Artificial intelligence may feel invisible, but the infrastructure behind it uses land, electricity, cooling equipment, grid capacity and, in some cases, substantial quantities of water. Before more agreements are signed and developments approved, the public should be able to see the full case.
Artificial intelligence can feel almost effortless. A person types a question into a box and an answer appears within seconds, creating the impression that the technology exists somewhere distant.
Unfortunately, the reality is more physical. AI runs inside buildings filled with servers, cables, cooling systems, backup generators, security equipment, and other industrial infrastructure. Those buildings need land, roads, fibre, electricity connections, transmission capacity and, depending on how they are designed and operated, access to water and wastewater systems.
New Zealand is already part of this expansion. Microsoft opened its New Zealand cloud region in December 2024, Google has announced plans for a New Zealand cloud region, and government procurement arrangements make it easier for eligible public agencies to purchase services from large international cloud providers. At the same time, new data-centre developments are being proposed or built in several parts of the country.
These developments are often described through the language of digital transformation, innovation, foreign investment, and economic growth. Those benefits may be real, but the language can make the change sound simpler than it is.
Every large data centre creates a series of decisions about physical resources, public infrastructure, environmental effects, financial risk, and long-term dependence on private technology companies. Those decisions are often made separately, by different organisations, under different laws, and with different pieces of information.
In Part One of this series, we looked at Local Water Done Well and asked who New Zealand’s changing water system is being built to serve.
In Part Two, we examined how decisions about cloud procurement, overseas investment, electricity, water, land use, council infrastructure, and public support can be made through separate processes without anyone presenting the whole public case.
In Part Three, we ask the practical question that should be answered before the next major data centre is approved or supported:
What should everyday New Zealanders be told before shared resources and public infrastructure are committed?
The same development can look different from every seat at the table
A large data centre does not look the same to everyone involved because each organisation and community sees the part closest to its own responsibilities.
To a minister, the development may represent foreign investment, economic activity, and evidence that New Zealand is open for business. To a technology company, it may represent a strategic location, a growing market, and long-term demand for cloud and artificial-intelligence services.
An electricity generator may see a large new customer that could support investment in new generation, while a network company may see the need for substations, transmission lines, and other capacity upgrades. A council may see construction activity, new rates or charges, pressure on local infrastructure, and a planning application that must be assessed under existing rules.
Businesses may see faster cloud services and new commercial opportunities, while residents living near the site may experience construction traffic, noise, changes to the landscape, and uncertainty about water or electricity demand.
A ratepayer may reasonably wonder whether council infrastructure will have to be expanded and who will pay for it. An electricity consumer may ask whether the developer will meet the full cost of the new capacity it requires, or whether some of that cost will eventually appear in household power bills.
Government agencies may see a faster and easier way to buy digital services, while the people whose information is being stored may have questions about privacy, control, foreign law, cybersecurity, and whether an essential public service could move to another provider later.
None of these perspectives is automatically wrong. The problem begins when one of them is presented as though it represents the complete public interest.
A development can attract international investment while still creating substantial local infrastructure costs. It can improve access to cloud services while increasing New Zealand’s dependence on a small number of global providers. It can create hundreds of jobs during construction while employing far fewer people once it is operating.
It can also use very little drinking water under its expected operating model while holding permission to use much more if circumstances change.
That is why good public decision-making must do more than gather separate technical assessments. It must show how the benefits, costs, risks, and responsibilities connect.
The Technology May Be Digital, But The Demand Is Physical
A major data centre may require a substantial electricity connection, new generation, transmission and distribution upgrades, substations, fibre links, road access, cooling equipment, backup generation, emergency planning, ongoing security, and systems for dealing with stormwater, wastewater, and heat.
The exact requirements will differ between developments, which is why general statements about innovation or investment are not enough.
A community needs project-specific information showing what one facility will require from the electricity system, water resources, council infrastructure, and surrounding environment.
