailed implementation runs from 17 August to 14 September 2026 according to Infrastructure NSW.
For a feasibility, the practical reading is that connection and augmentation cost should be modelled as a real, project-specific, potentially very large capital line, not as a headworks allowance carried across from an industrial job. It should also be modelled as a range, because the number is not knowable at the point most developers want to exchange contracts.
What planning pathway does a data centre use in New South Wales?
New South Wales is the only jurisdiction with a purpose-built data centre planning framework, and it runs on a power-consumption threshold rather than a value threshold.
The Department of Planning’s own page on state significant development for warehouses and data centres sets out the pathways. It states that data centres with a total power consumption of more than 10 megawatts, or 15 megawatts from 1 June 2023, are state significant development, lodged with the department and determined by the Independent Planning Commission or the minister. Data centres that do not meet that criterion but have a capital investment value of more than $30 million are regionally significant development, lodged with council and determined by a Sydney district or regional planning panel. Anything below both is local development determined by council.
The department also confirms the land use definition. A “data centre” was added to the Local Environmental Plan Standard Instrument and means “a building or place the principal purpose of which is to collect, distribute, process or store electronic data using information technology”, and the definition of “high-technology industry”, itself a type of light industry, was amended so that data centres are a sub-type. The department lists the zones in which the transport and infrastructure policy makes data centres permissible with consent: B5 Business Development, B6 Enterprise Corridor, B7 Business Park, IN1 General Industrial, IN2 Light Industrial, IN3 Heavy Industrial, E3 Productivity Support, E4 General Industrial and E5 Heavy Industrial, or equivalent zones, with local plans able to make them permissible elsewhere.
The design consequence of a power threshold rather than a floor area threshold is worth sitting with. A facility designed for 14 megawatts and one designed for 16 megawatts can occupy an identical building on an identical site and sit in completely different assessment regimes, with different consent authorities, different documentation and different timeframes. That is a design decision with a planning consequence, and it should be made deliberately rather than discovered. Our guide to the state significant development pathway in New South Wales covers what the higher pathway involves in practice.
What does the NSW Data Centre Policy Framework require?
The framework was announced on 17 August 2026 and is policy plus guidelines, not a change to the assessment threshold. It has three pillars: the NSW Data Centre Guidelines supporting planning assessment, regulatory reform on energy cost recovery, and a review by the Independent Pricing and Regulatory Tribunal of New South Wales into water pricing for data centres.
The guidelines set performance measures built around six principles, which Infrastructure NSW lists as applying world-class environmental and efficiency standards, imposing no net cost to consumers and communities, funding additional supply of water and energy, enhancing local community infrastructure and amenity, investing in future industries across the supply chain, and demonstrating a commitment to training and skills to support jobs.
The incentive is speed. The ministerial release states that projects meeting the performance measures “will receive a commitment to have their proposed developments assessed within 75 days, while still meeting all planning, environmental and community consultation standards”, supported by concurrent changes to Independent Planning Commission processes. Infrastructure NSW also states the guidelines will be reviewed annually.
Three things follow for a developer. The 75 days is an assessment commitment, not a total approval timeframe, and it sits after preparation, environmental assessment requirements and exhibition. The performance measures carry cost, because funding additional energy and water supply and enhancing local infrastructure are capital commitments that need to sit in the feasibility rather than in a good-intentions section of the application. And because the guidelines are reviewed annually, a project with a long lead time may be assessed against a version that does not exist yet.
What planning pathway applies in Victoria?
Victoria treats a data centre as a discrete land use term nested inside utility installation, and offers a ministerial pathway with a published cost threshold.
In the Victoria Planning Provisions, clause 73.03 lists “Data centre” as a land use term included in “Utility installation”. The data centre term itself has no definition in the table, and clause 73.03 provides that a term listed without a definition “has its ordinary meaning”. The nesting is what matters practically: a data centre picks up the permission status of utility installation in each zone’s land use table, rather than having its own row. Utility installation is defined as land used, among other things, “to transmit, distribute or store power” and “to collect, treat, transmit, store, or distribute water”, which does not obviously describe a data centre, so the characterisation rests on the inclusion rather than the definition. That is worth a planner’s view early rather than an assumption.
