#124 CommsDay Summit panel: Building Australia's digital infrastructure

WEBVTT
00:00:00.000 --> 00:00:12.400
Music.
00:00:12.291 --> 00:00:17.451
And welcome to the show. Today, we're going to be revisiting a panel session
00:00:17.451 --> 00:00:20.071
from Comms Day Summit earlier this month.
00:00:20.251 --> 00:00:27.051
This was a panel session looking at the issues around the construction of Australia's
00:00:27.051 --> 00:00:31.091
digital infrastructure, particularly with the view to making it AI ready.
00:00:31.511 --> 00:00:38.011
The panel featured Brookfield Global Data Center CEO, Uday Mathielagan,
00:00:38.871 --> 00:00:43.551
IberConnect CEO Mark Rafferty, Equinix Australia MD Guy Danskin,
00:00:44.411 --> 00:00:50.851
NextDC Chief Operating Officer Simon Cooper and the MC, Steve Kinane from the ABC.
00:00:51.851 --> 00:00:56.491
So this panel is going to be all about building Australia's digital infrastructure.
00:00:56.791 --> 00:01:01.191
The panel will examine the forces reshaping digital infrastructure investment
00:01:01.191 --> 00:01:02.791
across Australia and New Zealand.
00:01:03.011 --> 00:01:07.731
We'll begin by exploring how macroeconomic pressures are impacting build decisions
00:01:07.731 --> 00:01:10.451
and capital flows. We'll then turn to AI.
00:01:10.991 --> 00:01:15.491
Then the panel will also tackle concerns around overbuild and capital discipline
00:01:15.491 --> 00:01:16.931
in the data centre market.
00:01:17.111 --> 00:01:20.991
And we're also going to discuss the biggest barriers to project delivery and
00:01:20.991 --> 00:01:26.011
then move on to the policy settings needed to secure Australia's digital infrastructure future.
00:01:26.491 --> 00:01:30.531
So I'm first of all going to get each of the panellists to introduce themselves.
00:01:31.151 --> 00:01:34.391
Good afternoon, final session of the two days. Amazing room.
00:01:34.591 --> 00:01:36.771
Mark Rufferty, CEO of Faber Connects.
00:01:37.731 --> 00:01:45.091
Oh, thank you. Good afternoon. Thank you for the invite. My name is Simon Cooper. I'm the COO at NextDC.
00:01:46.011 --> 00:01:50.871
Good afternoon. My name is Uday Mathialgan. I'm managing partner at Brookfield,
00:01:50.871 --> 00:01:54.191
and I'm also the CEO of our global data center platform.
00:01:55.191 --> 00:01:58.451
I'm Simon Debron Goydansk and managing director at Ecranks, and I commend you
00:01:58.451 --> 00:02:01.411
all for being in the room at 4.30 on the last day of the conference. So well done.
00:02:02.831 --> 00:02:07.211
Okay. So the first one we're going to tackle is how are macroeconomic disruptions?
00:02:07.411 --> 00:02:12.531
And by that, I'm including Trump's tariffs, rising costs of capital exchange rate movements.
00:02:12.711 --> 00:02:17.551
How are they influencing the digital infrastructure market in Australia and
00:02:17.551 --> 00:02:19.491
New Zealand? Mark, maybe you want to kick that one off.
00:02:20.111 --> 00:02:25.691
Me first. I'd say the only real impact in our market is during construction.
00:02:25.691 --> 00:02:28.111
We started FibreConnects in 2019.
00:02:28.591 --> 00:02:32.771
We've seen the labour costs and obviously the cost of quality assurance of that
00:02:32.771 --> 00:02:35.951
construction increase by over 40%.
00:02:36.691 --> 00:02:41.651
And it's sort of been a gradual increase and we've seen a lot of pressure on
00:02:41.651 --> 00:02:43.291
the construction companies in solvency.
00:02:43.511 --> 00:02:48.051
So we've had four major constructors go into liquidation in the last 12 months,
00:02:48.451 --> 00:02:50.071
focusing in our markets.
00:02:50.671 --> 00:02:53.271
Our biggest concerns are actually keeping those costs under control,
00:02:53.471 --> 00:02:55.651
but more importantly, the quality assurance process of that.
00:02:55.791 --> 00:02:57.711
So ours more impacted through labor costs.
00:02:58.451 --> 00:03:00.051
Simon? Thank you, Mark.
00:03:00.911 --> 00:03:06.851
Yeah, to carry on with that thread, sort of post-COVID price pressure has been significant.
00:03:07.251 --> 00:03:10.831
And on top of that, and I think we've probably seen the same in the fiber industry,
00:03:11.331 --> 00:03:15.211
the volume of work and maybe to some extent in the data center,
00:03:15.311 --> 00:03:18.351
the volume of possible work has gone through the roof as well.
00:03:18.471 --> 00:03:19.831
I think that's contributed to it.
00:03:20.071 --> 00:03:24.891
But ultimately, managing your costs and focusing on that hasn't really prevented
00:03:24.891 --> 00:03:26.791
us from continuing our development.
00:03:27.051 --> 00:03:33.111
On the supply chain side, we've gone through an incredible tightening where
00:03:33.111 --> 00:03:35.411
we were hearing things that would
00:03:35.411 --> 00:03:40.211
normally take three or four months to deliver could be 18 or 24 months.
00:03:40.431 --> 00:03:43.651
That's almost gone all the way back to normal.
00:03:44.631 --> 00:03:49.791
So that is what it is. These are regular cycles. I think with respect to interest
00:03:49.791 --> 00:03:55.471
rates and tariffs and so on, I think Australia is in a position where it's kind
00:03:55.471 --> 00:03:59.491
of somewhat uniquely insulated from some of that.
00:03:59.731 --> 00:04:03.891
I've not seen anything that's tariff related that's driven different decision
00:04:03.891 --> 00:04:05.911
makings from my perspective at this point.
00:04:06.131 --> 00:04:09.031
Maybe that'll change in the next six months as it really flows through.
00:04:09.271 --> 00:04:13.411
And from an interest rate perspective, you know, we deal with that as a business.
