Taking Stock - 20 August 2026

Edward Lee writes:

DATA centres are coming, but who pays for the power?

For most people, artificial intelligence exists somewhere in the cloud. We type a question into ChatGPT, ask Claude to summarise a document or use Google to search the internet, and an answer magically appears seconds later.

The reality of what happens behind the scenes is much more physical.

Behind “the cloud” are enormous data centres filled with computer chips, cooling systems, batteries and electrical equipment. They consume vast amounts of electricity and, as investment in artificial intelligence accelerates, their demand for power is becoming large enough to influence electricity markets around the world.

The International Energy Agency expects global data centre electricity consumption to almost double over the next five years, creating a new challenge for governments.

Whilst countries want the investment, the technology and the economic activity associated with AI, they also need to work out where all the additional electricity will come from and who pays for the generation and grid infrastructure required to support it.

New Zealand has now joined that debate, and developments over the past few months suggest it could become an increasingly important issue for investors as well.

Last month, the Green Party called for a one-year moratorium on consenting new data centres in New Zealand as they would like to add additional regulation to the consenting process. It argues that data centres should be required to bring additional electricity generation with them and pay the full infrastructure costs associated with their development.

The concern about who pays for that infrastructure is legitimate, but we think the proposed solution is wrong.

New Zealand cannot afford to respond to critical infrastructure investment by putting consenting on hold, especially when the market is currently behaving as it should.

Data centres, electricity generation and transmission infrastructure already take years to plan and build, while artificial intelligence is developing considerably faster. 

Delaying investment decisions creates uncertainty, adds cost and increases the risk that projects are built elsewhere.

Australia, the United States, Ireland and other countries are all competing for data centre investment, and international capital will not wait indefinitely for New Zealand to make up its mind.

More importantly, overseas experience shows that these issues can be addressed without stopping development. 

The better response is to establish clear expectations around who pays for new infrastructure, how additional generation is supported and what long-term commitments large electricity users must make, while allowing investment to continue.

New Zealand's electricity companies also have a strong commercial incentive to ensure new demand is accompanied by new generation. 

Recent announcements from Contact Energy and Mercury suggest that this is already starting to happen. With 31 prospective large electricity projects already in the pipeline, we do need to think carefully about who should pay for the generation, transmission and other infrastructure required to supply them.

Not all of these projects will be built. Some projects will be delayed, others will become smaller and some will disappear entirely. 

Even so, the scale of the pipeline is significant.

That does not mean we should stop them. It simply means the electricity system needs to be capable of accommodating them, and the costs of doing so need to be allocated fairly.

A good example of how this could work emerged last week when Contact Energy and CDC Data Centres announced that they are exploring the development of a major new data centre in Taranaki. 

The proposed facility would have 250 megawatts of computing capacity and could require approximately 350 megawatts of electricity at peak times.

For comparison, that one development could have peak electricity demand equivalent to almost 5% of New Zealand’s current record national peak.

That sounds like a lot of power, but the announcement included how Contact and CDC were proposing to supply the site with electricity. Contact intends that electricity would be supplied under a long-term contract supported by new geothermal, wind and solar generation.

A long-term contract of this scale materially strengthens the investment case for new generation, giving Contact greater certainty that the electricity produced by a new power station will have a committed buyer.

Importantly, this arrangement is being developed commercially rather than waiting for a government moratorium to determine how it should work. Contact and CDC have recognised the problem and are designing the generation, storage and data centre together.

This is exactly the model New Zealand should be exploring. 

Rather than asking whether the electricity system can somehow absorb another 350 megawatts of demand, the developer and generator are working together to support new generation, add batteries, make use of existing grid infrastructure and secure long- term electricity demand.

This proposal also makes this particularly relevant to New Zealand investors because CDC is not some distant international technology company. 

Infratil owns almost half of CDC Data Centres. CDC has become Infratil’s most important investment and is now the largest developer, owner and operator of secure data centre infrastructure across Australia and New Zealand.

Infratil expects CDC’s earnings to exceed A$1 billion in the 2028 financial year and to reach around A$2 billion once its currently contracted capacity is fully deployed. 

Data centres have therefore moved from being an interesting part of the Infratil portfolio to one of the major drivers of its future earnings.

The proposed development brings both sides of the investment story together. 

Infratil shareholders participate through CDC, while Contact shareholders participate through the electricity generation, batteries and long-term supply contract required to power it.

Mercury provides another example. It has entered into a 15-year, 140 megawatt power purchase option agreement with Datagrid for the proposed data centre in Southland. The electricity involved represents roughly 3% of current national electricity demand.

Mercury has subsequently gone further, investing $53 million for a 12.7% equity interest in Datagrid.

Mercury is not just looking at a data centre as another large electricity customer. It is participating in the value created by the data centre itself while also providing a long- term electricity arrangement to support it, effectively benefiting from both sides of the coin.

