Craig Scroggie

Well, right now, the scale of demand for artificial intelligence is greater than the world's ability to be able to meet that with power, the energy required and the data center physical build out.

So there is more demand to host GPU infrastructure than there is physical energy and buildings constructed today.

It would be at least an order of magnitude greater, Alan, than what we've seen at any other point in time in history.

So a big data center used to be 20, 30, 40 megawatts only five years ago.

A big data center today will be 100, 200, 300 and going to very soon be in excess of 1,000 megawatts of power.

So you're talking about the size of a small city.

Well, depending on the type of data centre you build, if you're building to support the enterprise, so banks and online, you build them close to the city, they're low latency.

If you're in a hospital and your data is being accessed, medical systems, you want the computers close to the user.

Whereas in artificial intelligence, we can go to more remote regions or further out from cities where power is available and allow us to really do a couple of things.

One is to build flat, so single-story buildings, whereas when we build in metropolitan areas, you build multi-story buildings to get more efficient use out of the land.

so as artificial intelligence becomes a much larger feature of the digital infrastructure landscape they will be built very large and low lying and they're actually built in quite a different way they're not built necessarily with the same security or resiliency requirements because artificial intelligence has a slightly different use case in the production of tokens so yes they do take up a significant amounts of land they can be built in

areas further out from cities or in regional areas.

And predominantly, you'll see them focused around areas where we can secure large-scale energy or connectivity to the NEM so that access to transmission and generating infrastructure is a key feature of being able to produce thousands of megawatts of power to run artificial intelligence.

Did you have a problem with any of them?

Well, they said largely what was already the case for the industry.

Australia is a little bit different to the US.

And so if you see a lot of the headlines in the US and the media at the moment, many of the questions are in relation to developers of infrastructure using community resources.

In Australia, and as long as we've been building data centers for 15 years, we've always had to pay for our own substation infrastructure, transmission to connect to electricity market, and then the ability to be able to sponsor through PPAs or power purchase agreements, financial instruments, the development of new renewable energy that goes into the grid.

So the release of those guidelines was really just the government meeting or re-communicating what was already legislated and already the case in Australia for the developers of data centre and other digital infrastructure assets.

Well, in the main, the majority of things that the government talked about in those requirements were giving the public confidence that these