One of the more popular ones, and a recurring social media trope, claims that America’s Army Corps of Engineers was prepared to build a proper motorway between Wellington and Auckland after the Second World War. And that the New Zealand government turned them down.
A more restrained version claimed that the Americans would have built Transmission Gully to facilitate movement from their base at Paekākāriki through to the port at Wellington.
Both versions are false, of course. Highways are expensive. They tend not to be provided as gifts. And if one had been offered, would the government really have been so silly as to decline it?
But some things that sound too good to be true are, nevertheless, real.
I think there is a very brief window in which mostly-American hyperscale datacentres could massively expand renewable energy generation in New Zealand, at their own expense, to the country’s much broader benefit.
I worry, though, that the country will miss the opportunity.
A little over two years ago, Leopold Aschenbrenner explained the basic logic of international competition in artificial intelligence. The owner of an artificial intelligence competent enough at programming to help build its own successor will have a durable advantage over others that follow.
Being a month behind the competition matters a lot if your competitor was able to run a few hundred thousand, or more, capable AI programmers to help it make progress during that month.
Companies have been betting on that scenario, as well as general AI uptake and diffusion, for a few years now. The pace of datacentre buildout has been astonishing. A large portion of current US economic growth is investment in datacentres and the things they need.
Tales of datacentres’ water requirements are largely urban legend, or at least depend on not really understanding industrial and agricultural-scale water use. Compared to a household, they’re large. Compared to a meatworks or an irrigated paddock, they’re not that large.
But one of the things they really do need is energy. Lots of it. Large datacentres are measured by the amount of energy their compute requires.
America’s electricity grid cannot keep up. Getting a connection to the grid now takes between three and seven years. And the queue of projects wanting connection represents more power than the entire country’s current generation capacity.
Not falling behind competitors really matters. So the frontier labs want more compute and they want it right now. If a hyperscale datacentre (really large, in lay-terms) cannot get a grid connection, it will set up its own power generation.
The kinds of generation that datacentres can build on-site in a hurry are not particularly efficient. Their electricity is expensive, compared to the power generated by the kinds of powerplants that electricity utilities build. But some of them can be built in as little as 18 months, which is much better than waiting up to seven years for a connection to the electricity grid.
You may remember Caterpillar as the maker of the lovely large yellow equipment that toddlers young and old love to see working. They’ve also made generators for off-grid construction projects. They’re now massively scaling that side up and are delivering gigawatts of natural gas generator sets to datacentres.
Hyperscalers are willing to pay a lot if they can get access to energy in a hurry, without large fishhooks along the way.
Two years ago, Aschenbrenner’s logic led me to one obvious conclusion. A country that could quickly say yes to a proposed geothermal plant sited next to a new datacentre could attract billions of dollars of investment, very quickly.
Over the past two years, the scale of that opportunity has increased considerably.
The stakes have also increased. For a relatively short period, the American government restricted access to a leading model, allowing only American citizens to use it. It would be surprising if the American government did not impose more substantial export restrictions on its frontier models over the coming years. If American companies have access to frontier models and New Zealand companies do not, the country’s productivity challenges will worsen.
Having local datacentres able to run the best open-source models would provide some insurance. If no one builds them now, a future government may well want to build them itself, at taxpayers’ expense.
New Zealand has enormous amounts of untapped conventional geothermal energy. The country’s electricity generators know how to build geothermal plants. The broader Taupo volcanic zone is, with Iceland, one of the two best places in the entire world for both conventional geothermal energy and the next generation of ‘supercritical’ geothermal energy.
A clear, rapid, and certain consenting path for new datacentres and the energy that powers them, whether geothermal, solar or otherwise, able to deliver decisions in a couple of months, could see enormous investment in very short order.
The consenting path is only part of it. In their submissions on the Government's draft geothermal strategy, Contact, Mercury and Eastland told MBIE that the Geothermal Energy Regulations 1961 add time and cost from consenting through to operation. The rules conflict with newer health and safety law. Pipeline requirements apply a standard written for long-distance hydrocarbon transmission. And regulations drafted in 1961 never anticipated supercritical geothermal at all. The law has no category for the thing the country is best placed to build.
Contact asked the Government to accelerate its planned review of the rules rather than wait until next year, and to run an end-to-end review with industry. On the Government's current schedule, the review would have little chance of finishing before the window to attract investment starts closing.
At the scale at which these investments are made internationally, the New Zealand government’s current support for developing next-generation supercritical geothermal generation looks like spare change. The companies that would fund the energy and datacentre buildout could very plausibly fund a hastening of current work on next-generation geothermal.
If the work on supercritical geothermal pans out, the country would have near-limitless, low-cost electricity for generations to come. Some of it would power datacentres; the rest of it could power everything else.
This isn’t an urban legend like the one about the Army Corps of Engineers that would have built New Zealand highways for free. All of it could be funded by the companies who right now are willing to pay a substantial premium for power that can be online quickly enough to matter in the current AI race. Whether or not they are right about the stakes in that race, more generation here would be valuable.
The opportunity is real, but the largest part of it is time-limited. The government should accelerate its regulatory review and set a shorter statutory clock for decisions on co-located datacentres and the generation that powers them.