These examples point towards a much more sophisticated relationship between data centres and electricity generators than simply supplying another large customer with power.

There is also a broader economic effect that should not be overlooked. 

If these projects proceed, the economic activity extends far beyond the data centre itself. 

New generation needs to be constructed, batteries installed, transmission strengthened and substations expanded. That creates work for construction companies, engineers, electricians, equipment manufacturers and the businesses that supply them. 

Further upstream, it adds demand for copper and other critical minerals needed for cables, transformers, batteries and electrical infrastructure. 

The result can be a substantial chain of investment, employment, wages and tax revenue flowing through the economy. At a time when New Zealand is searching for stronger productivity, investment and economic growth, that opportunity should not be dismissed lightly.

Building new renewable generation requires considerable capital upfront, while the asset itself might operate for several decades. Developers and their lenders therefore want confidence that there will be customers prepared to buy the electricity once the project is completed.

New Zealand solar developer Lodestone Energy provides a useful example. It has used long-term commercial offtake agreements as part of the development of its growing solar portfolio, including an agreement to supply renewable electricity across sites operated by The Warehouse Group.

A data centre potentially takes the same concept to another level. Instead of contracting electricity across a collection of shops or industrial sites, a single customer might be prepared to contract hundreds of megawatts for 15 or 20 years. 

That long-term demand can provide the revenue certainty needed to finance new generation.

This is why we should be careful about viewing data centres solely as a threat to electricity supply. A 350 megawatt data centre arriving tomorrow and competing for New Zealand’s existing generation would clearly create additional pressure. 

A 350 megawatt data centre whose long-term contract enables hundreds of megawatts of new generation and storage to be financed is a very different proposition.

The electricity demand itself can therefore become part of the financing solution. Rather than stopping development through a one-year moratorium, New Zealand should continue building while ensuring the rules governing new generation and infrastructure costs keep pace.

New Zealand already has an enormous infrastructure deficit. Electricity generation and transmission need significant investment, and demand is likely to increase as transport, industry and other parts of the economy electrify. 

Adding another layer of delay to major investment is unlikely to improve that position.

Data centres are critical infrastructure in their own right. Artificial intelligence, cloud computing, financial services, government systems and much of the modern economy increasingly depend upon them. 

New Zealand should be thinking about how to attract this investment and ensure we have the infrastructure to support it, rather than debating how to temporarily stop it.

For investors, this debate illustrates how quickly the artificial intelligence investment story is broadening.

The first stage of the AI boom was largely about semiconductor companies. Nvidia became one of the world’s most valuable businesses because its chips were essential for training and running increasingly powerful AI models. 

Microsoft, Alphabet, Amazon and Meta then began spending extraordinary amounts of money constructing the data centres required to deliver those models.

The next stage increasingly looks like an electricity and infrastructure story. Data centres require generation, generation requires transmission, and renewable generation increasingly requires storage and backup.

All of it requires transformers, electrical equipment, copper and enormous amounts of capital.

For New Zealand investors, this is already appearing in companies listed on our own sharemarket. 

Infratil provides exposure to the physical data centre infrastructure through CDC. Contact is considering a model that combines CDC with long-term electricity supply, new renewable generation and battery storage, whilst Mercury has entered into a long- term electricity arrangement with Datagrid and taken an equity interest in the company. 

Further down the chain, greater investment in electricity infrastructure means greater demand for the commodities required to build it. 

Companies such as BHP and Rio Tinto therefore have an indirect relationship with the AI boom through copper and other materials required for electrification. 

Internationally, Cameco provides another angle as nuclear power attracts renewed attention from technology companies seeking reliable electricity around the clock.

Artificial intelligence may have started as a technology investment story, but this is becoming an infrastructure and electricity story.

The Green Party's proposed moratorium has brought the debate into the open, but pausing development to add further regulation and rules would be a mistake. New Zealand needs investment, more electricity generation and greater productivity, and we should want to participate in one of the most important technological changes occurring in the global economy, especially when the market is already behaving as it should, without further regulation.

The objective should not be to protect New Zealand's existing electricity supply from data centres. It should be to use data centre investment to expand it.

If the world's largest technology companies want access to New Zealand's renewable electricity, there is nothing unreasonable about expecting them to help fund the generation and infrastructure needed to supply it.

Done badly, the AI boom could place further pressure on an already constrained electricity system. Done well, it could provide billions of dollars of investment, significant employment, greater demand for New Zealand businesses, additional tax revenue and large customers prepared to underwrite the construction of new generation.

The electricity required by artificial intelligence has to come from somewhere.

New Zealand's challenge is making sure that if we help power it, New Zealand captures its fair share of the economic benefits.

Travel

Johnny Lee - Taupo - 1 September

Johnny Lee - Hamilton - 2 September

Johnny Lee - Tauranga - 3 September

Johnny Lee - Christchurch - 7 September (FULL)

Chris Lee & Partners Limited