Transpower has said data centres could add as much as 5,000 gigawatt-hours to future electricity demand, an amount similar to the annual demand of the New Zealand Aluminium Smelter at Tiwai Point. New Zealand’s total annual electricity demand is currently close to 40,000 gigawatt-hours, which shows why these projects cannot be treated as ordinary commercial connections.
New Zealand is already trying to electrify transport, industrial heating, and other parts of the economy. The population is growing, existing electricity infrastructure requires investment, and households already face concerns about affordability and security of supply.
Data centres will join that demand, which means the public needs to know whether developers will fund the infrastructure required for their operations, whether other consumers could carry part of the cost, and what obligations will apply during periods when electricity is scarce.
It also means that claims about renewable electricity need to be considered carefully. A company may purchase renewable energy or renewable energy certificates, but the wider public question is whether its additional demand requires new generation and whether the electricity system can meet that demand without increasing costs or relying more heavily on fossil-fuel generation at other times.
Not Every Data Centre Uses Water In The Same Way
Some data centres use substantial quantities of water for cooling, while others rely more heavily on outside air, closed-loop systems, recycled water, or designs intended to reduce direct water use.
The right questions are specific to the development being proposed.
The public should be told what cooling system will be used, how much water the facility expects to use during ordinary and peak operation, where the water will come from, what will happen to it afterwards, and what demand could arise if the facility expands or the cooling system changes.
People should also be able to see who reviewed the developer’s estimates, what happens during droughts or restrictions, and whether the consent conditions remain appropriate if the climate or operating model changes.
Those questions should be answered while the proposal is still being considered, rather than after a facility has been built and the major commitments have already been made.
Southland Shows Why the Consented Maximum Matters
The proposed Datagrid development near Makarewa provides a real example of why the public needs to see both what a developer expects to use and what it has received legal permission to use.
Environment Southland granted a group of resource consents associated with the proposed data centre in March 2026. Those consents cover groundwater, stormwater, wastewater, discharges to air, and effects on a wetland.
The proposed campus covers approximately 49 hectares, with Datagrid describing plans for a 280-megawatt hyperscale AI facility. The company says the development will use Southland’s cool climate and access to renewable electricity to support large-scale computing.
One of the resource consents allows Datagrid to take up to 604,800 litres of groundwater each day. That is the maximum legally permitted amount rather than proof that the facility will use that volume every day, but it remains a significant allocation of a shared natural resource.
Datagrid has said that it does not expect to draw groundwater during normal operation because Southland’s cool temperatures will allow the facility to rely mainly on outside air for cooling. The company says the limited water needed during warmer conditions would be supplied through rainwater harvested from the roofs of the data-centre buildings.
That explanation may prove accurate, and Southland’s climate may allow the facility to operate with very little direct water use.
However, the difference between what the company expects to use and what the consent legally allows is precisely why the public needs clear information and strong monitoring.
If Datagrid expects not to use groundwater, people are entitled to ask why permission was sought for more than 600,000 litres a day, what circumstances could lead to that permission being used, and whether the community would be told before groundwater extraction began.
They are also entitled to know how frequently actual water use must be recorded and reported, whether the consent allows future expansion, and what would happen if the technology, cooling system, ownership, or commercial purpose of the development changed.
The question is not whether Datagrid currently intends to use the maximum amount. The question is whether the consent conditions are designed for the full range of activity that has been approved over the life of the project.
A Consented Maximum Still Matters When The Expected Use Is Lower
Developers are right to say that a resource consent often provides an upper limit rather than a forecast of ordinary daily use.
A consented maximum may allow for unusual weather, backup arrangements, future flexibility, or conditions that arise only occasionally. It should not be described as though the maximum volume will automatically be used every day.
However, the maximum still defines what the holder is legally permitted to do.
When the authorised volume is more than 600,000 litres a day, the public should not be expected to dismiss it simply because the company’s present design anticipates using less.
The transparent way to describe the position is that Datagrid is permitted to take up to 604,800 litres of groundwater a day, while the company says its expected operating model will rely on Southland’s cool air and harvested rainwater and will not require groundwater during normal operation.
