The artificial intelligence boom is driving a global expansion in data centres - and with it, public debate over their heavy demand for power, water and land.
Authors
- Alan Brent
Professor and Chair in Sustainable Energy Systems, Te Herenga Waka — Victoria University of Wellington; Stellenbosch University
- Andrew Crossland
Professor in Practice, Department of Engineering, Durham University
Increasingly, governments are weighing their economic benefits against concerns about pressure on electricity grids, local communities and the environment.
That debate has now reached Aotearoa New Zealand. Singapore-based Datagrid has secured approval for a NZ$3.5 billion, 280 megawatt (MW) AI data centre near Invercargill.
Scheduled to open in 2028, it will become the country's second-largest electricity user after the Tiwai Point aluminium smelter .
While Datagrid's " AI factory " is expected to more than double the country's data processing capacity, it has drawn some public unease . The Green Party recently proposed a one-year pause on new AI data centres while stronger rules are developed.
Similar regulatory debates are playing out overseas in places such as New York , Amsterdam and Australia .
In Ireland - now a global hub for AI infrastructure - data centres accounted for nearly a quarter of the country's metered electricity consumption last year, forcing difficult questions about its future grid capacity.
That's particularly relevant to New Zealand, where electricity shortages can already emerge during dry years when hydro lake levels are low.
But do data centres inevitably have to become another energy-hungry burden on electricity systems? And, if designed and operated differently, could they help support a cleaner, more flexible grid?
Can data centres be grid assets?
The answer lies in how these centres consume electricity. If they can adjust when and how they use power, they could become increasingly valuable as New Zealand adds more wind and solar generation to its electricity system.
But how, precisely?
Many of the computational tasks performed in data centres are not time-critical. AI model training, large-scale data processing and other batch computing can often be scheduled for periods when there is plenty of renewable electricity, wholesale prices are low or the grid is under less strain.
Rather than increasing demand during evening peaks, data centres could help absorb surplus wind and solar generation that might otherwise be curtailed. This gives renewable developers greater confidence that the electricity they generate will have a buyer, making new projects more financially viable.
Transpower estimates every gigawatt reduction in peak demand could avoid around NZ$1.5 billion in system costs. The Energy Efficiency and Conservation Authority has also identified flexible electricity demand as one of the cheapest ways to reduce future network investment.
Similar thinking is already emerging overseas.
In Australia, flexible data centre demand is helping support new solar and battery investment. Britain's electricity system operator has suggested new data centres could be built faster in Scotland where renewable electricity is most abundant and the grid can best support them.
Data centres can also provide services besides simply consuming electricity. Most already contain sophisticated electrical infrastructure, including uninterruptible power supply (UPS) systems, batteries, backup power systems and, increasingly, on-site renewable generation .
With appropriate market rules, these assets could help stabilise the grid by providing fast frequency response and reserve capacity. Data centres themselves could temporarily reduce demand during periods of network congestion.
Many existing battery and UPS systems are already technically capable of providing these services, although market arrangements would need to evolve before they could participate more widely.
As well, data centres' large and predictable electricity demand could underpin more renewable generation.
Long-term power purchase agreements could improve the financial viability of wind and solar projects - particularly smaller community-owned developments that can otherwise struggle to secure finance.
Datagrid's power purchase agreement with Mercury covers 140MW of its 280MW electricity demand. This leaves open the possibility that the remaining demand could underpin a new generation of community-owned renewable energy projects across Southland.
Getting the incentives right
There is no doubt data centres are large electricity consumers. But that does not mean they have to be a liability for the electricity system.
Poorly planned projects could increase peak demand, require costly transmission and distribution upgrades, worsen local network congestion and compete with other users for renewable electricity.
But with the right design and market incentives, data centres could instead become part of the solution. By shifting when they consume electricity, they can help absorb variable wind and solar generation, reduce pressure on the grid and make it easier to operate a highly renewable electricity system.
For regulators, the key question isn't whether data centres are good or bad for the electricity system - but how to create the planning rules, market incentives and operating arrangements that allow them to become part of the solution.
If New Zealand can get those settings right, data centres can help integrate renewable energy, support community energy projects and improve grid resilience.
With more projects likely to follow Datagrid's Southland development, Aotearoa New Zealand's goal should be to make data centres grid assets rather than grid constraints .
Alan Brent is a director of Wellington-based engineering consultancy Novaris Grid & Energy Systems. It has no commercial relationship with Datagrid.
Andrew Crossland is employed by the National Energy System Operator (NESO), a UK government-owned public corporation.
