As cities run out of space and power, data centres are shifting to regional locations. This episode asks whether data centres can thrive beyond the grid and whether opening access to digital infrastructure can attract long-term investment. We also look at how ultra-low latency and speed to market are increasingly shaping the next wave of development, not just access to energy.
Can data centres thrive outside the grid?
Meeting demand for AI and digital services through decentralisation
16 March 2026
0:00
/
0:00
00;00;01;11 - 00;00;04;08
Nexus, Publish By GHD.
00;00;04;08 - 00;00;05;27
Where ideas connect.
00;00;08;20 - 00;00;09;20
Hello,
everyone.
00;00;09;20 - 00;00;12;19
This is part two of our podcast.
00;00;13;01 - 00;00;17;29
In part one, we covered how infrastructure
constraints are driving innovation.
00;00;18;09 - 00;00;20;25
Now we shift focus to what's next.
00;00;20;28 - 00;00;23;12
What's the future look like?
What's the next frontier?
00;00;23;15 - 00;00;28;23
And pleasingly today we're again
joined by hacks Sam and Ty.
00;00;29;02 - 00;00;32;02
So thanks for staying with us
to continue the conversation.
00;00;32;05 - 00;00;33;09
Welcome, guys.
00;00;33;12 - 00;00;35;27
Hello. Good to be back here.
00;00;36;00 - 00;00;38;08
So in this episode,
we're going to chat through
00;00;38;08 - 00;00;41;12
the untapped
potential of regional data centers.
00;00;41;19 - 00;00;44;11
And AI inferencing hubs.
00;00;44;14 - 00;00;47;24
We'll look at how these emerging models
are enabling latency
00;00;47;27 - 00;00;52;04
sensitive applications and reshaping
infrastructure investment strategies.
00;00;52;19 - 00;00;53;20
And it's something that I think
00;00;53;20 - 00;00;56;28
is, really a nice bridge
from our previous episode
00;00;57;01 - 00;01;00;23
where we looked at how technology
is really driving innovation and the rise.
00;01;00;26 - 00;01;03;06
And, you know,
I was kind of quite exciting
00;01;03;06 - 00;01;06;06
at the end of that to hear about,
what the future might hold.
00;01;06;18 - 00;01;10;19
But I also think that there's a big future
that we didn't discuss, which is looking
00;01;10;24 - 00;01;14;17
at some of our regional locations,
which perhaps traditionally
00;01;14;20 - 00;01;16;17
haven't been the focus in the market.
00;01;16;18 - 00;01;18;06
The market has primarily been
00;01;18;06 - 00;01;22;04
focused in big city
locations, often for latency reasons.
00;01;22;07 - 00;01;25;07
But, you know,
I think that we are seeing a big shift
00;01;25;12 - 00;01;28;29
to the opportunities
that are available in regional locations
00;01;29;07 - 00;01;32;02
for things
like AI training, as one example.
00;01;32;05 - 00;01;34;24
But maybe to start with you, Sam,
are you able to
00;01;34;25 - 00;01;38;03
perhaps give us some insights
as to why regional locations are becoming
00;01;38;09 - 00;01;41;09
more and more viable for our data centers?
00;01;41;15 - 00;01;46;17
I guess this does link quite nicely to,
the previous discussion we had, but,
00;01;47;00 - 00;01;50;07
the regional appeal is coming
because, the resource
00;01;50;10 - 00;01;55;00
availability in the central
populated cities is diminishing.
00;01;55;14 - 00;02;00;02
So, with
I guess, lack of access to the approvals
00;02;00;05 - 00;02;03;12
for development, land,
water, energy in the cities.
00;02;03;25 - 00;02;07;25
Then, you see these,
developers are looking elsewhere
00;02;07;28 - 00;02;10;27
for solutions to satisfy that demand.
00;02;11;00 - 00;02;14;06
So that's where they cost
the net larger and broader,
00;02;14;09 - 00;02;17;25
and they find out where, the markets
that don't require
00;02;18;04 - 00;02;21;21
the immediate service of inner city,
connectivity,
00;02;22;00 - 00;02;25;25
they become a viable customer
in these regional locations.
00;02;26;15 - 00;02;27;10
Yeah. It's interesting.
00;02;27;10 - 00;02;31;03
I also think one of the other things
that I am personally keen to say
00;02;31;06 - 00;02;34;06
is that making sure we don't get like a,
a great digital divide
00;02;34;11 - 00;02;37;19
where it becomes sort of, you know,
those that live in cities have access
00;02;37;22 - 00;02;40;29
to the greatest, technology
and AI infrastructure,
00;02;41;02 - 00;02;43;19
digital infrastructure
compared to those in the region.