00:04:13.411 --> 00:04:14.731
I guess there was a number of
00:04:14.731 --> 00:04:18.891
years where we didn't have to because they were low and now we do have to.
00:04:19.331 --> 00:04:24.171
So I'm hoping we continue on that downward trend, but we'll continue to deal with it.
00:04:24.331 --> 00:04:29.211
So really, it comes back to opportunity and then just dealing with local cost
00:04:29.211 --> 00:04:33.311
and resource challenges as there's lots going on both in our industry and others.
00:04:34.685 --> 00:04:38.985
Yeah, look, I think Ian Martin touched on it a bit earlier. Cost of capital
00:04:38.985 --> 00:04:40.425
is not necessarily going up.
00:04:40.805 --> 00:04:47.385
I think the issue is more around factors with supply chain and construction
00:04:47.385 --> 00:04:50.525
costs a lot, and maybe just a global perspective.
00:04:50.845 --> 00:04:54.625
And look, there's a lot of discussion on tariffs, but the tariff thing is,
00:04:54.825 --> 00:04:59.625
to a large extent, it depends on whether it's related to the US or not.
00:04:59.625 --> 00:05:02.585
So when you look at the supply chain, let's break it down, say,
00:05:02.705 --> 00:05:07.705
data centers or fiber networks, a lot of the componentry comes from other markets as well.
00:05:07.885 --> 00:05:10.825
So it captures the other things. And also, it's not just tariffs.
00:05:10.985 --> 00:05:14.745
It's export controls, which could have an impact as part of a trade war.
00:05:15.345 --> 00:05:19.985
Yeah, true of that. And if you're going there, so clearly the world is almost
00:05:19.985 --> 00:05:24.605
separating into three different, I wouldn't call them blocks necessarily,
00:05:24.605 --> 00:05:27.305
but you've got the whole U.S. ecosystem.
00:05:27.825 --> 00:05:31.125
The Chinese ecosystem is very strong in certain markets.
00:05:31.265 --> 00:05:34.985
And there's a whole bunch of countries in the middle, particularly when you
00:05:34.985 --> 00:05:38.565
start thinking about and talking about sovereign AI capability,
00:05:38.805 --> 00:05:42.765
who are starting to get more concerned and worried in terms of where should
00:05:42.765 --> 00:05:44.665
they source their inputs.
00:05:44.885 --> 00:05:47.145
And so I think we can talk a little more about AI in a minute,
00:05:47.305 --> 00:05:51.765
but just on your initial points, I think cost of capital itself is not such a big issue,
00:05:52.025 --> 00:05:55.345
but it's the risk that And when you think about the Australian context,
00:05:55.545 --> 00:06:00.685
I think construction costs have become more unpredictable and supply chain is
00:06:00.685 --> 00:06:04.145
still, you know, if you don't have it stitched up, you've got some challenges.
00:06:04.205 --> 00:06:07.265
So I think it's worth managing the risk and delivering large,
00:06:07.365 --> 00:06:09.785
complex projects. All right, Guy, what are your thoughts on this?
00:06:10.585 --> 00:06:13.745
Yeah, on the export controls, I'm not sure how familiar people are with the
00:06:13.745 --> 00:06:17.125
AI diffusion rule, which was retired under the Trump administration.
00:06:17.825 --> 00:06:21.065
So while it's been retired, it's actually pretty indicative of how the US government
00:06:21.065 --> 00:06:23.985
thinks about Australia in terms of a strategic ally in this region.
00:06:24.225 --> 00:06:27.725
And so as we think about digital infrastructure in Australia and New Zealand, I think it.
00:06:28.558 --> 00:06:31.418
It's one signal that we're very well placed from a strategic position.
00:06:31.678 --> 00:06:34.358
I agree with the panelists largely on tariffs.
00:06:34.558 --> 00:06:38.998
We're a global organization and we see opportunity potentially there.
00:06:39.158 --> 00:06:43.838
For example, we've got about $500 million of components on pre-buy that we've
00:06:43.838 --> 00:06:45.138
got strategically placed around.
00:06:45.478 --> 00:06:48.778
So there's opportunity there, but also risk. Just a quick follow-up about that
00:06:48.778 --> 00:06:52.018
guy, because in the free trade agreement between the US and Australia,
00:06:52.198 --> 00:06:54.898
they don't mention digital services.
00:06:55.638 --> 00:07:00.778
But Jamison Greer, when he was being questioned in the Congress by Mark Warner,
00:07:00.898 --> 00:07:05.558
who was getting very fired up on behalf of Australia, did talk about barriers
00:07:05.558 --> 00:07:10.238
to trade in the digital space in Australia. Do you think that could become an issue?
00:07:10.658 --> 00:07:16.478
I think it is an issue. I was in the US in March, and the US government has
00:07:16.478 --> 00:07:20.498
a particular view that we have trade barriers here in terms of how we think
00:07:20.498 --> 00:07:23.178
about taxation of American technology companies.
00:07:23.178 --> 00:07:26.958
And then, of course, in Australia here, we've got the digital tax we think about
00:07:26.958 --> 00:07:28.338
on those similar companies.
00:07:28.518 --> 00:07:34.198
So I think both countries see some potential trade conflict there in terms of non-tariff barriers.
00:07:34.558 --> 00:07:38.118
Yeah. Okay. Guy, while you've got the mic, I'll come to the second question about AI.
00:07:38.738 --> 00:07:42.138
How is AI reshaping business models and investment strategies?
00:07:42.338 --> 00:07:45.958
Are we looking at game-changing innovations or just another tech trend?
00:07:46.538 --> 00:07:50.178
That's not another tech trend. I hope not because otherwise a lot of us are out of a job.
00:07:51.078 --> 00:07:53.538
Yeah, look, I think we're through that hype cycle, you know,
00:07:53.638 --> 00:07:57.458
the Gartner hype cycle in terms of, you know, we're two years in here.
00:07:57.558 --> 00:08:00.378
And so really it's now the adoption and the real use cases of AI.
