Around the world, the boom in artificial intelligence is driving a parallel race between major tech companies to secure the physical infrastructure that enables it.
This race has now reached New Zealand, with Singapore-based Datagrid rolling ahead with plans to open the country’s first “AI factory” near Invercargill.
The scale of this major data centre, designed to power energy-hungry global AI systems, is substantial – both in its 78,000 square-metre size and its multi-billion-dollar cost.
It is not the only major investment by global players. Just last year, Amazon Web Services announced plans for a large new cluster of data centres in Auckland, billed as a NZ$7.5 billion investment in New Zealand’s digital infrastructure.
For smaller countries such as New Zealand, projects like these are often presented as wins – bringing jobs, investment and a pathway into the global AI economy.
But they also carry less visible consequences, many of which are more complicated than they first appear.
The hidden economics of AI infrastructure
Behind these projects sits a rapidly expanding global demand for computing infrastructure, driven by the growth of AI systems such as ChatGPT and Claude.
Increasingly, governments are leaning into this shift by competing to attract data centres as part of their approach to AI and the digital economy.
In New Zealand’s case, officials have pointed to the country’s renewable energy, cool climate, available land and political stability as reasons it could appeal to foreign investors. New Zealand Trade and Enterprise has even suggested the country could become an “international data centre hub”.
Datagrid’s project has been presented as an early example of that opportunity. The company has described it as “the most significant upgrade to New Zealand’s digital infrastructure in a generation”.
Some of the potential downsides are straightforward.
These systems rely on powerful computer chips and so use far more energy than typical digital services.
The Datagrid facility, for instance, is expected to draw up to 280 megawatts of electricity, or around 6% of national demand – making it New Zealand’s second-largest electricity user after the Tiwai Point aluminium smelter.
But other drawbacks are less obvious.
My research on data centre markets suggests that, while the infrastructure is built locally, the systems it supports usually operate across multiple countries.
Datagrid has indicated it expects to serve international AI and cloud providers, positioning New Zealand as a potential host for global AI workloads.
These workloads are likely to come from AI markets already dominated by a handful of big tech companies. This, in turn, raises questions about where value is created – and how much of it is retained locally.
Who builds the infrastructure – and who controls it?
For smaller economies like New Zealand, relying on large international investors is often seen as a practical necessity, given the scale of capital and market reach required to build and run these facilities.
But large data centre projects can also have significant impacts on the shape of local cloud service providers – companies that typically offer data storage, computing services and IT support to businesses and government.
These smaller domestic players often find it difficult to compete directly with global firms on price and influence.
As such, they are often forced to find new opportunities. As AI increases demand for land, power and connectivity, local firms are taking on roles in building and running the infrastructure that supports them.
This tends to shift them toward the physical side of the industry, securing sites, brokering access to energy supply, building facilities to overseas market specifications and meeting both local and international legal requirements.
Global companies such as Amazon Web Services, Microsoft and Google, meanwhile, provide the platforms and software that run on top, often by leasing capacity from local operators.
Recent developments in New Zealand’s own cloud sector illustrate how these arrangements can work in practice.
Amazon Web Services, for example, recently disclosed it had shifted away from plans for a large standalone Auckland data centre build, while continuing to expand its local cloud operations through co-location agreements with local data centre providers.
For countries hosting this infrastructure, the relationship can become uneven, with hosts supplying the land, energy and networks while key decisions about how these systems operate are often made elsewhere.
This raises broader questions for governments. If domestic firms are concentrated in infrastructure roles, how does that affect the development of higher-value capability over time?
And how should countries weigh immediate gains – jobs, investment and connectivity – against longer-term control and positioning in the digital economy, including what this means for where value is created and retained?
Projects like Datagrid’s will still deliver clear benefits. But their wider impact will depend on how trade-offs are managed, and how much influence countries retain over the systems they host.
Angus Dowell does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