00;02;43;20 - 00;02;45;21
So that's the other pleasing thing,
00;02;45;21 - 00;02;49;09
is to hopefully enable
that for the regional communities.
00;02;49;21 - 00;02;52;09
And that's a great point
because, connectivity
00;02;52;09 - 00;02;56;01
in regional zones
is increasing incredibly quickly.
00;02;56;04 - 00;03;01;05
The interconnections between countries
are, increasing at an incredible rate.
00;03;01;17 - 00;03;05;01
And internally,
infrastructure developments are now
00;03;05;04 - 00;03;10;14
including high, connectivity for,
these restricted access areas.
00;03;11;05 - 00;03;15;16
But, what we are trying
to, use as our target market
00;03;15;19 - 00;03;20;25
is that regional interest
and finding a demand where we can connect
00;03;20;28 - 00;03;27;17
to landfall locations with quite a small,
latency and still provide
00;03;27;20 - 00;03;32;13
a reliable, connected service
that can satisfy that, regional demand.
00;03;32;26 - 00;03;34;09
So, okay, that's quite interesting.
00;03;34;12 - 00;03;38;12
So you're finding that
the regional component offers
00;03;38;19 - 00;03;42;16
the opportunity to connect closer to it,
where there's nowhere near regional area
00;03;43;01 - 00;03;46;00
to achieve the latency parameters,
because your
00;03;46;03 - 00;03;49;02
your nearby regional town
or regional area.
00;03;49;05 - 00;03;50;29
Yeah, it goes down to, physics.
00;03;50;29 - 00;03;53;16
And forgive me for,
not being too technical
00;03;53;16 - 00;03;56;20
on this topic
and asking that jumped in, but,
00;03;57;10 - 00;04;00;17
I've got the great reference
that, we're looking at Fiji and would love
00;04;00;20 - 00;04;03;23
to, have our breakout,
development there.
00;04;04;07 - 00;04;08;08
And, we've got a round trip
of 30 milliseconds from Sydney.
00;04;08;18 - 00;04;12;23
And so that's a service
that, really is quite functional
00;04;12;26 - 00;04;16;11
for quite a lot of different
consumers of data,
00;04;16;25 - 00;04;20;18
which is where we targeting
to, build up our customer base.
00;04;21;02 - 00;04;22;08
Super interesting.
00;04;22;11 - 00;04;25;15
I think it's fantastic
because also with the shift away
00;04;25;18 - 00;04;30;03
from the big city centers and also,
you know, looking to advance manufacturing
00;04;30;06 - 00;04;34;08
and build communities
and sort of the old haven't spoke model,
00;04;34;14 - 00;04;36;17
we need to be able to provide
the digital infrastructure
00;04;36;17 - 00;04;39;03
to enable those cities
and communities to thrive.
00;04;39;03 - 00;04;42;02
And so I think that's the opportunity
there to,
00;04;42;05 - 00;04;44;10
you know, not just for private investment,
but for government
00;04;44;10 - 00;04;48;06
to start to incentivize investment
in regional locations, to really make sure
00;04;48;09 - 00;04;51;12
that the industry is keeping pace
with the rest of the globe.
00;04;52;10 - 00;04;55;28
And I guess on investors tie, I'm
interested to hear from your point of view
00;04;56;01 - 00;04;59;20
and what you've been seeing in the market
around some of the opportunities
00;04;59;23 - 00;05;03;08
that investors can unlock the value
in regional locations.
00;05;04;02 - 00;05;06;09
There's so many things
you could potentially unlock,
00;05;06;09 - 00;05;06;26
but I mean, how
00;05;06;29 - 00;05;07;05
when you
00;05;07;07 - 00;05;08;20
when you think about like communities
00;05;08;20 - 00;05;11;23
and landlocked areas
or infrastructure investors,
00;05;11;26 - 00;05;14;26
like when they use the word value,
what do you think they mean
00;05;15;17 - 00;05;17;01
in terms of from an investor perspective?
00;05;17;01 - 00;05;21;21
I think it means long term
sustainable returns and having their brand
00;05;21;24 - 00;05;26;05
associated with some positive change
for society in the world.
00;05;26;25 - 00;05;28;00
So I think it's a mix of those.
00;05;28;00 - 00;05;32;26
And I think if you look at this
AI space over the next five years or so,
00;05;33;10 - 00;05;36;10
it's going to be the biggest deployment
of capital in human history.
00;05;36;19 - 00;05;40;29
And so the early investors who I think
deploy capital wisely
00;05;41;08 - 00;05;45;28
will get very good returns and be able
to sort of grow exponentially with this.
00;05;46;14 - 00;05;49;03
You know, we're just starting
with really early in the race.