Dr Eric Crampton is Chief Economist at the New Zealand Initiative. This article was sourced HERE
But some things that sound too good to be true are, nevertheless, real.
I think there is a very brief window in which mostly-American hyperscale datacentres could massively expand renewable energy generation in New Zealand, at their own expense, to the country’s much broader benefit.
I worry, though, that the country will miss the opportunity.
A little over two years ago, Leopold Aschenbrenner explained the basic logic of international competition in artificial intelligence. The owner of an artificial intelligence competent enough at programming to help build its own successor will have a durable advantage over others that follow.
Being a month behind the competition matters a lot if your competitor was able to run a few hundred thousand, or more, capable AI programmers to help it make progress during that month.
Companies have been betting on that scenario, as well as general AI uptake and diffusion, for a few years now. The pace of datacentre buildout has been astonishing. A large portion of current US economic growth is investment in datacentres and the things they need.
Tales of datacentres’ water requirements are largely urban legend, or at least depend on not really understanding industrial and agricultural-scale water use. Compared to a household, they’re large. Compared to a meatworks or an irrigated paddock, they’re not that large.
But one of the things they really do need is energy. Lots of it. Large datacentres are measured by the amount of energy their compute requires.
America’s electricity grid cannot keep up. Getting a connection to the grid now takes between three and seven years. And the queue of projects wanting connection represents more power than the entire country’s current generation capacity.
Not falling behind competitors really matters. So the frontier labs want more compute and they want it right now. If a hyperscale datacentre (really large, in lay-terms) cannot get a grid connection, it will set up its own power generation.
The kinds of generation that datacentres can build on-site in a hurry are not particularly efficient. Their electricity is expensive, compared to the power generated by the kinds of powerplants that electricity utilities build. But some of them can be built in as little as 18 months, which is much better than waiting up to seven years for a connection to the electricity grid.
You may remember Caterpillar as the maker of the lovely large yellow equipment that toddlers young and old love to see working. They’ve also made generators for off-grid construction projects. They’re now massively scaling that side up and are delivering gigawatts of natural gas generator sets to datacentres.
Hyperscalers are willing to pay a lot if they can get access to energy in a hurry, without large fishhooks along the way.
Two years ago, Aschenbrenner’s logic led me to one obvious conclusion. A country that could quickly say yes to a proposed geothermal plant sited next to a new datacentre could attract billions of dollars of investment, very quickly.
Over the past two years, the scale of that opportunity has increased considerably.
The stakes have also increased. For a relatively short period, the American government restricted access to a leading model, allowing only American citizens to use it. It would be surprising if the American government did not impose more substantial export restrictions on its frontier models over the coming years. If American companies have access to frontier models and New Zealand companies do not, the country’s productivity challenges will worsen.
Having local datacentres able to run the best open-source models would provide some insurance. If no one builds them now, a future government may well want to build them itself, at taxpayers’ expense.
New Zealand has enormous amounts of untapped conventional geothermal energy. The country’s electricity generators know how to build geothermal plants. The broader Taupo volcanic zone is, with Iceland, one of the two best places in the entire world for both conventional geothermal energy and the next generation of ‘supercritical’ geothermal energy.
A clear, rapid, and certain consenting path for new datacentres and the energy that powers them, whether geothermal, solar or otherwise, able to deliver decisions in a couple of months, could see enormous investment in very short order.
The consenting path is only part of it. In their submissions on the Government's draft geothermal strategy, Contact, Mercury and Eastland told MBIE that the Geothermal Energy Regulations 1961 add time and cost from consenting through to operation. The rules conflict with newer health and safety law. Pipeline requirements apply a standard written for long-distance hydrocarbon transmission. And regulations drafted in 1961 never anticipated supercritical geothermal at all. The law has no category for the thing the country is best placed to build.
Contact asked the Government to accelerate its planned review of the rules rather than wait until next year, and to run an end-to-end review with industry. On the Government's current schedule, the review would have little chance of finishing before the window to attract investment starts closing.
At the scale at which these investments are made internationally, the New Zealand government’s current support for developing next-generation supercritical geothermal generation looks like spare change. The companies that would fund the energy and datacentre buildout could very plausibly fund a hastening of current work on next-generation geothermal.
If the work on supercritical geothermal pans out, the country would have near-limitless, low-cost electricity for generations to come. Some of it would power datacentres; the rest of it could power everything else.
This isn’t an urban legend like the one about the Army Corps of Engineers that would have built New Zealand highways for free. All of it could be funded by the companies who right now are willing to pay a substantial premium for power that can be online quickly enough to matter in the current AI race. Whether or not they are right about the stakes in that race, more generation here would be valuable.
The opportunity is real, but the largest part of it is time-limited. The government should accelerate its regulatory review and set a shorter statutory clock for decisions on co-located datacentres and the generation that powers them.
Dr Eric Crampton is Chief Economist at the New Zealand Initiative. This article was sourced HERE
1 comment:
If you really think our electricity generators have any interest in supplying cheap, plentiful electricity I have a bridge for sale.
Although the Iwi are certainly looking forward to the use of their geothermal property and the massive payouts they will rort from the rest of us.
Post a Comment
Thank you for joining the discussion. Breaking Views welcomes respectful contributions that enrich the debate. Please ensure your comments are not defamatory, derogatory or disruptive. We appreciate your cooperation.