00:08:00.638 --> 00:08:03.698
And the thing that's exciting, certainly for our industry, and I think a lot
00:08:03.698 --> 00:08:07.138
of the folks in this room over the last couple of days, is that as we move into
00:08:07.138 --> 00:08:11.178
inferencing and the carriage and network that's required to drive that either
00:08:11.178 --> 00:08:13.418
from a mobile perspective, a sub-sea perspective,
00:08:13.978 --> 00:08:17.658
long-haul networking, you know, consumption is driven through connectivity.
00:08:18.018 --> 00:08:21.298
And I think for the folks here at Comms Day, I think that's pretty exciting.
00:08:22.018 --> 00:08:23.458
Uday, do you want to follow up on that one?
00:08:24.544 --> 00:08:30.924
Probably a couple of comments on AI. Clearly, I think it's a general purpose sort of technology.
00:08:31.064 --> 00:08:33.984
People have given lots of different language to it. I won't add to it,
00:08:34.124 --> 00:08:38.924
but I'll give you two real examples in terms of what's happening in our firm.
00:08:38.924 --> 00:08:43.324
I mean, we own dozens of businesses across our infrastructure and private equity
00:08:43.324 --> 00:08:49.184
practice, and we've got 300 live projects just in terms of how AI can change
00:08:49.184 --> 00:08:51.704
business processes and a whole bunch of stuff.
00:08:51.824 --> 00:08:55.004
So, you know, I'm sure the same thing's happening across the board.
00:08:55.004 --> 00:08:59.324
So it's a huge impact. But with an investor hat on, particularly around digital
00:08:59.324 --> 00:09:04.084
infrastructure, we're seeing, and back to my earlier point, there are countries
00:09:04.084 --> 00:09:07.224
that are really taking sovereign AI seriously.
00:09:07.684 --> 00:09:11.224
And so the opportunity to invest has expanded.
00:09:11.524 --> 00:09:15.044
So we've made two announcements recently.
00:09:15.044 --> 00:09:19.004
So we're working with the French government who've announced a 100 billion euro
00:09:19.004 --> 00:09:21.584
program to create a sovereign AI capability.
00:09:21.584 --> 00:09:25.344
We're going to contribute $20 billion of that, primarily through our data center
00:09:25.344 --> 00:09:29.064
business called Data4, but it's actually a very large campus in the northern
00:09:29.064 --> 00:09:32.604
part of France, which is quite different from a typical data center.
00:09:32.764 --> 00:09:35.724
It's going to bring together industry in a different way.
00:09:36.044 --> 00:09:39.204
Similar announcement in Sweden recently. So I think, you know,
00:09:39.244 --> 00:09:40.164
I'm just posing the question.
00:09:40.324 --> 00:09:43.304
I think there'll be countries that decide that they're going to have to take
00:09:43.304 --> 00:09:48.024
control of some elements of their AI when it comes to sovereignty.
00:09:48.164 --> 00:09:51.664
And are you seeing that here yet? But I think it's playing out.
00:09:51.904 --> 00:09:54.684
You know, I think it's, look, I think Australia is blessed with,
00:09:54.824 --> 00:10:00.224
we've got lots of data-centric, you know, good companies, great infrastructure.
00:10:00.584 --> 00:10:03.844
But I think it's, the question's out there.
00:10:04.004 --> 00:10:08.044
It's like, to your earlier point as well, as public policy changes,
00:10:08.404 --> 00:10:11.884
how much of it do you really want to truly control?
00:10:12.104 --> 00:10:15.184
You know, so I think this is a dynamic that's playing out every day in front
00:10:15.184 --> 00:10:18.584
of us, you know, in terms of how different countries are reacting.
00:10:19.304 --> 00:10:24.504
Simon, how do you see this space evolving? Yeah, so I think innovation is really the key thing here.
00:10:24.664 --> 00:10:30.004
And not to sort of talk to the innovation that NVIDIA or the inventor of the
00:10:30.004 --> 00:10:36.464
Tensor equivalent has driven, but just kind of how business is going to get created and driven.
00:10:36.764 --> 00:10:43.504
So to your point, I don't think at a federal level, it's really been gotten its arms around yet.
00:10:43.684 --> 00:10:47.984
But we do have land, we've got the capability to grow the power supply.
00:10:47.984 --> 00:10:52.884
And we're a nation that punches well above its weight in terms of technology adoption.
00:10:53.164 --> 00:10:57.944
So it's coming. We did an announcement last week about another site that we're
00:10:57.944 --> 00:11:00.724
going to be developing in central Melbourne, near Port Melbourne,
00:11:01.064 --> 00:11:04.944
with a significant amount of state-level government involvement.
00:11:05.204 --> 00:11:07.184
And that's great for a couple of reasons.
00:11:07.684 --> 00:11:12.984
Obviously, the subtle changes that are required in terms of planning alignment are important.
00:11:13.184 --> 00:11:18.124
But more importantly, from my perspective, and I think from the whole industry's perspective, to see.
00:11:19.134 --> 00:11:21.814
Politicians and their underlying sort of
00:11:21.814 --> 00:11:25.014
planning and thinking actually kind of get behind
00:11:25.014 --> 00:11:32.594
a growth opportunity and figure out how to enable it to actually grow and then
00:11:32.594 --> 00:11:38.414
how to absorb it into the sort of the local infrastructure so that it does benefit
00:11:38.414 --> 00:11:42.694
and create jobs and create opportunity as opposed to the other way around.
00:11:42.834 --> 00:11:47.214
And I would say 18 months or 24 months ago, there was a real danger here where
00:11:47.214 --> 00:11:50.414
there was, you know, stop signs going to get chucked up because nobody knew
00:11:50.414 --> 00:11:52.954
how to cope. And we'd seen that in Ireland.
00:11:53.154 --> 00:11:55.674
We've seen that in Singapore. It's a cycle, you get through it.
00:11:55.774 --> 00:11:58.594
But I'm really excited that we're on that journey.
00:11:58.894 --> 00:12:01.754
Lots of problems to solve, but we seem to be getting genuine support.
00:12:02.334 --> 00:12:05.274
Mark, how are you feeling about it? How are we feeling about AI?
00:12:06.634 --> 00:12:13.054
Look, obviously, our biggest sort of focus area at the moment is obviously we
00:12:13.054 --> 00:12:15.354
build conduit networks that connect all the data centers.