00;05;49;06 - 00;05;52;18
And I think, as Andrew mentioned,
we need to make sure the digital divide
00;05;52;21 - 00;05;53;22
isn't there.
00;05;53;25 - 00;05;56;14
And if you look at the growth
of usage of ChatGPT
00;05;56;17 - 00;05;59;29
or other models we're at,
that's about a billion people using them.
00;06;00;04 - 00;06;01;29
There's 8 billion people on the planet.
00;06;02;02 - 00;06;05;09
And I think all these efficiencies
are happening are fantastic.
00;06;05;12 - 00;06;06;20
And I think we need to go further.
00;06;06;20 - 00;06;08;13
You know,
we need to reduce the cost of these models
00;06;08;13 - 00;06;12;17
from $200 a month for the high end ones,
or $20 a month for the regular
00;06;12;20 - 00;06;17;08
use ones, down to $2 or $0.20
for 8 billion people around the world.
00;06;17;11 - 00;06;20;11
Regardless of you know which situation
they're in to be able to use it.
00;06;20;23 - 00;06;22;16
And so we need to enable that
through efficiency.
00;06;22;16 - 00;06;25;15
And technology is
the biggest efficiency driver you'll find.
00;06;25;15 - 00;06;27;01
And we've seen this in the past.
00;06;27;01 - 00;06;29;07
You know, the cost comes down
orders of magnitude.
00;06;29;10 - 00;06;31;29
But the growth in terms
of users will go up.
00;06;32;02 - 00;06;36;15
And I think providing that level of
intelligence to every human in the world.
00;06;36;18 - 00;06;40;06
I'm really optimistic that there's
some smart young people around the world
00;06;40;09 - 00;06;42;14
that are going
to create some magic with it.
00;06;42;17 - 00;06;47;01
And I think that's our responsibility is,
you know, who's using behind the meat's,
00;06;47;06 - 00;06;50;03
sustainable power,
putting the data centers in place,
00;06;50;03 - 00;06;51;06
providing the GPUs,
00;06;51;06 - 00;06;54;06
bringing down the cost of these models,
especially on the inference side.
00;06;54;11 - 00;06;56;29
Is our responsibility
to better enable that future.
00;06;57;02 - 00;07;00;24
And so I'm really curious,
once we get to 2030, 2035
00;07;00;27 - 00;07;05;15
and hopefully we've got the cost down
to something like $0.20 a month for folks
00;07;05;18 - 00;07;06;11
to use.
00;07;06;14 - 00;07;09;09
It's going to be really interesting
to see what comes out of that.
00;07;09;12 - 00;07;11;19
I want jump in on that one as well,
because I'm quite heavy
00;07;11;19 - 00;07;13;12
in this topic currently.
00;07;13;12 - 00;07;18;14
And one of the great synergies
that we have with ocean is that,
00;07;18;24 - 00;07;23;15
we are looking at power generation
and data center provision as a bundle.
00;07;24;03 - 00;07;27;16
If we can do data center
as a single service, fantastic.
00;07;27;19 - 00;07;32;17
But for, the discussions around
infrastructure investment and development,
00;07;33;04 - 00;07;36;10
they are still locked into thinking
infrastructure
00;07;36;13 - 00;07;41;02
is energy, water
and other basic services to the community.
00;07;41;05 - 00;07;44;24
Data is not an infrastructure
kind of thing.
00;07;44;27 - 00;07;47;14
It's still just a telco investment.
00;07;47;17 - 00;07;53;09
But linking that data into development,
into something that can underwrite
00;07;53;21 - 00;07;57;00
the development of energy
as a massive development.
00;07;57;03 - 00;08;01;19
Knowing that you've got a data center
providing a guaranteed consumer
00;08;01;22 - 00;08;04;22
for X amount of years, that stability
00;08;05;02 - 00;08;07;18
breeds more investment.
00;08;07;21 - 00;08;11;11
And so, yeah, it's
kind of using the data center demand
00;08;11;14 - 00;08;16;17
as a leverage to actually do
greater energy development
00;08;16;20 - 00;08;20;17
that is going to service the community
where that development
00;08;20;20 - 00;08;23;20
may not have been financially feasible
prior to that.
00;08;24;04 - 00;08;24;25
Yeah.
00;08;24;28 - 00;08;28;02
I think my Andrew, my colleague Salmon
Hawks have helped me answer that question.
00;08;28;05 - 00;08;29;08
Yeah.
00;08;29;11 - 00;08;31;23
One of the things we haven't touched on
is that, you know, in
00;08;31;23 - 00;08;35;18
regional locations is the ability
to build closer to the generation.