The faster route is the Development Facilitation Program. The department’s expedited planning pathways guidance lists “Digital technologies” as a sector, with “Data centre” as the land use, and thresholds of $10 million estimated development cost in regional Victoria and $20 million in metropolitan Melbourne. Notably, the same table carries a separate sector for “Utility installation (other than a data centre)”, which confirms the department treats data centres as a carve-out rather than as ordinary utility infrastructure.
Two features of the pathway change the risk profile. The department states that the estimated development cost “must be supported by a report prepared by a suitably qualified quantity surveyor”, so the threshold is evidenced rather than asserted. And decisions made by the Minister for Planning under the provisions cannot be appealed to the Victorian Civil and Administrative Tribunal, which removes third-party merits review risk from the timeline. Other than for state or public authority projects and renewable energy facilities, requests must be supported by written advice from the Chief Executive Officer of Invest Victoria confirming the project has demonstrated investment certainty or capacity to secure funding, which is a gate worth knowing about before building a programme around the pathway. Our guide to the Development Facilitation Program covers how the wider programme operates.
How do the other states and territories treat data centres?
South Australia has gone furthest after New South Wales, and in a direction other states may follow. PlanSA states that amendments to the Planning, Development and Infrastructure (General) Regulations 2017, made on 11 September 2025, prescribe that data centres “can now, if sponsored, be assessed as Crown development under section 131 of the PDI Act”, meaning the Planning, Development and Infrastructure Act 2016 (PDI Act), and that where a data centre application is lodged under the general scheme, the State Commission Assessment Panel is the relevant authority. The requirement that should interest developers most is documentary. Applications for data centres must now be submitted with “advice from SA Water that there is sufficient water supply to meet the requirements of the data centre” and “a certificate from the Technical Regulator confirming that it complies with their requirements relating to the reliability, security and stability of the state’s power system”. South Australia has converted the two questions that decide the project into lodgement prerequisites. See the PlanSA note on the regulation changes for the wording.
Western Australia has no data-centre-specific policy, but the significant development pathway is available on value. The Department of Planning, Lands and Heritage states that the Part 11B pathway “came into effect on 1 March 2024” and that “a ‘significant development’ in Part 11B is one valued at $20 million or more in the Perth and Peel regions and $5 million or more everywhere else”, determined by the Statutory Planning Committee of the Western Australian Planning Commission with a 120-day statutory timeframe subject to stop-the-clock provisions, a mandatory pre-lodgement meeting and a minimum 28-day public consultation period. Details are on the significant development pathway page. Most data centre projects would clear the value threshold comfortably.
Queensland has no published data centre policy that could be located, and no data-centre-specific land use term could be verified. Two state pathways may be relevant depending on scale. Ministerial infrastructure designation under the Planning Act 2016 provides, in the department’s words, “an alternative to lodging a development application with local government”, though the examples the department publishes are hospitals, schools, emergency services facilities, power lines and water and sewerage infrastructure, and data centres are not among them. Separately, the Office of the Coordinator-General administers coordinated projects, prescribed projects and state development areas under the State Development and Public Works Organisation Act 1971, and describes coordinated projects as “large-scale developments requiring rigorous impact assessment”. Whether either route is available to a private data centre is a question for a Queensland planner rather than an assumption.
Tasmania’s projects of state significance route, administered by the Tasmanian Planning Commission under the State Policies and Projects Act 1993, has an unusual feature worth knowing: the commission states that for a declared project, “the requirements of the planning scheme or any other Act do not apply”, enabling a fully integrated assessment. A proposed project must meet at least two of seven declaration criteria including significant capital investment and significant infrastructure requirements, and a declaration order must be approved by both houses of parliament. No data centre has been declared to date, and the process is slow by design.