00:12:15.714 --> 00:12:18.894
We have to be very considerate as into the architecture of that.
00:12:19.074 --> 00:12:23.334
I mean, you can't go and rip up everyone's front lawn for three years and then
00:12:23.334 --> 00:12:26.054
say, well, I've got it wrong. I have to actually change this architecture.
00:12:26.754 --> 00:12:30.894
So we've seen the architecture requirements evolve in preparation,
00:12:31.574 --> 00:12:32.774
especially Sydney and Melbourne.
00:12:33.074 --> 00:12:38.514
But where a lot of our focus is now is, one, the capacity going into these data
00:12:38.514 --> 00:12:40.694
centers for what we believe the data centers are going to be used for,
00:12:40.914 --> 00:12:42.974
but how a customer is going to connect to it.
00:12:43.214 --> 00:12:47.654
Yeah. So we are building a brand new fiber fabric across the country and we
00:12:47.654 --> 00:12:50.794
are building to all the corporate buildings in a predefined way.
00:12:50.974 --> 00:12:55.194
The way the stock exchange has connected any customer to its trading platform
00:12:55.194 --> 00:12:58.654
for over 20 years is focused on latency, direct point of sight.
00:12:58.774 --> 00:13:02.794
And I think the inferencing requirements and how fiber is connected to those
00:13:02.794 --> 00:13:07.554
inferencing requirements to customers is something the current architecture of Fibre doesn't do.
00:13:08.354 --> 00:13:13.034
You have to get that capacity correct. You can't build this three times.
00:13:13.274 --> 00:13:15.954
You have to actually build it once in preparation.
00:13:16.234 --> 00:13:20.154
We think we're probably about three years from seeing some real inferencing
00:13:20.154 --> 00:13:22.294
products coming into the corporate market,
00:13:22.474 --> 00:13:25.894
into the enterprise sort of buildings or in customers, but we're spending a
00:13:25.894 --> 00:13:28.854
lot of our time sort of talking with partners that we believe are going to be
00:13:28.854 --> 00:13:34.394
managing that product and those inferencing zones for their end customers to
00:13:34.394 --> 00:13:40.334
make sure that that connectivity is not the issue and is an assistance to guaranteeing that service.
00:13:40.794 --> 00:13:43.914
Simon, you mentioned problems to solve. What are the problems you're seeing?
00:13:44.614 --> 00:13:51.574
So there's new supply chain problems. So let's just focus, for instance, on liquid chip.
00:13:52.214 --> 00:13:58.634
Okay. It's great. It's going to enable us to build less data centers to support way more compute.
00:13:59.482 --> 00:14:03.802
All the other challenges are still there. Power, how you reject heat from the
00:14:03.802 --> 00:14:06.302
data center, et cetera, et cetera. Great.
00:14:06.982 --> 00:14:11.182
Now, to deliver that liquid to chip into the architecture that's been selected
00:14:11.182 --> 00:14:15.942
by the likes of NVIDIA, which is essentially the tertiary liquid flows into
00:14:15.942 --> 00:14:19.822
the device and back out, you've got a hygiene challenge there.
00:14:19.942 --> 00:14:21.122
You've got flow rate challenges.
00:14:21.462 --> 00:14:27.162
And what we're seeing is we're moving what was traditionally a deployment in
00:14:27.162 --> 00:14:30.822
the mechanical space, and there were several subcontractors here in Australia
00:14:30.822 --> 00:14:31.982
that were very, very good at it,
00:14:32.262 --> 00:14:37.042
you've now got to a point of hygiene and stainless steel and so on that we're
00:14:37.042 --> 00:14:40.162
finding, oh, the food industry can do this. Can they cope with the workload?
00:14:40.422 --> 00:14:43.602
So that's a challenge that will be solved.
00:14:43.802 --> 00:14:48.682
You've got the power side where our understanding is, and we've seen this in
00:14:48.682 --> 00:14:52.302
some of our conversations with our client base and some of the research that's done,
00:14:52.942 --> 00:14:59.802
that a 100-megawatt compute deployment, which in cloud world would slowly ramp,
00:15:00.122 --> 00:15:04.862
get deployed in incremental periods, and many, many customers will be using it quite differently,
00:15:05.462 --> 00:15:08.542
it could be entirely synchronized at a data center.
00:15:09.002 --> 00:15:12.202
And then depending on the particular problem or piece of work that it's doing,
00:15:12.442 --> 00:15:15.722
could all ramp up and all ramp down. And we're talking in milliseconds.
00:15:16.102 --> 00:15:20.382
Because we have renewable energy targets for 2030, right? That's a different thing.
00:15:20.822 --> 00:15:23.562
That's a different thing. Is that going to make it harder to hit then?
00:15:24.282 --> 00:15:29.442
Personally, I think it should make it easier because growth in demand creates
00:15:29.442 --> 00:15:34.122
an opportunity to invest in power generation into a growth market, not a replace market.
00:15:34.442 --> 00:15:39.282
Now, there's tons of detail underneath that. But for me, at least that's a better thing.
00:15:39.782 --> 00:15:43.502
System strength, which is what this is talking about, is a challenge.
00:15:43.782 --> 00:15:47.282
It'll be a challenge in any market eventually. So it'll be solved.
00:15:47.502 --> 00:15:50.882
There are solutions to it. It's about thinking it through. And the last thing
00:15:50.882 --> 00:15:55.622
I would say is that the hyperscalers are leading, but very differently.
00:15:56.222 --> 00:16:00.222
So as we speak to one, they're almost ready and they tell us things they've
00:16:00.222 --> 00:16:02.882
done when they've deployed to help us learn and deploy for them.
00:16:03.022 --> 00:16:06.082
We speak to another and they change their mind at the last minute and say,
00:16:06.182 --> 00:16:09.282
no, we're going to stick with the air-cooled GVUs for now because we've not
00:16:09.282 --> 00:16:10.502
quite made our decision yet.
00:16:10.642 --> 00:16:13.102
So there's a little bit of turbulence there, if you like.