00;08;35;21 - 00;08;38;19
So looking at, you know, being able
to build closer to renewable energy
00;08;38;22 - 00;08;43;14
sources, being able to build closer
to different water, supplies.
00;08;43;23 - 00;08;47;18
And, you know, that enables things
like AI training to be done
00;08;47;27 - 00;08;49;15
in regional locations.
00;08;49;18 - 00;08;51;16
So those sorts of opportunities
00;08;51;16 - 00;08;55;03
out in the regions to get investment into
AI training is huge.
00;08;55;26 - 00;08;59;15
Also, just due to geopolitical challenges
at the moment and,
00;08;59;18 - 00;09;02;09
where things are being built
and being invested,
00;09;02;12 - 00;09;05;09
there's sovereign capability
with a lot of uncertainty in the globe
00;09;05;09 - 00;09;09;13
at the moment,
being able to train your own capability
00;09;09;16 - 00;09;12;19
and knowledge of data
and the ability to control data and to get
00;09;12;21 - 00;09;16;01
smarter is is kind of become
a sovereign capability, if not already.
00;09;16;04 - 00;09;20;05
So I think that, you know, being able
to do AI training on, on home
00;09;20;08 - 00;09;21;16
soil in regional locations
00;09;21;16 - 00;09;24;14
is going to be become more
and more prevalent across the globe
00;09;24;17 - 00;09;27;19
and will not just be limited
to a select few nations.
00;09;28;19 - 00;09;31;11
And look, we did just mentioned
latency there.
00;09;31;11 - 00;09;35;21
And some of the, I guess, challenges,
but also opportunities in the regions.
00;09;35;24 - 00;09;37;28
But one of the challenges
with regions is latency.
00;09;37;28 - 00;09;42;15
But we might actually zoom back in
to the opposite end of the spectrum.
00;09;42;23 - 00;09;46;05
And perhaps, Hawkes,
I might start with you just to talk around
00;09;46;08 - 00;09;49;18
what ultra low latency is
and why it's important
00;09;49;21 - 00;09;53;00
for data centers and what what's
the sort of some of the types of uses
00;09;53;03 - 00;09;56;16
where ultra low latency
is going to be important in data centers?
00;09;57;05 - 00;09;58;00
Yeah, absolutely. Andrew.
00;09;58;00 - 00;10;01;04
So, I think as you mentioned,
there will be some regional
00;10;01;07 - 00;10;03;06
and there'll be
some national capabilities.
00;10;03;06 - 00;10;03;15
You know,
00;10;03;18 - 00;10;06;20
I think there's general intelligence
and then there's sovereign models,
00;10;06;23 - 00;10;07;12
as you mentioned.
00;10;07;12 - 00;10;09;15
And geopolitics is providing
00;10;09;18 - 00;10;12;25
another nuance on top of that, especially
with the rise of sort of nationalism
00;10;13;12 - 00;10;16;01
in terms of ultra low latency,
where it becomes
00;10;16;01 - 00;10;18;13
really important is on the
AI inferencing side.
00;10;18;16 - 00;10;22;03
So inferences where AI is monetized,
that's where you deploy the models
00;10;22;06 - 00;10;24;12
and people prompt them,
and that's where you pay the subscription.
00;10;24;12 - 00;10;26;08
Whichever model you choose to use
00;10;26;11 - 00;10;28;14
to be able
to get the replies to your prompt
00;10;28;14 - 00;10;30;15
and the intelligence and tokens
that you need.
00;10;30;18 - 00;10;33;16
So inference to scale
is going to be distributed.
00;10;33;19 - 00;10;37;04
Being close to
I say users is really important
00;10;37;07 - 00;10;41;02
in that sort of ultra low latency
and uses today I think are humans,
00;10;41;17 - 00;10;44;15
but users in the future might be machines
and so really depends.
00;10;44;18 - 00;10;45;25
Although we think about it as cities,
00;10;45;26 - 00;10;48;03
it also depends on
what entities users are.
00;10;48;06 - 00;10;51;05
So I think as we unpack that,
as we go forward as an industry,
00;10;51;08 - 00;10;53;29
it's about being close
to whoever that user is.
00;10;54;02 - 00;10;56;08
And we've found to be able
to deliver distributed
00;10;56;08 - 00;11;00;06
AI and inference, you need a really
a standard rapid deployment model.
00;11;00;09 - 00;11;02;22
You need something that's standardized
that you can play in multiple ways.
00;11;02;22 - 00;11;05;21
And this is where the prefab model
does really well.
00;11;05;24 - 00;11;09;20
And so the inference side to your point
is a data security piece.
00;11;09;23 - 00;11;10;03
So you know,
00;11;10;03 - 00;11;11;22
if you think about the sovereign models,
00;11;11;25 - 00;11;14;14
the sovereign models
are going to be for national security.
