Auckland Light Rail ran for six years, spent $228 million, and laid zero metres of track.
At its peak the project was paying about $920,000 a week to two engineering firms. Not to build anything. To plan, re-plan, and re-plan again, through three different versions of the route. In January 2024 a new government cancelled it as part of a 100-day plan, and the public got nothing back for a quarter of a billion dollars except some very expensive PDFs.
I want to be upfront about what this series is, because the material is political and the timing is an election year. This isn't a political hit job. It's pattern recognition on the public record, and the pattern has both parties' fingerprints all over it.
The same meeting, every three years
I've spent a chunk of my career in exec rooms, at startups and at big companies, and I've been through my share of strategy rewrites. Even one rewrite, done inside one company, with everyone trying their best, costs you about a year. Roadmaps get binned. Half-finished work gets written off. Your best people quietly update their CVs. The new strategy is usually about 70% the old strategy with new diagrams.
New Zealand runs like a company that's contractually required to put its entire exec team up for replacement every three years. Each incoming team arrives with a mandate to show movement in a hundred days, and the fastest way to show movement is to bin whatever the last team was building. The reorg is the announcement. The announcement is the work.
Any one of these execs might be competent on their own. The dysfunction is the team, and the team is permanent: it spans parties, decades and ideologies, and it cannot leave a predecessor's decision alone. Meanwhile the shareholders of NZ Inc can't sell their shares. A decent chunk of them are still at primary school.
That's the frame. Now the receipts.
The Cook Strait ferries are the cleanest one. In 2021 KiwiRail signed a fixed-price contract for two new rail-capable ferries, with port upgrades to match. By December 2023 the cost had escalated badly and the incoming government cancelled the lot. By then $507.3 million had already been spent, and 1News later confirmed we paid Hyundai another $144 million to walk away from the shipbuilding contract. So: $651 million, no ferries. Then we ordered different ferries anyway, for $1.86 billion, due 2029. The old rail-enabled Aratere got retired in the meantime, which severed the Cook Strait rail link for roughly five years.
We paid $144 million to not get ships, and then bought ships.
Three Waters took seven years from the Havelock North outbreak to a full legislative framework, then got repealed under urgency in an afternoon in February 2024. You can think the reform was wrong and still notice the maths: the $120 to $185 billion of ageing pipes that justified it didn't get repealed with it.
The Resource Management Act replacement is my favourite, in the way a slow-motion replay of a faceplant is someone's favourite. Parliament spent years building the RMA's successor, passed it in August 2023, and the next government repealed it 122 days later. The replacement's replacement arrived in late 2025, and The Spinoff noted its principles "all but mirror" the laws that got repealed. Thirty-plus years of everyone agreeing the RMA must go, and the RMA is still here, having outlived its own successor.
Te Pūkenga merged sixteen polytechnics into one national institute in 2020. In January 2026 it finished un-merging them back into ten. Between those two restructures: 855 staff gone, $9.5 million in redundancy payouts, and $325 million allocated to stand up the new polytechs that look a lot like the old polytechs. If you've ever survived a corporate restructure followed by a de-restructure, you know exactly what those six years felt like from inside.
The Smokefree generation law passed in December 2022 as a world first, got studied by half the planet, and was repealed before a single clause took effect.
And if you want the purest specimen of the whole genre, it's the bright-line test. National invented it in 2015 at two years. Labour stretched it to five, then ten. National snapped it back to two in 2024. Four settings in nine years, on a tax rule that decides what happens when ordinary people sell a house. Both teams, same dial.
Worth noting who started and stopped each of these. Labour built ALR, iReX, Three Waters and Te Pūkenga; National stopped them. National invented the bright-line test; Labour extended it; National reverted it. And Labour announced a $785 million harbour cycle bridge in June 2021, then cancelled it itself by October, which proves you don't even need to change the government to get the U-turn. The cycle doesn't care who's driving.
Only money actually spent
Time to add it up, and this is where I need to be careful, because the temptation with a topic like this is to quote the scariest number available.