00:16:13.302 --> 00:16:17.562
But more exciting, I think, from everyone's perspective, is you've got this
00:16:17.562 --> 00:16:22.042
NeoCloud or the NCPs, as they were known. And it's remarkable to me that the
00:16:22.042 --> 00:16:26.022
infrastructure as a service layer has almost instantly re-emerged.
00:16:26.202 --> 00:16:30.982
And you've got CoreWeave and probably seven or eight others not of that scale,
00:16:31.142 --> 00:16:33.542
but that are capable of getting there, having conversations.
00:16:33.902 --> 00:16:38.022
So, yeah, there's a huge opportunity and lots of problems to solve in a good way. Okay.
00:16:38.502 --> 00:16:42.222
Uday, I want you to kick off on our third topic. Is there a risk of overbuild,
00:16:42.442 --> 00:16:43.942
overinvestment for DCs?
00:16:44.022 --> 00:16:47.282
How does a platform operator differentiate themselves in the market?
00:16:47.282 --> 00:16:53.022
With so much money flowing, are investors doing the right deals or is this a case of FOMO?
00:16:53.730 --> 00:16:59.630
I might take a counter view to that point. I just don't think there's almost
00:16:59.630 --> 00:17:03.970
enough money to build out all the data set to capacity that will be required in time.
00:17:04.130 --> 00:17:08.690
That doesn't stop speculation from happening in the interim.
00:17:08.850 --> 00:17:14.690
When you add up all the numbers, the amount of capacity required is quite large.
00:17:14.830 --> 00:17:17.310
And to Simon's point earlier, the market has grown quite a bit.
00:17:17.410 --> 00:17:20.890
So we used to talk about hyperscalers. Now, you've got the neoclouds and you've
00:17:20.890 --> 00:17:26.030
got, you know, the whole range of, you know, just AI platforms who themselves
00:17:26.030 --> 00:17:28.070
are going directly to procure capacity.
00:17:28.230 --> 00:17:32.350
So macro level demand is very high, a lot of industry constraints.
00:17:32.750 --> 00:17:37.970
But capital itself, you know, like if you use this rough math,
00:17:38.450 --> 00:17:43.950
$10 million US per megawatt, it'll change a little bit depending on the geography.
00:17:44.290 --> 00:17:48.610
For every 100 megawatts, it's a billion dollars. and 100 megawatts is nothing these days.
00:17:48.830 --> 00:17:53.010
You touch a campus, we're building campus hundreds of megawatts each.
00:17:53.190 --> 00:17:55.170
So each project is worth a few billion.
00:17:55.390 --> 00:17:59.750
For every billion we put in, our customers to put in three to four billion in
00:17:59.750 --> 00:18:02.810
terms of chips and liquid cooling and whatever else.
00:18:02.990 --> 00:18:05.490
So there's quite a lot required sort of macro level.
00:18:05.710 --> 00:18:10.870
So I think there's a parallel development taking place in the financial world where.
00:18:11.600 --> 00:18:17.360
You know, more sophisticated capital is required at different stages of development
00:18:17.360 --> 00:18:18.340
of digital infrastructure.
00:18:18.380 --> 00:18:20.460
So I think this is, this will evolve over time.
00:18:20.600 --> 00:18:24.620
So you're going to start seeing even large operators selling down different
00:18:24.620 --> 00:18:27.700
pieces and parts of their businesses over time, different answers.
00:18:27.800 --> 00:18:31.620
So I wouldn't say there's too much money flowing in, but clearly there's quite
00:18:31.620 --> 00:18:35.420
a lot of speculation, particularly those who have land, who sometimes,
00:18:35.540 --> 00:18:39.800
you know, take a little bit of power and, you know, that's kind of dressed up as almost a data center.
00:18:39.980 --> 00:18:43.360
There's a long, way to go from left to right to achieve that.
00:18:43.480 --> 00:18:47.800
But more capital at the right stage of development is going to be required than
00:18:47.800 --> 00:18:50.320
is available today in the market. Guy, what's your view on that?
00:18:50.840 --> 00:18:54.620
Yeah, look, similar to Udo, I think the demand signals are incredibly strong
00:18:54.620 --> 00:18:58.640
at the moment, and there is no shortage of capital coming into our space.
00:18:58.840 --> 00:19:01.720
I was interested in the CommScope gentleman yesterday. I don't know if folks
00:19:01.720 --> 00:19:04.720
caught that, but he was talking about from a supply chain perspective,
00:19:04.980 --> 00:19:06.600
their concern about there being
00:19:06.600 --> 00:19:09.520
either boom or bust in our space and then trying to hedge against that.
00:19:09.980 --> 00:19:13.240
The thing I would say is, you know, Equinix has been operating for 25 years
00:19:13.240 --> 00:19:14.680
here in Australia and around the world.
00:19:15.060 --> 00:19:18.280
Folks that are now coming up with plots of land and a little bit of energy,
00:19:18.400 --> 00:19:22.760
as Udo said, there's a long way to go from there to operating a data center,
00:19:22.940 --> 00:19:25.740
mission critical facility, I should say.
00:19:25.940 --> 00:19:29.540
And I do think that there is some speculative investment on some of that,
00:19:29.660 --> 00:19:32.140
that, you know, I think some caution would be advised.
00:19:32.400 --> 00:19:35.220
But also, as earlier said, in Australia, we're very fortunate.
00:19:35.400 --> 00:19:37.100
There's a handful of very credible
00:19:37.100 --> 00:19:40.420
established operators so there's no shortage of quality supply here.
00:19:40.580 --> 00:19:45.000
I think as you go out into Asia and elsewhere, that starts to diminish in terms of quality.
00:19:45.580 --> 00:19:50.880
Mark, any risk of overbuild or overinvestment? I'm just a telco, mate. No.
00:19:52.040 --> 00:19:57.040
Look, I think on the capital conversation, we're in an interesting time in the
00:19:57.040 --> 00:19:58.760
market just in general. So, we've got the luxury.
00:19:58.980 --> 00:20:03.700
We have to build every data center operator. So, I think there's 32 in Sydney
00:20:03.700 --> 00:20:06.040
at the moment getting built. There's about 12 in Melbourne.