00;11;14;14 - 00;11;17;02
But there's also going to be civilian
capabilities as well.
00;11;17;02 - 00;11;20;13
So a country will make available
government services
00;11;20;24 - 00;11;23;03
to all of its citizens, essentially.
00;11;23;06 - 00;11;26;13
And so that will need to be inference
for citizens.
00;11;26;16 - 00;11;27;14
And,
00;11;27;15 - 00;11;32;01
you know, returns will be significant
here in the sense of lowest risk returns.
00;11;32;08 - 00;11;34;13
The sovereign models
and the sovereign deployments,
00;11;34;13 - 00;11;37;17
I think will be most attractive
to investors, and then it will go down
00;11;37;20 - 00;11;41;04
to investment grade, and then it's a more
of the more speculative stuff.
00;11;41;07 - 00;11;41;22
Essentially.
00;11;41;22 - 00;11;46;10
So ultra low latency data security
and investors is really all comes together
00;11;46;23 - 00;11;49;28
with our inferencing around
general intelligence and sovereign models.
00;11;50;25 - 00;11;51;02
Yeah.
00;11;51;02 - 00;11;55;28
Just, as one example of the machines,
it's things like self-driving cars.
00;11;56;01 - 00;11;58;14
You know, I certainly hope
that it doesn't take too long
00;11;58;14 - 00;12;01;09
to make a decision
when something runs in front of the car.
00;12;01;09 - 00;12;01;27
You certainly want
00;12;01;27 - 00;12;06;04
the ultra low latency there
so that, a decision can be made by AI.
00;12;06;07 - 00;12;07;12
Yeah.
00;12;07;15 - 00;12;10;12
So I guess the other thing to,
00;12;10;12 - 00;12;13;14
with you on the line heart is, you know,
what do you think?
00;12;13;17 - 00;12;17;08
You know, sort of next wave of data
center projects with, like
00;12;17;11 - 00;12;21;11
you said, speed to market, ability
to get projects online quickly.
00;12;21;21 - 00;12;22;05
What are those
00;12;22;05 - 00;12;25;10
sort of investors looking for
in the next wave of data center projects?
00;12;26;09 - 00;12;28;02
Yeah, I think as we look at the next wave
00;12;28;03 - 00;12;30;00
and as I mentioned,
we're early in this space,
00;12;30;00 - 00;12;32;29
but it is going to be the biggest
deployment of capital in human history,
00;12;33;02 - 00;12;33;20
as I mentioned.
00;12;33;20 - 00;12;38;08
And so the early investors who get in
and pick the right bets, I would say,
00;12;38;20 - 00;12;39;23
are going to do really well
00;12;39;25 - 00;12;42;26
and probably have really
attractive returns in the longer term.
00;12;43;09 - 00;12;46;00
It's almost like
we're at the start of the internet era,
00;12;46;03 - 00;12;47;25
but it's going to accelerate
a lot quicker.
00;12;47;25 - 00;12;48;27
What happened in 20 years
00;12;48;27 - 00;12;51;29
and the internet is going to happen
in five years with AI essentially.
00;12;52;02 - 00;12;56;06
So there's huge opportunities here
if investors pick the right places
00;12;56;09 - 00;12;57;16
to back, essentially.
00;12;57;19 - 00;13;01;14
And I think, as you mentioned
earlier, power is the biggest constraint.
00;13;01;20 - 00;13;03;29
And so it's about
how do you get the most out of that power.
00;13;04;02 - 00;13;07;01
And the way to get the most out of
the power is to be able to deploy the most
00;13;07;04 - 00;13;10;03
tokens, deploy the latest Nvidia chips,
00;13;10;06 - 00;13;12;28
and have data centers
that evolve with the speed of technology.
00;13;12;28 - 00;13;15;06
And that's what we're focused on
is to be able to do that.
00;13;15;06 - 00;13;18;24
And so whether it's behind the meter,
the power that Sam generating
00;13;19;05 - 00;13;21;14
or this grid power,
you know, it's about saying, okay,
00;13;21;14 - 00;13;24;07
we have a limited amount of power,
how do we get the most amount of tokens?
00;13;24;07 - 00;13;24;28
And that's how you grow.
00;13;24;28 - 00;13;25;10
Revenue
00;13;25;10 - 00;13;27;07
essentially, is the number of tokens
that you're able to deliver
00;13;27;07 - 00;13;28;29
and the intelligence
you're able to deliver.
00;13;29;02 - 00;13;31;07
So I think you need to make sure
you look across the whole chain.
00;13;31;07 - 00;13;34;10
As we've mentioned before,
the whole stack, the whole ecosystem.