For the Australian Capital Territory and the Northern Territory, the land use definitions and zone tables are published only as documents that could not be verified for this guide. Both are small markets for this asset class, but the characterisation question should be checked directly rather than inferred from another jurisdiction.
What does the Australian Government expect of data centre developers?
The Department of Industry, Science and Resources published Expectations of data centres and AI infrastructure developers on 23 March 2026. It is not law. The document states that “the expectations work alongside existing national, state and territory laws” and “do not alter any obligations to comply with Australian law”. What it does carry is prioritisation. The document states that “the Australian Government will prioritise proposals most closely aligned with the expectations” and that “energy-intensive data centre proposals not closely aligned with the expectations will not be prioritised by Commonwealth regulatory assessments”.
The scope excludes small facilities. The expectations apply to “new or expanded developments within Australia”, with co-location sites, hyperscale operations and large-scale artificial intelligence compute centres asked to consider how they apply, and expressly do not apply to “small-scale edge or on-site enterprise data centres”.
Five expectations are set out: prioritising Australia’s national interest including national security and data sovereignty; supporting the energy transition, including securing “new and additional clean energy generation and/or storage to offset demand” and covering “their share of transmission and distribution infrastructure costs”; sustainable and efficient water usage, including using non-potable water where possible and providing “ongoing transparent reporting about their water usage and efficiency”; investment in Australian skills and workforce; and research, innovation and local capability.
One line is directed squarely at speculative land banking dressed as development: “Non-genuine data centre proposals that congest approval pathways are not welcome.” The document also states the government will work with states and territories to implement the expectations through their processes, particularly through the Energy and Climate Change Ministerial Council, which is what the New South Wales framework and the market commission’s ministerial advice are doing.
How much water does a data centre need, and who pays for it?
Cooling is the water question, and it is now a pricing and approval question as well as an engineering one. The design choice between evaporative cooling, which uses a lot of water and less electricity, and closed-loop or air-cooled systems, which use far less water and more electricity, is a trade between two constrained resources, and both are being repriced.
In New South Wales, the third pillar of the framework is a review by the Independent Pricing and Regulatory Tribunal of New South Wales. The ministerial release states the government is commissioning the tribunal to review the water pricing framework for data centres, and that “the review will consider how water pricing can reflect the full costs associated with servicing data centres, ensuring other water users on the network are protected and the impacts of drought and water scarcity are appropriately managed”. The Minister for Water’s stated position is that the guidelines “direct data centres to use water efficiently, prioritise recycled water where possible, and contribute to the additional water supply needed to support future growth”.
South Australia has already made it a lodgement gate, requiring advice from SA Water that there is sufficient supply before an application is accepted. Other jurisdictions have not, yet.
For a feasibility, that means water should carry two lines, not one: a servicing and headworks capital cost that may be materially higher than an industrial equivalent, and an operating cost exposed to a pricing framework under active review. A twenty-year hold model built on today’s published water tariff is modelling a number that a regulator has publicly announced it is reconsidering.
What else can stop a data centre site?
Noise from cooling plant is the constraint most often underestimated, because a data centre runs continuously and the assessment that bites is the night-time one. In New South Wales the relevant instrument is the Environment Protection Authority’s Noise Policy for Industry, which sets both an intrusiveness level, controlling emergence above background at residences, and an amenity level, controlling overall noise by receiver type. A large array of chillers or dry coolers running at 3am next to a residential interface is a genuine design constraint, and the mitigation, whether acoustic screening, plant selection or setback, costs money and floor area. The site selection lesson is that an industrial site backing onto residential is a different proposition for a data centre than for a warehouse that shuts at 6pm.
