Hyperscalers need one national rulebook, not a council-by-council lottery

New Zealand has the renewables, the cool climate and the political stability that AI infrastructure investors want. What it doesn’t have is a national rulebook. That gap is already visible. Singapore-based Datagrid has won resource consent for a $3.5 billion data centre at Makarewa, Southland. The project could draw up to 280MW, enough to make it the country’s second-largest electricity user behind Tiwai Point. It went through a local council. The national framework is still being written.

The rules are arriving after the precedent, not before it. For an industry built on 15-year power contracts, that sequencing matters.

Boardrooms want certainty, not a ban

Infratil chief executive Jason Boyes calls the AI infrastructure boom “a 10-year theme”. He says New Zealand is well placed to become a hub, but warns: “I don’t think New Zealand will miss out, but we risk being late if we delay the benefits of investments.” His one request of the next government is to “establish clear national rules, rather than regional ones, around the issues communities raise.”

Harbour Asset Management’s Shane Solly explains why. Building has become harder in the US and Europe because of community pushback and power constraints, and in the Middle East it “has stopped”. Grid drag, noise, water use and visual effects are what local communities fight over. For investors, the real risk is timing. A multimillion-dollar power purchase agreement with a hyperscaler can fall apart if a consent stalls.

Makarewa exposed the gap

Green co-leader Chlöe Swarbrick says the Southland facility could use about 6% of national electricity supply. She also points out that councils can consent these projects “on a non-notified basis with no public input required.” Her answer is a one-year pause on new consents, and that is the wrong one. Global capital won’t wait a year for New Zealand to make up its mind. Her diagnosis is harder to dismiss, though. When a project of national significance is decided under district plan rules, both sides lose. Communities get no clear say, and investors get no predictable process to rely on.

Additionality is the right instinct

The Government’s response is better than the Greens’ alternative. Prime Minister Christopher Luxon has asked Nicola Willis and DPMC to develop a national “compact”, described as “the deal by which you come into New Zealand if you want to build out a data centre.” At its centre is additionality, meaning large operators must bring new generation with them.

Energy Minister Simeon Brown says hyperscale facilities should underwrite new generation to match their own demand. Data centres use about 0.6% of national power now, and he expects that to reach about 3% by 2030. He rejects putting up a “you’re-not-welcome sign” because “we’re a small country at the bottom of the world.” That is sound. The catch is that Brown also concedes the enforcement mechanics are still under development. Investors can’t price a principle.

Treasury’s memo is the reality check

Officials are more cautious in private. A July 2026 Treasury memo released under the OIA puts current data centre load at 8-104MW. It counts 1,140MW more in Transpower’s connection pipeline, against about 19,000MW of mostly intermittent generation seeking connection. Officials also warn that hosting foreign-owned data is “unlikely to deliver large economic or productivity gains”. They note that most of the jobs come during construction. Their core concern is timing, with demand likely to outpace the build-out of generation and transmission.

The scale was already clear a year ago. In September 2025, the Electricity Authority found data centres made up 64% of new load capacity in its 1,780MW pipeline.

Designed well, they can help the grid

The burden story is only half the picture. Many AI workloads aren’t time-critical, so they can be scheduled to soak up surplus wind and solar that would otherwise be wasted. That improves the economics of new renewable projects. Transpower modelling suggests every gigawatt of avoided peak demand saves around $1.5 billion in system costs. The same analysis warns that poorly planned projects could “worsen local network congestion and compete with other users for renewable electricity.” How the rules are designed decides which outcome New Zealand gets.

What investors need before the next consent

DLA Piper argues New Zealand should build its own framework rather than copy one from overseas. That means separating credible projects from speculative ones and dealing with water, noise and Treaty obligations early. BCG analysis reported by BusinessDesk has called for data centres to be explicitly included in fast-track consenting, with clarity on how Investment Boost applies. Connection pricing needs fixing too. Back in 2024 the Electricity Authority admitted the rules had “not kept pace” with the size of connection applications.

Whoever forms the next government inherits the compact half-built. Over the next six to twelve months it has to become an enforceable additionality test, a national consenting pathway and consistent connection pricing. If it doesn’t, New Zealand will repeat the pattern already driving hyperscalers out of parts of the US and Europe, where every project is fought out one council hearing at a time. Makarewa should be the last big data centre decided that way.

Sources

• NZ Herald: AI data centres: Business leaders urge clear national rules for New Zealand (2026-10-29)
• RNZ: Urgent need for rules and protections on data centres, Greens say (2026-07-26)
• 1News: Govt wants ‘common-sense’ rules around big data centres – PM (2026-07-27)
• Treasury (via PSGR): Data centres in NZ – Memorandum to Ruth Fairhall (2026-07-23)
• Electricity Authority: Generation investment pipeline updates and insights (2025-09-16)
• BusinessDesk: $70b datacentre opportunity could be seized with policy changes, BCG says
• Electricity Authority: Proposals to help businesses and infrastructure get connected (2024-10-25)