One unbuilt data centre is already rewriting NZ energy policy

The load nobody planned for

Data centres used to be someone else’s problem, big sheds full of servers, invisible to the grid. Not anymore. The proposal now driving New Zealand’s energy debate is Datagrid’s AI facility at Makarewa, north of Invercargill, which has Overseas Investment Office sign-off to scale from an initial 280MW to a full 1GW across a 49-hectare site.

The numbers are what make this a policy problem rather than a construction story. At full capacity the $5 billion facility would consume nearly twice the electricity of the 572MW Tiwai Point aluminium smelter. New Zealand’s peak real-time demand is about 6GW. Datagrid and Tiwai combined would be 1.5GW, a quarter of the whole country’s peak, from just two sites.

And this is not demand you can manage around the evening peak. Data centres run 24/7, 365 days a year because powering servers down damages them. It is permanent baseload. MBIE’s own 2024 modelling assumed data centres run at a 90% load factor.

The modelling is already obsolete

Back in July 2024, MBIE’s Electricity Demand and Generation Scenarios projected data centre demand of 1.8 TWh by 2030 in its reference scenario, rising to 4.6 TWh in the growth case. It identified 233MW of committed capacity from the likes of Microsoft, AWS and CDC. The Datagrid proposal alone, at 1GW, blows past that entire national figure. As a planning guide, the 2024 numbers are dead.

The grid itself is in reasonable shape. In the March 2026 quarter, renewables hit 94.5% of generation, solar jumped 50.2% to a record 373 GWh, and industrial demand rose 4.8%. New wind is coming online in the south. But the system leans heavily on hydro, which means dry-year risk is permanent, and Datagrid’s proposed diesel backstop may only cover a few hours, not the extended stretches when lake levels bottom out.

Why this is a power bill story

Here is the part that matters for every business paying an electricity account. If a giant new load draws from the existing grid without bringing its own generation, the extra system costs get socialised across everyone. Newsroom reported in March 2026 that the Datagrid facility raised exactly this question over household and business bills.

The scale of the number is stark. Transpower modelling cited in a July 2026 analysis estimates every gigawatt reduction in peak demand avoids around $1.5 billion in system costs. Read it the other way and a gigawatt of unmanaged new baseload could add that much, flowing straight through to prices.

Additionality is the whole ballgame

The government has landed on the right principle, though not yet the rule. On 27 July 2026, Prime Minister Christopher Luxon signalled a ‘compact’ for hyperscale operators built around ‘additionality’, the requirement that large centres contribute new generation rather than just consuming existing supply. ‘Large ones, absolutely, we need to make sure that there is additionality,’ Luxon said, with Economic Growth Minister Nicola Willis leading the work.

Right now there are no national rules. Consenting sits with local councils, often non-notified. The Greens want a one-year pause to write standards first, mirroring New York state’s July 2026 moratorium. Luxon calls that alarmist. He may be right that a blanket pause is clumsy, but a principle without a rule protects nobody’s power bill.

The upside if the rules are good

This does not have to be a threat. Victoria University’s Alan Brent and Durham University’s Andrew Crossland argued on 31 July 2026 that with the right pricing signals, data centres could ramp up when wind and solar are generating surplus and dial back at peak, acting as a giant flexible load that smooths the grid. EECA calls flexible demand one of the cheapest ways to cut future network investment.

University of Auckland’s Gehan Gunasekara went further on 26 July 2026, warning of a ‘digital Strait of Hormuz’ if growth goes unmanaged, and proposing long-term supply contracts plus a hard requirement that centres draw from new generation. That is additionality with teeth.

The investment is real and welcome. But whether it lands as an asset or a cost depends entirely on whether the compact becomes an enforceable rule before the concrete sets. Get additionality binding and NZ gets the load, the capital and a more flexible grid. Leave it as a principle and every other consumer quietly picks up the tab.

Sources

• Explainer: What is a data centre and why is everyone freaking out now? (2026-08-02)
• Great southern data centre: Is this the end of the world as we know it? (2026-06-11)
• Govt wants ‘common-sense’ rules around big data centres – PM (2026-07-27)
• NZ sleepwalking into a digital Strait of Hormuz (2026-07-26)
• Fears new data centre could bump up power bills (2026-03-17)