00;13;34;20 - 00;13;37;21
And then also make sure
you find the right sort of models
00;13;37;24 - 00;13;41;02
and the right risk
adjusted, projects to get involved in.
00;13;41;23 - 00;13;42;03
Yeah.
00;13;42;03 - 00;13;46;11
And so I'm just interested in your views
as well, in terms of, where you see
00;13;46;18 - 00;13;50;00
the next wave of projects coming,
particularly perhaps
00;13;50;03 - 00;13;53;04
in some of those regional locations
and what it might mean for ocean Bay.
00;13;53;29 - 00;13;56;14
Certainly a portion of the market
that I'm watching
00;13;56;14 - 00;13;59;19
closely is the sustainable development
of these data centers.
00;13;59;23 - 00;14;03;18
So we're seeing constriction
in all major cities.
00;14;03;21 - 00;14;07;10
The bottlenecks are stretching,
and that constriction
00;14;07;13 - 00;14;09;21
isn't going to go away
until the demand goes away.
00;14;09;21 - 00;14;14;06
And, I mean, Hawks is pretty confident
this demand is here to stay, and I agree.
00;14;14;20 - 00;14;20;08
And so the next wave is moving
to the tiered structure of deployments
00;14;20;11 - 00;14;25;11
of data centers that, not reducing value
for the communities around them,
00;14;25;13 - 00;14;29;28
that they're not impacting the resources
required for everyday life.
00;14;30;11 - 00;14;33;00
And, adding that extra value.
00;14;33;03 - 00;14;36;02
And really,
that's that's what we are targeting.
00;14;36;02 - 00;14;38;05
We are trying to target,
00;14;38;05 - 00;14;41;02
the demand that is growing
at these regional data centers.
00;14;41;05 - 00;14;42;27
They're being pushed
out of the major cities.
00;14;42;27 - 00;14;46;00
They're they're being sent
to locations that
00;14;46;03 - 00;14;49;07
people kind of go, well,
we don't have the power available yet.
00;14;49;10 - 00;14;51;24
So we got, you know,
what are we going to do?
00;14;51;24 - 00;14;53;22
We want the data,
but we don't have the space.
00;14;53;22 - 00;14;56;13
We don't have the energy,
we don't have the water.
00;14;56;16 - 00;15;00;25
And, Ocean Beat wants to be there saying,
well, we've got a solution for that.
00;15;00;28 - 00;15;05;00
And it's baseload renewable energy
and we're not going to take any land.
00;15;06;01 - 00;15;08;14
I think
I'll just add to what Sam mentioned there.
00;15;08;14 - 00;15;12;10
From an investor perspective,
there's about $70 trillion,
00;15;12;13 - 00;15;17;08
which is going to pass from baby boomers
to Gen Z in the next 20 years.
00;15;17;24 - 00;15;21;24
And the idea is that that $70 trillion
as it pass to the next generation, that
00;15;21;27 - 00;15;24;26
that would be invested
in more of the ESG, sustainable,
00;15;24;29 - 00;15;28;10
renewable
kind of opportunities, projects and funds.
00;15;28;13 - 00;15;31;13
And so there's a huge amount of capital
which is going to transfer
00;15;31;21 - 00;15;32;29
and be invested in these areas.
00;15;32;29 - 00;15;35;29
So I think you will see
that capital will flow over time as well
00;15;36;02 - 00;15;39;04
into this spaces,
and it'll become more of the norm now.
00;15;39;06 - 00;15;40;20
I think one of the other things that we
00;15;40;23 - 00;15;42;25
perhaps haven't touched on,
and just before we move to
00;15;42;25 - 00;15;46;04
finalize, is the reason why
there's a demand for data centers.
00;15;46;07 - 00;15;47;26
And you know why the data centers exist.
00;15;47;26 - 00;15;51;19
It's because it's going to enable progress
of the human race.
00;15;51;22 - 00;15;54;06
And it's a really important
it's the next evolution.
00;15;54;06 - 00;15;56;10
And it's really exciting to be part of.
00;15;56;10 - 00;15;59;15
And I think the challenge
and the opportunity at the same time
00;15;59;18 - 00;16;03;18
for all of us is what is the role
that we can play to enable
00;16;03;21 - 00;16;06;24
that change and the legacy that we leave
for the next generation.
00;16;07;03 - 00;16;07;21
As you mentioned,
00;16;07;21 - 00;16;11;12
hock $70 trillion worth of funds
being passed from the baby boomers
00;16;11;15 - 00;16;14;27
to the next generation, and certainly
a significant portion of that will be
00;16;15;07 - 00;16;18;24
on digital infrastructure to enable
that productivity and efficiency.