Security of critical infrastructure obligations may attach to the completed asset. Section 12F of the Security of Critical Infrastructure Act 2018 (Cth) defines a critical data storage or processing asset. The test turns on who the end-user is and what data is involved, and the section sets out the categories that bring an asset in, so whether a particular facility is caught is a question to put to the section itself rather than to a rule of thumb. Where it applies, obligations may include registration, a risk management program and cyber incident reporting, and the Cyber and Infrastructure Security Centre publishes what attaches to the sector. This is an operator obligation rather than a developer one in most structures, but it shapes what a tenant will require of the building, and it is a question for the lawyer advising on the lease or operating agreement rather than something to design around blind.
The ordinary industrial due diligence still applies and still kills sites: contamination, flood, easements, aviation and telecommunications interference, and access for very heavy and very large plant deliveries. Fibre connectivity, and specifically diverse physical fibre routes rather than a single carrier’s presence, is a genuine site test that has no equivalent in industrial development.
Why does a data centre feasibility break a normal development model?
Because the denominator is wrong. A model built around square metres, whether of gross floor area or lettable area, will misprice a project where cost and revenue both scale with electrical capacity rather than floor plate. Gross floor area still matters for the planning application, but it is not what drives the numbers.
What sits in the cost stack, and why is cost per square metre misleading?
There is no reliable published Australian benchmark for data centre construction cost per megawatt. That is not an evasion, it is the state of the evidence. Government building approvals data does not separate the category, and no professional body publishes a figure that could be verified for this guide. Anyone quoting a confident dollar-per-megawatt range is quoting either overseas data or their own project experience, and both may be some distance from what your design in your location with your redundancy level actually costs.
What can be said is how the stack is structured. Land and site works behave much like industrial. Shell and core, the concrete, steel, roof and slab, is a relatively modest share. The electrical and mechanical fit-out, meaning the substation and high-voltage works, switchgear, uninterruptible power supplies, standby generation and fuel, and the cooling system, tends to dominate, and it scales with megawatts and with the redundancy tier the tenant requires. Grid connection and augmentation sits outside the building contract entirely and is negotiated with the network. Long-lead equipment procurement, particularly transformers and switchgear, has its own price and programme risk that a standard construction contingency was not sized for.
The honest position for a developer is that the quantity surveyor’s cost plan against a specific design, at a specific redundancy level, with current procurement lead times, is the only number worth putting in a feasibility. A rate carried from another project or another country is a placeholder. Build the total development cost from the plan, and carry the connection cost as a separate, explicitly uncertain line rather than burying it in headworks.
How is revenue expressed?
Generally as a rate per kilowatt of contracted capacity per month, applied to the critical information technology load the facility can deliver, rather than as a rent per square metre. Contracts tend to be long and are often signed before or during construction, which changes the risk profile substantially compared with a speculative warehouse. In a wholesale or hyperscale arrangement, a single counterparty may take the entire facility.
The practical implication for modelling is that the revenue line is driven by contracted megawatts, the rate, and the ramp at which capacity is taken up, not by a lease-up curve measured in floor area. Where the end value is derived by capitalising stabilised income, the capitalisation rate applied to that income does more work in the feasibility than almost any other input, and a small movement in it may swamp a large movement in build cost.
What does staging do to the cashflow?
It changes the shape of the model more than anything else. Data centres are typically built and energised in stages, hall by hall, as capacity is contracted, and the market operator notes that large facilities “typically ramp-up over 5-10 years” and incorporates that into its forward outlooks. A model that assumes practical completion followed by full revenue is not describing the asset.
That means the cashflow model needs to carry staged capital expenditure, staged energisation, staged revenue and a debt structure that survives a long period of partial income. It also means the peak funding requirement may occur well after the first hall is operating, which is a different profile from a build-to-sell project where peak debt sits just before settlements begin.
What does a data centre developer actually sell?
There are broadly three exits, and they demand different things from the project.