00:20:06.520 --> 00:20:10.600
We have to be in all of them. We know what all their plans are and some of them
00:20:10.600 --> 00:20:12.840
might have slowed, but they're definitively getting built.
00:20:13.740 --> 00:20:16.860
They're not all going after the same type of customers. Yeah,
00:20:16.900 --> 00:20:19.480
I think you're going to be left with about three or four real specialists in
00:20:19.480 --> 00:20:21.000
the country that will do that.
00:20:21.140 --> 00:20:26.480
In my world, I don't think there's enough space for three or four people trying
00:20:26.480 --> 00:20:29.560
to build a new core data center connectivity network.
00:20:29.600 --> 00:20:34.140
I think we're running an averaged capital across all regions,
00:20:34.280 --> 00:20:38.720
across all products. We can't look at a data center on its own merits to say,
00:20:38.800 --> 00:20:41.960
will we get enough business out of that one data center, which is risky.
00:20:42.080 --> 00:20:43.300
It's very, very risky for us.
00:20:43.440 --> 00:20:48.140
So, but we can't be half pregnant and only in a certain amount of data centers at the same time.
00:20:48.400 --> 00:20:52.280
So our biggest concern when we see all the data centers being built is who's
00:20:52.280 --> 00:20:55.160
going to go in them. Yeah. Are they going to be a single tenant?
00:20:55.792 --> 00:21:00.312
And so if you don't win that customer, you have wasted your money building to that network.
00:21:00.612 --> 00:21:05.212
Is the future going to be that customers will be required to go to that data center?
00:21:05.492 --> 00:21:10.272
So like our two highest routes are between NextDC and Equinix,
00:21:10.792 --> 00:21:13.752
you know, individual services and things like that, but our highest capacity
00:21:13.752 --> 00:21:15.132
are between the self-built.
00:21:16.012 --> 00:21:21.292
Simon, what do you see are the biggest logistical challenges for digital infrastructure builders?
00:21:22.172 --> 00:21:25.552
So just joining on to the conversation we just had, I think,
00:21:25.712 --> 00:21:29.372
I mean, there are markets where overbuild may have happened.
00:21:29.532 --> 00:21:33.792
So if we look at the Johor Bahru area, just north of Singapore there,
00:21:33.992 --> 00:21:38.972
we're starting to see data centers that were mostly built, sort of looking for
00:21:38.972 --> 00:21:40.492
change of ownership, for instance.
00:21:40.492 --> 00:21:45.532
So, but I think fundamentally what the other panelists have said,
00:21:45.692 --> 00:21:52.052
it's such an enormously complicated investment from a scale of funding and sophistication.
00:21:52.212 --> 00:21:55.972
I don't think that's an issue. So, from a logistics point of view,
00:21:56.132 --> 00:21:58.652
if I go back to the beginning, which I think is more interesting.
00:21:58.992 --> 00:22:06.472
The risk is the number of real developments and the number of new entrants and
00:22:06.472 --> 00:22:11.452
potential solutions to those real developments can create deep confusion in
00:22:11.452 --> 00:22:12.392
part of the supply chain.
00:22:12.512 --> 00:22:15.932
So, whether that's the manufacturers who are reasonably used to dealing with
00:22:15.932 --> 00:22:21.752
that, but certainly the planners, the power networks, I think the fiber providers
00:22:21.752 --> 00:22:25.512
are over it. What about permits? And then that could actually choke things if we're not careful.
00:22:25.712 --> 00:22:30.832
And so we've done a lot of work with a couple other self-builders and data center
00:22:30.832 --> 00:22:36.392
developers to try and lay out a fairly neutral view of what's happening and help,
00:22:36.572 --> 00:22:40.072
whether it's AEMO or the water authorities or so on and so forth,
00:22:40.232 --> 00:22:42.292
get a zoom in on what's genuinely real.
00:22:42.292 --> 00:22:45.312
And then it really comes down to how does
00:22:45.312 --> 00:22:48.252
the construction industry respond and that's
00:22:48.252 --> 00:22:51.592
going to be very different in every state in queensland it's reasonably
00:22:51.592 --> 00:22:55.032
soft at the moment which is great but at some point they are going to actually
00:22:55.032 --> 00:22:58.932
start responding to the fact they've got the olympics yeah and so that'll change
00:22:58.932 --> 00:23:03.632
and new south wales is different to victoria so it's it's about volume and securing
00:23:03.632 --> 00:23:08.152
the expertise that you need but not too soon because you can't build until you
00:23:08.152 --> 00:23:09.152
you've you got your certainty.
00:23:09.712 --> 00:23:12.532
Uday, what's the biggest logistical challenges in your view?
00:23:13.452 --> 00:23:18.552
I'd say two things. I mean, power is very obvious in terms of particularly with
00:23:18.552 --> 00:23:21.172
a data center hat on, having sufficient power.
00:23:21.412 --> 00:23:25.012
You asked the question about renewable energy earlier, and I think renewable
00:23:25.012 --> 00:23:27.132
is going to fill a lot of the gap.
00:23:29.452 --> 00:23:32.612
We've got a platform here in this country called Neo, and we're seeing,
00:23:32.732 --> 00:23:37.172
and in some of the markets too, a data center is seem to be the single biggest
00:23:37.172 --> 00:23:38.672
driver in terms of our demand.
00:23:40.112 --> 00:23:43.272
But the challenge is going to be in the grid. You know, there's going to be
00:23:43.272 --> 00:23:48.592
– renewable can't solve everything because the energy sources are not continuous
00:23:48.592 --> 00:23:52.172
and predictable. So, you're going to have to have batteries. You're going to have to.
00:23:56.197 --> 00:24:01.357
Support is required from both public and private investment in those spaces.
00:24:01.517 --> 00:24:04.897
I think power and strengthening the grid is going to be an important issue.
00:24:05.177 --> 00:24:09.057
Two, I think we've touched on earlier on construction, but just people,
00:24:09.217 --> 00:24:10.497
just finding the right talent.