The unresolved issue is why that level of permission was required, what conditions would lead to its use, how actual demand will be independently monitored, and whether the consent can respond if the development expands or changes.
That is not an argument against the company. It is an argument for a public record that clearly distinguishes between what is expected, what is permitted, and what could happen later.
The Warning From a Water Expert
Engineer Evan Vaughters, who chairs a climate-change research group at Water New Zealand, has warned that New Zealand’s resource-management framework may not be equipped to assess the long-term effects of hyperscale data centres.
Speaking to RNZ’s Nine to Noon, Vaughters said the volume permitted under the Datagrid groundwater consent was comparable to the amount of water used by a town the size of Kaikōura. He questioned whether the consent conditions provided the monitoring needed to understand long-term consequences and asked why such a large groundwater permission was required if the company did not expect to use it.
His wider concern was about resource allocation. A developer may be able to demonstrate that one proposed groundwater take will not create an unacceptable local effect, while the larger question of whether this is the best use of finite water and electricity resources sits outside the narrow assessment.
Vaughters also argued that regulatory systems often plan for the environment of the past rather than the environment of the future. That matters because consents can last for decades, while climate conditions, rainfall, technology, electricity demand, and commercial plans may all change during that time.
This is the gap that New Zealand needs to confront.
A consent process may determine whether one activity meets the relevant legal test at one site. It may not decide whether several individually acceptable developments will collectively place too much pressure on a water source, an electricity network, or a region’s infrastructure.
It may also be unable to answer whether a proposed use of water and electricity is the best long-term choice for New Zealand, because no one consent authority has been given responsibility for making that wider decision.
We Are Approving Infrastructure For a Changing Climate
New Zealand’s planning and resource-management systems were not designed around the demands of hyperscale artificial-intelligence infrastructure.
These developments combine very high electricity demand with questions about water, heat, land, transmission capacity, overseas ownership, public procurement, digital dependency, and possible future expansion.
Each issue may fit inside a different regulatory box, but the combined effect does not sit within only one of them.
The environment will also not remain fixed during the life of these facilities. Rainfall patterns may change, droughts may become more frequent or severe, and competing demand for water and electricity may grow.
For that reason, approvals should include ongoing monitoring, clear reporting, review conditions, and a process for reconsidering resource use where the development changes materially.
Local Water Done Well does not answer the allocation question
Part One of this series examined Local Water Done Well and asked who New Zealand’s future water system is being built to serve.
That reform is largely concerned with how drinking water, wastewater, and stormwater services are governed, financed, and delivered. It does not by itself provide a national framework for deciding how large industrial users should access groundwater, how competing demands should be prioritised, or what information communities should receive before a major development is approved.
Electricity May Be The Larger National
Issue
The Southland debate has focused heavily on water, but electricity may become the larger national constraint.
Datagrid describes its proposed Makarewa campus as having up to 280 megawatts of hyperscale capacity. A development of that size would be an industrial electricity user on a scale that very few New Zealand sites approach.
Transpower’s modelling shows that data centres could collectively add up to 5,000 gigawatt-hours of demand. That is significant in a country whose current annual electricity consumption is close to 40,000 gigawatt-hours.
The public therefore needs more than an assurance that the electricity will be renewable.
People need to know what new generation will be required, which transmission and distribution upgrades will be needed, who will pay for them, and whether the development will make electricity more expensive or less reliable for other users.
They also need to know what happens if a developer reserves substantial network capacity and then delays, reduces, or cancels the project.
These are questions about how a shared and essential system should be managed when one new user may require as much electricity as a large industrial operation or an entire region.
The Benefits May Be Real, But They Need To Be Measured
Large scale digital infrastructure may bring genuine benefits to New Zealand. It could Improve access to cloud and AI services, provide additional computing capacity, support research, create construction work, and allow some information and services to be hosted closer to home.
Southland may also benefit from investment, contracts for local companies, training opportunities, and specialist employment.
Those are reasonable arguments in favour of development, but “possible” benefits are not the same as demonstrated public value.