The first is powered land, sometimes called a powered shell. The developer acquires the site, secures the planning approval and, critically, secures the grid connection, then sells the approved and connected site to an operator who builds the facility. This is the shortest capital cycle and the one where the developer’s value creation is almost entirely the connection and the approval rather than the building. It also fails completely if the connection cannot be secured, which is why the grid work has to be done before the land is committed rather than after.
The second is develop and lease, then hold or sell the stabilised asset. The developer builds to a tenant’s specification under a long agreement for lease, and the completed value comes from capitalising the contracted income. The lease structure does a great deal of work here, particularly on who bears power cost, who bears the risk of a network tariff change, and what happens if the tenant’s capacity requirement changes.
The third is develop and operate, which is a different business rather than a different exit. It carries operating risk, customer risk and technology obsolescence risk, and most property developers who go this way do so with an operating partner.
The choice made at the front end determines what has to be true at the back end, and it should be made before the site is bought, because the powered-land exit and the develop-and-lease exit justify very different land prices for the same parcel.
How does data centre development work in New Zealand?
New Zealand has no data-centre-specific planning instrument, and the regime it does have is in the middle of being replaced. Data centres are assessed as ordinary land use under district plans through the resource consent process, with the added complication that plan making is largely frozen.
The Ministry for the Environment’s National Policy Statement for Infrastructure was made on 15 December 2025 and came into force on 15 January 2026. Whether a data centre gets its benefit is genuinely unclear: the statement’s list of additional infrastructure covers schools, health facilities, fire and emergency services, defence, corrections, stormwater networks, resource recovery and flood control works, and does not name data centres, so coverage would turn on the Resource Management Act 1991 definition of infrastructure. That is a question for a New Zealand planner, not an assumption to build a programme on.
The plan stop matters more in the near term. The ministry’s explanatory note states that an August 2025 amendment “suspends the requirement to review plans and policy statements and prevents notification of new plan or policy statement changes or variations until the end of 2027, when the new resource management system will be in effect”, and that consequently “plans and policy statements will not be updated to give effect to this National Policy Statement”. In practice, a developer is working with the district plan as it stands. Our guide to resource consent in New Zealand covers the process itself.
The Fast-track Approvals Act 2024 is the alternative route. The ministry states that the Act “establishes a permanent fast-track approvals regime for a range of infrastructure and development projects” and includes 149 projects in Schedule 2 that can apply directly to an expert panel. Data centres are not a listed sector, but the regime is not closed to them: an unlisted proponent may apply to the Minister for Infrastructure for referral, and combined energy and data centre proposals have used that route.
On the electricity side, Transpower’s published position is unusually clear and directly useful. Its grid connection process page states that “the New Zealand power system is run on open access principles” and that “you cannot buy or reserve transmission capacity on the transmission system”. It states that “most small assets (<40 MW) choose to connect via the local distributor’s network as transmission level connections tend to be more expensive”, while “large assets (>50 MW) will usually need to connect directly to Transpower’s assets (e.g. a substation) if there is capacity”, and that establishing a new greenfield connection may cost “upward of $20 million”. On timing, it states that a small, discrete and simple connection may be running “in fewer than 18 months”, while “large-scale complex connections may take longer than three years to commission”, with a standard concept assessment turnaround of 35 to 45 working days. Transpower also states plainly that “deciding where to connect to the national grid will have the most significant impact on connection cost”, which is the same site-selection lesson as the Australian market, stated by the network itself.
For scale, Transpower’s October 2024 release on its future demand work noted that data centres “are also expected to contribute to increased demand by as much as 5,000 GWh”, comparable to the New Zealand Aluminium Smelter. That figure is now some years old and should be treated as an indication of magnitude rather than a current forecast.
The wider reform is unfinished. The Planning Bill and Natural Environment Bill, which together replace the Resource Management Act 1991, were reported back by the Environment Committee on 20 July 2026 and had not passed as at August 2026.
What is still moving, and when?
More of this sector is draft than settled, which is unusual and worth stating plainly.