00:24:10.977 --> 00:24:13.537
I mean, digital infrastructure is physical. At the end of the day,
00:24:13.637 --> 00:24:18.697
it's got to be designed, built, operated, require specialist people.
00:24:19.097 --> 00:24:22.617
A number of countries are struggling, I think, to get the right people.
00:24:22.617 --> 00:24:28.697
Also, any policy settings or changes that can support getting more skills at
00:24:28.697 --> 00:24:30.397
the right time is going to make a difference.
00:24:30.537 --> 00:24:34.077
We're suddenly reducing immigration, including putting a cap on foreign students,
00:24:34.157 --> 00:24:36.177
aren't we? Yeah, look, it's a...
00:24:38.897 --> 00:24:41.757
Guy, do you want to make a comment on the logistical challenges?
00:24:42.257 --> 00:24:44.077
I don't want to make a comment on foreign students.
00:24:45.377 --> 00:24:48.857
Yeah, look, picking up on what Udo said, that's something we think about a lot
00:24:48.857 --> 00:24:51.877
at Equinex is around people bringing the right folks in.
00:24:51.997 --> 00:24:55.757
As this data center industry proliferates, there's just so many being built
00:24:55.757 --> 00:24:57.157
at such scale so quickly.
00:24:57.497 --> 00:25:00.277
And these, again, are mission-critical facilities. This isn't something you
00:25:00.277 --> 00:25:02.597
can walk in tomorrow and run within two weeks.
00:25:02.737 --> 00:25:06.457
And so thinking about that pipeline of talent through vocational colleges,
00:25:06.717 --> 00:25:10.357
through universities, outreach with programs, there's diversity challenges in our industry.
00:25:10.557 --> 00:25:13.977
And so thinking through all of that, and it's not an immediate near and present
00:25:13.977 --> 00:25:16.477
concern, but it's certainly something we've got to think about long-term,
00:25:16.817 --> 00:25:19.097
not just as my organization, but as an industry.
00:25:19.517 --> 00:25:21.697
The other thing I'd add on to the sustainability piece, Steve,
00:25:21.757 --> 00:25:25.277
because I think you asked that question, renewable projects in Australia, there are lots.
00:25:25.757 --> 00:25:30.377
Quality renewable projects with immediate or near-term connection to the grid, there are not as many.
00:25:30.657 --> 00:25:34.397
And so as we scale up, we scale up for these customers that are very demanding
00:25:34.397 --> 00:25:39.437
on the sustainability side, as we are on ourselves, that's a challenge we're
00:25:39.437 --> 00:25:41.117
going to really address next kind of three to five years.
00:25:42.397 --> 00:25:45.177
All right. Mark, you want to make a comment on the logistical challenges?
00:25:46.017 --> 00:25:47.057
No. Okay. We'll move on then.
00:25:47.317 --> 00:25:51.717
So our last topic before we wrap up is what are the public policy challenges
00:25:51.717 --> 00:25:55.517
for Australia's digital infrastructure future? Who wants to take that one on first?
00:25:56.177 --> 00:26:00.617
Simon? Happy to start there. Again, as I said earlier, I think that the mood
00:26:00.617 --> 00:26:01.957
has significantly changed.
00:26:02.157 --> 00:26:05.697
So that was a challenge and I was quite concerned about it 12,
00:26:05.817 --> 00:26:07.857
18 months ago. So that's great.
00:26:08.237 --> 00:26:11.017
What changed and who's changing it?
00:26:11.757 --> 00:26:15.837
We went we seem to be going down a pathway where
00:26:15.837 --> 00:26:19.017
data centers were at significant risk of being blamed
00:26:19.017 --> 00:26:21.737
for not allowing some of the things that we want to do
00:26:21.737 --> 00:26:25.017
to happen and and the the grid transition was
00:26:25.017 --> 00:26:29.257
a primary one water was lurking in the background as we you know it's a it's
00:26:29.257 --> 00:26:32.917
a challenge in many parts of the country naturally and we want to expand housing
00:26:32.917 --> 00:26:38.377
and so on but i think through some significant efforts and and credit to the
00:26:38.377 --> 00:26:42.317
folks who've been involved listening and us as a group saying,
00:26:42.517 --> 00:26:47.697
we understand we have to contribute and we will just open the doors and let us talk.
00:26:47.917 --> 00:26:50.797
So that's great. I think there's...
00:26:51.336 --> 00:26:56.756
The continuous challenges of what regulation is required and things we've maybe
00:26:56.756 --> 00:27:02.496
not even really solved yet a bit further upstream, whether it be on the social media side or so on.
00:27:02.656 --> 00:27:06.296
There's a whole bunch of sort of regulation catch up that has to happen.
00:27:06.476 --> 00:27:09.976
And then there was the point that you made or someone made earlier about how
00:27:09.976 --> 00:27:14.936
do folks invest in the country and we've got different tax rules and import and so on.
00:27:15.096 --> 00:27:17.936
And will that stifle growth maybe at some point?
00:27:18.136 --> 00:27:23.436
But again, I've seen generally the large investors in the production platforms
00:27:23.436 --> 00:27:26.116
like the hyperscalers and so on generally move towards that.
00:27:26.596 --> 00:27:31.156
So I think for me, it really comes down to not over-regulating and putting regulations
00:27:31.156 --> 00:27:35.476
in place that allow speed versus historically, you've got to be reliable,
00:27:35.476 --> 00:27:38.836
but you're only going to grow at 1% Mr. Water Network or Mr. Power Network.
00:27:38.996 --> 00:27:41.696
And those changes are a challenge. Guy, you want to chip in there?
00:27:42.296 --> 00:27:45.876
Yeah, I think one of the things we'd certainly like to see the government leaning
00:27:45.876 --> 00:27:48.716
on is around an artificial intelligence strategy.
00:27:49.276 --> 00:27:53.476
We're one of the few developed countries in the world that doesn't have a government-led
00:27:53.476 --> 00:27:55.456
AI strategy. Korea's got one.
00:27:55.856 --> 00:27:57.896
Japan has one. Singapore, United States.
00:27:58.356 --> 00:28:03.256
And that's kind of, it's not giving us as much direction as we should have as a nation.
