Demand for AI computing is rising faster than ever, and while computing chips are increasing in quality each year, they need electricity to keep up.<\/p>\n
\u201cWhen data centres were growing off a really low base, you could grow your energy demand at that kind of rate. But now we\u2019ve already got quite a big installed base, tripling the amount of energy consumed by data centres just won\u2019t happen.<\/p>\n
\u201cWhen the constraint becomes power … the winners will be the markets that are power-rich. And the answer to that right now is China.\u201d<\/p>\n
Despite energy pressures, Newfield believed safety would be the first constraint on investment, acknowledging recent development of AI agents acting outside their remit and the potential repercussions.<\/p>\n
As a result, the social licence for AI technology is already reaching a crossroads, with protests against data centres and the technology prevalent in the US. They\u2019ve also begun to happen in New Zealand.<\/p>\n
With those pressures mounting, Newfield offered a prediction.<\/p>\n
\u201cI actually believe the US is going to hit a constraint very soon, probably straight after the midterms, which will be a lack of electricity and lack of social licence and that will create this massive overflow opportunity of the AI industry needing to move out of the US.\u201d<\/p>\n
As for where that demand will go, Newfield said several characteristics would support decision-makers.<\/p>\n
These included countries with abundant renewable energy, geopolitical stability, locally owned data centres and forward-thinking regulation.<\/p>\n
Newfield also suggested membership of the Five Eyes defensive pact would be beneficial, as AI companies including Anthropic look to demonstrate trustworthiness to government departments such as the Pentagon.<\/p>\n
So where is that investment most likely to go? Australia.<\/p>\n
\u201cI know the flags are quite similar, but right now New Zealand should be eating the US\u2019s lunch and Australia will. Anthropic are coming down saying we expect Australia to be our biggest, after the US, our biggest global centre for AI development because you have all of those characteristics.<\/p>\n
\u201cActually New Zealand is naturally way more advantaged than Australia, we just haven\u2019t yet made the case strong enough and we haven\u2019t pushed fast enough. This is something we can\u2019t afford just to be a fast follower.\u201d<\/p>\n
Newfield pointed to research from Boston Consulting Group (BCG) which showed if New Zealand captured between 2-3% of the Asia-Pacific AI data centre market, it would equate to a US$60-$70 billion ($107-$125b) opportunity, based on about 600 megawatts of capacity over 10 years.<\/p>\n
He stressed the target was too conservative, noting Morrison\u2019s Australian data centre business signed a single 550 megawatt contract earlier this year, and argued there could be hundreds of billions of GDP opportunity for New Zealand if it gets its settings right.<\/p>\n
One area Newfield suggested could change was energy generation, noting his view had changed over the past few years.<\/p>\n
\u201cIf you\u2019d asked me two years ago, I would have said don\u2019t mess with a world-leading model that New Zealand has, because what New Zealand\u2019s model has done has been really good at preventing overbuild and efficiently adding generation as it\u2019s required.<\/p>\n
\u201cThe change is with AI, there is a national competitive advantage from bringing in excess capacity, and the only way you bring in excess capacity is not short-term market signals. It will be some form of long-term underwrites, and you need to do that in a minimally disruptive way.\u201d<\/p>\n