00;16;18;27 - 00;16;22;02
So it's a really exciting time,
for all of us.
00;16;22;05 - 00;16;26;28
And I guess with that, I'm keen to hear
if you had $10 billion,
00;16;27;01 - 00;16;31;04
so $10 billion, pick a currency,
let's call it US dollars in the industry.
00;16;31;07 - 00;16;34;01
Where would you be investing?
So you've got $10 billion.
00;16;34;01 - 00;16;35;04
Where would you invest?
00;16;35;07 - 00;16;37;17
I'll start with you, Sam.
00;16;37;20 - 00;16;41;10
So I did introduce myself
as a co-founder of a, startup.
00;16;41;15 - 00;16;44;14
So, no, I want to take that out.
00;16;44;17 - 00;16;48;12
To be fair,
I would be looking at the strategy
00;16;48;15 - 00;16;51;27
of the standardization
and prefabrication and,
00;16;52;26 - 00;16;53;13
maybe I'll.
00;16;53;15 - 00;16;57;05
If I had that 10 billion,
maybe I'd try and, come across the pond.
00;16;57;19 - 00;17;03;08
But, I think that is increasingly
the required strategy to be
00;17;03;11 - 00;17;06;23
the best in field is ensuring
that you've got a standardized,
00;17;06;26 - 00;17;11;13
repeatable process that ensures
that you can get the best efficiency
00;17;11;16 - 00;17;15;24
out of that construction
phase, reduce lost time,
00;17;16;11 - 00;17;19;18
and really just ensuring
that you've got best in class, fast
00;17;19;21 - 00;17;24;27
development, capable,
tied in with a solution that brings
00;17;25;00 - 00;17;27;29
in that social license
to continue to operate because,
00;17;28;03 - 00;17;29;24
you know, fast development is great.
00;17;29;24 - 00;17;33;16
But without that continued social losses,
that $10 billion is not coming back.
00;17;33;28 - 00;17;35;09
Well said.
00;17;35;12 - 00;17;38;11
And tie,
where would you invest $10 billion.
00;17;38;14 - 00;17;40;29
Yeah, $10 billion is a lot of money.
00;17;40;29 - 00;17;44;07
But it sounds not like a lot
compared to the 70 trillion
00;17;44;10 - 00;17;46;05
that's about to be transferred
from one generation
00;17;46;05 - 00;17;47;25
to another, or indeed the multi-trillion
00;17;47;25 - 00;17;50;00
that's going to be spent in the next
ten years.
00;17;50;03 - 00;17;53;03
Can I cheat
and give five to Sam and five to Hawks?
00;17;53;22 - 00;17;55;24
But that's cheating for me.
00;17;55;27 - 00;17;58;03
I do, you shorten. We're sorry.
00;17;58;06 - 00;17;58;17
You're right.
00;17;58;20 - 00;18;01;11
No, no. But, that's cheat answer. But,
you know, maybe.
00;18;01;11 - 00;18;05;07
Is there something in,
like setting up a, a global competition
00;18;05;18 - 00;18;10;16
where the person who gets to a P of X
Hawks in number?
00;18;10;21 - 00;18;14;11
What's a magic number
that you'd love to get to in terms of P or
00;18;14;13 - 00;18;16;04
in terms of investment? P.
00;18;17;08 - 00;18;19;10
It needs to be less sub 1.1.
00;18;19;13 - 00;18;20;04
Yeah. Right.
00;18;20;04 - 00;18;25;12
So the first person that can get to 1.05
gets $10 billion.
00;18;25;27 - 00;18;28;04
Sam, what's the P of I should be.
00;18;28;07 - 00;18;28;15
Yeah.
00;18;28;16 - 00;18;32;06
Well, I was just going to say
when the power that we're generating is,
00;18;32;09 - 00;18;38;06
for moving a lot of seawater that is,
coming out at the system at 17°C,
00;18;38;09 - 00;18;42;05
we have excess cooling that requires
no energy to move it
00;18;42;08 - 00;18;45;08
because we've consumed it
in the power generation cycle.
00;18;45;17 - 00;18;47;28
I believe we are very close to one.
00;18;48;01 - 00;18;49;21
Therefore, you might win the 10 million.
00;18;49;21 - 00;18;51;20
Congratulations
00;18;51;23 - 00;18;53;23
with great competition to.
00;18;53;24 - 00;18;55;29
I love it.
00;18;56;02 - 00;18;59;24
And Huck's interested from your point of
view where you would invest, $10 billion.