The Australian Energy Market Commission’s access standards rule, including the proposed 30 megawatt threshold and the ride-through requirements, is a draft. The commission’s project page gives an expected final determination of 29 October 2026. Until then the applicable technical requirements for a connection are the current rules plus whatever interim guidelines the market operator publishes.
The commission’s advice to energy ministers on renewable offsetting, firming contracts and market registration is advice, not rules. The commission itself says further design work is needed and that governments will determine implementation through legislation, new electricity rules, or both.
In New South Wales, the guidelines are live and the 75-day assessment commitment applies now, but the energy cost recovery reforms sit in a bill and a consultation open until 14 September 2026, and the water pricing review by the Independent Pricing and Regulatory Tribunal of New South Wales has been commissioned rather than completed. The guidelines are reviewed annually.
South Australia’s requirements are in force. Victoria’s pathway is in force. Queensland, Western Australia, Tasmania, the Australian Capital Territory and the Northern Territory have no data-centre-specific policy published as at August 2026, which does not mean they will not, given the Commonwealth’s stated intent to work through the Energy and Climate Change Ministerial Council.
A project with a three to five year lead time is going to be assessed under rules that do not exist yet. The sensible response is to model the cost of compliance with the direction of travel, which across every published document is that the developer funds the additional energy and water supply it consumes, rather than modelling today’s rules and treating the change as upside risk.
What to ask your planning consultant, electrical engineer and quantity surveyor
The questions below are the ones that decide a data centre project on your specific facts, and that this guide deliberately does not answer.
For the electrical engineer or energy adviser, before you commit to the land:
- What is the realistic firm capacity available at this connection point today, and what augmentation would be required to reach our target load?
- What is the network’s current indicative timeframe from connection enquiry to energisation for a load of this size in this part of the network, and what has that timeframe actually been on recent comparable connections?
- What proportion of the augmentation cost would fall on us as proponent under the current arrangements, and how would that change if the New South Wales cost recovery reforms or the national package proceed as proposed?
- Would this facility fall above or below the proposed 30 megawatt large inverter-based load threshold, and what would compliance with the draft ride-through standards cost in equipment terms?
- If the final rule differs from the draft, what in our design would have to change, and at what point in the programme does that become expensive?
- What are the current lead times on the transformers, switchgear and generation we would need, and what does that do to our programme?
For the planning consultant:
- Which pathway does our design fall into, and how close are we to the threshold that would move us into a different one?
- Is the site’s zoning one where a data centre is permissible with consent, and does the local instrument change that answer?
- What would meeting the state’s data centre performance measures require of us in capital terms, and is the accelerated pathway worth that cost on this project?
- What is the realistic total timeframe from now to consent, including pre-lodgement, assessment requirements, environmental assessment preparation and exhibition, rather than the assessment period alone?
- What is the noise exposure at the nearest sensitive receivers, and what mitigation would the relevant policy require of continuously operating plant?
- Are there third-party appeal rights on this pathway, and what is the realistic delay if they are exercised?
For the quantity surveyor:
- What is the cost plan for this specific design at this redundancy level, split between shell and core, electrical, mechanical and site works?
- How does the cost move if the redundancy level changes, and which components does that actually affect?
- What contingency is appropriate given current procurement lead times on long-lead electrical equipment, and is that different from the contingency you would normally carry on an industrial job?
- How should we treat the grid connection and augmentation cost, given it is not part of the building contract and is not fully known at this stage?
- What is the cost difference between the water-intensive and the electricity-intensive cooling options on this design, and how does that interact with the water and energy commitments we are being asked to make?
For the lawyer, on the lease or agreement for lease:
- Who bears the risk of a change in network tariffs, water pricing or a new statutory obligation on the operator during the term?
- What happens to the agreement if the connection is delayed beyond the target energisation date, and who wears that?
- If the tenant’s capacity requirement changes, what obligation does that create for us?
- Do the security of critical infrastructure obligations attach to us as owner, to the operator, or to both, and how is that allocated in the documents?