Politicians on every side do exactly that. When light rail was cancelled, ministers cited builds costing "$15 billion, rising to $29.2 billion". Those are projections of money we now won't spend. Counting avoided future costs as waste is fantasy accounting, and the moment you do it, anyone with a calculator can dismiss your whole argument.
So the rules for this series are boring on purpose. Count money that actually left the public account on things that were then cancelled or reversed: sunk planning, design, contracts, land, plus the penalties paid to walk away. Never count "would have cost" figures. Attribute anything contested to whoever claims it, like Labour's Tangi Utikere putting the full iReX cancellation at $1.16 billion once you include ongoing maintenance, a figure that bundles in things I won't count.
On those rules: light rail's $228 million, iReX's $507.3 million plus the $144 million exit, the taxpayer share of Let's Get Wellington Moving ($109.7 million of the $180.7 million it spent over nine years, mostly on consultants; Wellington's ratepayers carried the rest), Te Pūkenga's redundancy bill, and the $80 million-plus in public service redundancy payouts across 24 agencies as the workforce swung up by 13,000 and back down again.
That's well over a billion dollars in directly sunk and cancellation costs from a single change of government, counted conservatively. A billion dollars is roughly half a new Dunedin Hospital, a project which has itself been rescoped so many times that 35,000 people marched down George Street about it.
I lived on the North Shore through the announcement years, and the ritual became familiar: the render, the press conference at the empty site, the revised render, the quiet line in a later Budget. SkyPath, then the standalone cycle bridge, then neither. Light rail to the airport in three different flavours. Somewhere around 2021 I stopped believing artists' impressions the way I'd stopped believing crypto whitepapers.
Here's the backdrop that turns this from annoying into expensive. Te Waihanga's 2022 stocktake found that from 2010 to 2019 New Zealand was the biggest infrastructure investor in the OECD as a share of GDP, while sitting near the bottom 10% of high-income countries for how much infrastructure each dollar buys. We're not under-spending. We're spending like a rich country and delivering like a country that re-litigates its decisions every electoral cycle, into an infrastructure deficit Te Waihanga and Sense Partners size at somewhere between $100 and $210 billion over 30 years.
I grew up in Christchurch, and the rebuild taught me what both versions of this look like on the ground. The bits that got locked in and left alone got built; the city got a genuinely world-class convention centre and a repaired arts centre. The bits that kept getting re-litigated dragged on so long they became civic jokes. The stadium took fifteen years of arguments and only opened this year. Same city, same decade, same money. The difference was whether each new set of decision-makers honoured the last set's call.
When stopping is the right call
A fair pushback: sometimes cancelling a project is the competent move, and an exec team that can't kill anything is its own kind of disaster.
True, and the record backs it in places. Treasury papers showed about 80% of the iReX cost escalation was seismic strengthening at the ports, work that's needed no matter whose ferries dock there. Pulling the pin on a runaway project can be governance working, and I'd rather have a government willing to stop a bad bet than one that rides it down out of pride.
The target of this series is narrower: the re-litigation of settled, long-horizon decisions. Ferries wear out on a schedule that doesn't care about coalition agreements. Pipes corrode on chemistry's timetable, not the electoral one. When the underlying problem has a 30-year horizon and the decision-making has a 3-year one, reversal isn't prudence. It's the exec team marking its territory, and the bill lands on shareholders who can't vote yet.
So that's Part 1: the cost, counted conservatively, with the fantasy numbers left out.
Part 2 asks why New Zealand specifically does this more than nearly anyone, because it turns out we're rigged for it in ways most democracies aren't, and the man who wrote the book on it in 1979 has updated his diagnosis.
Part 3 is where I stop reading the receipts and start pulling them. The entire statute book is published as open data: every act, every amendment, every repeal. I've started parsing all of it, back to 2008, to measure the churn properly. It's a personal project and a few evenings of code, not a royal commission. But nobody else has built it, most of NZ Inc's shareholders haven't been born yet, and somebody should be keeping the receipts.