Four rugby fields of nothing
When Prime Minister Christopher Luxon attended the launch of Amazon’s New Zealand cloud region in September 2025, the west Auckland site at Fred Taylor Drive sat bare. Four rugby fields of land Amazon had acquired between 2022 and 2024, with nothing on them but dirt. The photos and video Amazon supplied to media at the ceremony were of overseas facilities.
Amazon had spent roughly $40 million on the abandoned project, including approximately $33 million acquiring the land. Contractor Naylor Love, when approached by the NZ Herald, offered five words: “We’re under NDA. We can’t comment.”
Instead of building, Amazon pivoted to co-location, installing its server racks inside data centres owned by third parties, reportedly including Infratil-owned CDC. The NZ cloud region launched a year late, running on rented infrastructure.
The original promise versus the reality
The gap between announcement and delivery is worth tracing. In 2021, AWS told then-PM Jacinda Ardern it would invest $7.5 billion and build three physical data centres in Auckland. The Overseas Investment Office granted consent in March 2022, valuing the property and business establishment at $250-350 million. Amazon bought the land. Geotechnical work began.
Then it stalled. In early 2024, Auckland Council confirmed unresolved stormwater discharge issues had put the consent process on hold. Newsroom reported rising power prices as a key factor in the eventual abandonment. By early 2025, all drilling and digging had ceased.
The actual capital deployed in New Zealand: an empty site and some server racks in other companies’ buildings.
The $7.5 billion figure doesn’t survive scrutiny
AWS’s own economic impact study projected $7.5 billion in planned investment, $10.8 billion in GDP contribution, and more than 1,000 full-time equivalent jobs annually. The government repeated these numbers enthusiastically.
In August 2025, Forrester VP and principal analyst Sam Higgins warned that such studies “have become as common as cloud services these days” but are rarely validated after the fact. He called the GDP claims “rallying cries for market share rather than anything designed to prove the delivery of real or measurable outcomes.” Academic reviews suggest such studies can overstate benefits by 30-60 percent.
Tech commentator Peter Griffin in September 2025 advised taking the figures “with a pinch of salt,” noting AWS had extrapolated downstream economic activity and “probably adding in a healthy margin as well.” Catalyst IT managing director Don Christie estimated only a few dozen direct jobs at most, not 1,000. Dita de Boni in The Spinoff put it most sharply: “Amazon Web Services is somehow classing its own paying of its power bill as an investment in New Zealand.”
What this means for businesses running on AWS
The co-location model has practical consequences. Christie warned that AWS’s local availability zones are less robust than they should be given New Zealand’s seismic risk. Purpose-built hyperscale data centres are engineered for earthquake resilience in ways that co-located racks in a third-party facility may not be.
Customers who signed up for the promised local data centres, including the Ministry of Justice, Air New Zealand, and ANZ Bank, are getting local data residency. But they may not be getting the resilience architecture they expected. And the identity of the facilities hosting their data remains secret.
There is also an energy question. Amazon’s power purchase agreement with Mercury to offtake roughly 50 percent of the 103-megawatt Turitea South wind farm could effectively export scarce renewable energy to service global clients, rather than supporting New Zealand’s own transition.
The signal that matters
Information security expert Anthony Grasso from Titanium Defence offered a more charitable reading in September 2025, suggesting that “renting space and making it secure is far cheaper than massive construction” and that the co-location approach was pragmatic rather than a failure.
Perhaps. But the practical signal is the same. Cloud demand in New Zealand is real. AWS’s own study cites data showing cloud infrastructure contributed $4.8 billion to the NZ economy in 2023. The market is here. Yet the world’s largest cloud provider assessed New Zealand’s power prices, consenting environment, and construction economics, spent $40 million finding out it was too hard, and decided to rent.
Until those economics change, every New Zealand business running critical operations on AWS is doing so on infrastructure that a company with $100 billion in annual revenue decided wasn’t worth building. That should bother more people than it does.
Sources
- NZ Herald: Amazon construction freeze – PM at tech giant’s launch but massive Auckland site sits bare (2025-09-02)
- Newsroom: Amazon aborts construction of west Auckland data centre (2025-09-02)
- Newsroom: Tech giant Amazon’s $7.5b NZ data centre plan quietly put on hold (2024-03-01)
- AWS New Zealand Economic Impact Study (2025)
- iStart: Unpacking AWS’ local data centre and training claims (2025-08-08)
- NZ Herald: Amazon’s $7.5b NZ data centre plan questioned over job, GDP claims (2025-09-02)
- The Spinoff: Luxon hypes AWS opening as critics call investment old news (2025-09-03)
- RNZ: Amazon and the case of the missing New Zealand data centres (2025-09-02)
- BusinessDesk: A year late, but the Amazon Web Services New Zealand region $7.5b investment is here (2025-09-02)