Before approval or substantial public support is committed, the public should be told how many jobs are expected during construction, how many permanent roles will remain afterwards, what skills and pay levels those jobs will involve, and how much local procurement is expected.
People should also be able to see whether the facility will mainly serve New Zealand organisations or international customers, what tax is expected to be paid here, and how the claimed regional and national benefits will be measured after the facility opens.
Company announcements explain what a development hopes to achieve. Public reporting is needed to show what it eventually delivers.
Who is Planning New Zealand’s AI future?
There does not appear to be one public organisation responsible for planning the full impact of New Zealand’s AI infrastructure and that is of concern to us.
Investment agencies promote opportunities to international companies. Cabinet and ministers set policy across digital services, energy, water, local government, and overseas investment.
Councils make decisions about land, roads, and local infrastructure. Regional councils assess groundwater and environmental effects. Electricity generators, distribution companies, and Transpower consider supply and connections.
Government agencies buy cloud services, while private companies decide what they want to build and how quickly they may expand.
Each organisation sees one part of the development. The public lives with the combined result.
The problem is not necessarily that one organisation has failed to do its job, it’s that no organisation appears responsible for bringing the whole picture together and putting it before New Zealanders while there is still time to influence the outcome.
Reporting Should Continue After Approval
Transparency should not end when a consent is granted.
Major data centres should report annually on their actual electricity and water use, emissions, backup generation, permanent employment, public assistance, environmental conditions, and progress against the benefits used to support the project.
The public should be able to compare the original promise with what happened.
What We Are Asking For
Lobby for Good is asking the Government to establish a national disclosure and assessment framework for hyperscale data centres before more major developments are approved or substantially supported.
That framework should require one public infrastructure account for every major proposal, independent review of water and electricity estimates, clear disclosure of public support, and annual reporting after approval.
It should also require cumulative demand to be considered, rather than assessing every project as though it exists alone.
Help Us Build The Public Record
The information needed to understand these developments is spread across council files, resource consents, electricity reports, procurement agreements, ministerial announcements, and company material.
Most everyday New Zealanders do not have the time or technical support needed to bring those records together.
Lobby for Good is building a public record so people can see which developments are proposed, what resources they may use, which public bodies are involved, and which questions remain unanswered.
If a data centre or another major water or electricity user is being proposed near your community, share the documents and information you have found through the Lobby for Good Portal.
One local report may answer one local question. Records from several regions can show whether the same infrastructure pressures, advisers, and information gaps are appearing across the country.
That is how local experience becomes civic intelligence.
So far, the public has mostly been invited to observe decisions rather than help shape them.
That needs to change.
Join Lobby for Good and help build the people’s lobby.
No one should have to fight alone.
References
Environment Southland: Datagrid NZ Partnership Limited resource consents
Environment Southland: Water permit for the Datagrid groundwater take
Environment Southland: Section 42A recommending report for the Datagrid applications
Transpower: Data centres could add up to 5,000 GWh of electricity demand
RNZ, Nine to Noon: interview with Water New Zealand climate-change research group chair Evan Vaughters on the Datagrid consents and the long-term regulation of hyperscale data centres.
Science Media Centre: Expert reaction to the Southland data-centre approval
The Guardian: Local concerns about the proposed Makarewa data centre
Microsoft: New Zealand’s first hyperscale cloud region opens
New Zealand Digital Government: Cloud adoption policy and strategy
New Zealand Digital Government: Google Cloud Framework Agreement
New Zealand Digital Government: Amazon Web Services Cloud Services Agreement
New Zealand Government Procurement: Microsoft Cloud, Software and Services Agreement
Read the full series:
Water Reform Settled Who Manages the System. It Did Not Ask Who the System Must Serve.
Three Waters and Local Water Done Well were fiercely debated. The future water demands of data centres and other large industrial developments received far less public attention.
NO ONE NEEDS AI.
WE NEED TO DEACTIVATE IT NOW BEFORE IT RUINS OUR NATURAL WORLD, OUR MINDS AND OUR ECONOMY.
STOP ALL AI.