00:28:03.496 --> 00:28:06.776
When you consider the components that go into artificial intelligence,
00:28:06.976 --> 00:28:10.276
you think about the critical minerals that are sitting underneath our feet.
00:28:10.436 --> 00:28:16.136
We have an incredible role to play in the future of AI, but without government-led
00:28:16.136 --> 00:28:18.496
strategy, that's definitely a gap that we see.
00:28:19.176 --> 00:28:21.616
So do you see a real opportunity there? Because at the moment,
00:28:21.736 --> 00:28:26.356
I can't remember what the figure is, but China has virtual monopoly of not just
00:28:26.356 --> 00:28:28.076
digging them out of the ground, but processing them.
00:28:28.196 --> 00:28:30.396
Do you see an opportunity there as well?
00:28:30.896 --> 00:28:34.836
I mean, long term for Australia, yeah, absolutely. I mean, taking more value
00:28:34.836 --> 00:28:39.076
out of that supply chain is something we need to do, but it's a very difficult process.
00:28:39.416 --> 00:28:42.276
I think the government recognizes that, and you can see that in some of the
00:28:42.276 --> 00:28:45.596
conversations federally, but not only on critical minerals.
00:28:45.736 --> 00:28:49.076
We just talked about workforce, how we think about that leading into 2030,
00:28:49.376 --> 00:28:51.316
how we bring more people into the collective industry.
00:28:51.556 --> 00:28:54.316
The opportunity is huge, but it does need some guidance from government.
00:28:55.019 --> 00:28:58.679
Mark, what do you think are the public policy challenges in this area?
00:28:58.839 --> 00:29:04.579
Our challenges are predominantly, so we deal with local, state and federal.
00:29:05.159 --> 00:29:10.259
Our problems are construction delays with state, state-owned assets,
00:29:10.459 --> 00:29:13.899
rail, waterways, motorways, transport.
00:29:14.299 --> 00:29:19.239
I mean, we're still three years waiting for approvals to cross some train and
00:29:19.239 --> 00:29:24.699
motorways. You know, we have to escalate to a TIO process that takes a year
00:29:24.699 --> 00:29:26.599
to review whether it's a low impact or not.
00:29:27.219 --> 00:29:34.299
So I think a review of telecommunications infrastructure rights and compliance
00:29:34.299 --> 00:29:37.519
is definitely something I think we need to be held accountable for.
00:29:38.259 --> 00:29:41.559
So, you know, as it stands today, once you get a federal license carrier,
00:29:41.719 --> 00:29:43.659
you have rights to actually construct.
00:29:43.919 --> 00:29:47.119
And I think there needs to be a bit more governance on how people do do that.
00:29:47.259 --> 00:29:52.119
But in return, I think we need to understand the speed of this infrastructure
00:29:52.119 --> 00:29:53.019
and how it's getting built.
00:29:53.479 --> 00:29:57.399
Yeah. The customers that are going into these data centers have seen delays
00:29:57.399 --> 00:29:59.619
due to these permitting process issues.
00:30:00.019 --> 00:30:04.879
They've had to roll back deployments. They've had to roll back workloads because they can't get ready.
00:30:05.119 --> 00:30:08.419
We've got our beautiful union in Melbourne. They always play a role for us,
00:30:08.559 --> 00:30:14.759
but the state government and or the actual approval process to actually deal
00:30:14.759 --> 00:30:19.639
with these state departments, it's been the biggest delay in more deployments
00:30:19.639 --> 00:30:22.459
in this country. Is it better in one state or the other?
00:30:23.119 --> 00:30:28.659
Predominantly, no. No. I mean, you're looking at a 90-week approval process to cross a railway.
00:30:29.339 --> 00:30:35.959
Yeah. And so with those processes, I think that review on us being held to account,
00:30:36.099 --> 00:30:39.379
but definitively the other departments helping us actually complete the infrastructure
00:30:39.379 --> 00:30:41.099
is critical. We need an escalation.
00:30:41.719 --> 00:30:44.579
Don't you get the final word on public policy challenges?
00:30:45.476 --> 00:30:48.256
Look, I think a lot of it has already been covered, but with,
00:30:48.476 --> 00:30:51.096
again, with an investor hat on, I'd say two things.
00:30:51.276 --> 00:30:57.156
One is, you know, from an investment perspective, we like to increase predictability.
00:30:57.396 --> 00:31:02.136
That matters a lot. Investors don't like a lot of change. So that covers a whole
00:31:02.136 --> 00:31:07.636
gamut of things from, you know, financing structures through to physical sort
00:31:07.636 --> 00:31:09.676
of support around infrastructure creation.
00:31:09.996 --> 00:31:14.476
And two, we've actually touched on a lot of bottlenecks. So anything that can
00:31:14.476 --> 00:31:19.836
be done to make sure that the digital infra lens is also applied,
00:31:19.836 --> 00:31:24.296
because often we're talking about a zero-sum game, it's the same resources.
00:31:24.496 --> 00:31:28.596
There's other public infrastructure being created, whether it's labor capital or whatever.
00:31:28.816 --> 00:31:33.916
So just bringing another pretty specific digital infra lens to things, because it matters.
00:31:34.216 --> 00:31:37.736
It is the lifeline of the future economy. So if you don't do that,
00:31:37.976 --> 00:31:38.836
it's going to get missed.
00:31:39.556 --> 00:31:44.856
Australia is being seen by these companies in the world as a really good place
00:31:44.856 --> 00:31:46.156
to put their infrastructure.
00:31:46.456 --> 00:31:49.536
And Sydney and Melbourne especially are only going to continue to grow.
00:31:49.776 --> 00:31:52.636
We have to make sure we stay ahead to keep them coming.
00:31:52.936 --> 00:31:57.396
We can't become a country that had an opportunity to leverage the digital infrastructure.
00:31:57.616 --> 00:32:01.436
And so the policy around AI and our plan as a country and how we're going to
00:32:01.436 --> 00:32:05.136
support it, we need to support the digital infrastructure expansion and make sure they stay.
00:32:05.896 --> 00:32:09.336
Guys, great to get all your thoughts. Thanks so much. Please thank our panel.