00;19;00;15 - 00;19;02;26
Yes. Andrew,
we did some maths on investment
00;19;02;26 - 00;19;05;23
required in the data center space
and in the AI space
00;19;05;26 - 00;19;10;04
generally, and to build a one gigawatt
data center, to just build a data
00;19;10;07 - 00;19;14;09
center is about $12.5 billion to fit it
full of chips
00;19;14;16 - 00;19;18;11
and GPUs
is probably another 30, $35 billion.
00;19;18;29 - 00;19;21;21
And so to have a gigawatt of GPU cloud
00;19;21;21 - 00;19;24;20
is in the order
of about a $50 billion check size.
00;19;24;23 - 00;19;28;27
So I'd come back to you and ask for five x
what you're giving me and say I would,
00;19;29;06 - 00;19;33;26
I would build a gigawatt
GPU cloud powered by renewable,
00;19;33;29 - 00;19;38;08
sustainable energy
and ideally supporting some sovereign
00;19;38;14 - 00;19;42;10
AI workloads, which is the lowest
risk from an investment perspective.
00;19;42;21 - 00;19;45;21
And hopefully,
like you said, giving everybody
00;19;45;26 - 00;19;48;18
Einstein
level of intelligence in their pocket
00;19;48;21 - 00;19;52;19
and then seeing what does that enable
and what does that give to humans
00;19;52;22 - 00;19;54;24
around the world. Nice.
00;19;54;27 - 00;19;59;14
For what it's worth, I'd be picking
some communities in need and investing
00;19;59;17 - 00;20;03;21
co-locating data centers
with some community benefits alongside it,
00;20;03;23 - 00;20;06;09
so that we really change
some of the narrative around
00;20;06;12 - 00;20;09;04
data centers at the moment,
because I think the industry needs it.
00;20;09;07 - 00;20;12;26
And I think that being able to demonstrate
that data centers can give back
00;20;12;28 - 00;20;15;10
to the community
and be a really fundamental part
00;20;15;10 - 00;20;18;13
of the future for the next generation
and the legacy that it leaves
00;20;18;16 - 00;20;19;23
is is really important
00;20;19;23 - 00;20;20;26
because we're not going to change
00;20;20;26 - 00;20;22;19
the world of data centers with 10 billion,
00;20;22;19 - 00;20;25;20
but we will change the lives
of a few communities that really need it.
00;20;25;22 - 00;20;27;29
So that's where I've been visiting
AI 10 billion.
00;20;27;29 - 00;20;30;18
But,
look, we're unfortunately out of time,
00;20;30;18 - 00;20;33;18
but it has been an insightful
and amazing discussion
00;20;33;29 - 00;20;35;05
and we could have talked for hours.
00;20;35;05 - 00;20;36;15
I think it's been fantastic.
00;20;36;18 - 00;20;41;14
And we started with some of
the opportunities in regional locations,
00;20;41;24 - 00;20;43;04
and then we've moved to
00;20;43;07 - 00;20;47;06
some of the specialist examples
where speed to market is really important,
00;20;47;09 - 00;20;52;08
which is things like AI inference,
low latency locations and the benefits
00;20;52;11 - 00;20;56;29
that prefabrication
can bring to data centers with also
00;20;57;06 - 00;21;00;14
the license ocean bid with PPAs close
to one
00;21;00;22 - 00;21;03;19
that bring very sustainable solutions
to those regional areas.
00;21;03;22 - 00;21;06;08
And it's been a fascinating discussion.
00;21;06;11 - 00;21;07;06
Thank you Ty.
00;21;07;09 - 00;21;10;07
Thank you Hawks. Thank you Sam. You Andrew.
00;21;10;10 - 00;21;12;02
See ya, bye. Thanks, Andrew.
00;21;16;18 - 00;21;19;18
Brought to you by Nexus, Publish By GHD.
00;21;19;26 - 00;21;21;00
Where ideas connect
Nexus, Publish By GHD.
00;00;04;08 - 00;00;05;27
Where ideas connect.
00;00;08;20 - 00;00;09;20
Hello,
everyone.
00;00;09;20 - 00;00;12;19
This is part two of our podcast.
00;00;13;01 - 00;00;17;29
In part one, we covered how infrastructure
constraints are driving innovation.
00;00;18;09 - 00;00;20;25
Now we shift focus to what's next.
00;00;20;28 - 00;00;23;12
What's the future look like?
What's the next frontier?
00;00;23;15 - 00;00;28;23
And pleasingly today we're again
joined by hacks Sam and Ty.
00;00;29;02 - 00;00;32;02
So thanks for staying with us
to continue the conversation.
00;00;32;05 - 00;00;33;09
Welcome, guys.
00;00;33;12 - 00;00;35;27
Hello. Good to be back here.
00;00;36;00 - 00;00;38;08
So in this episode,
we're going to chat through
00;00;38;08 - 00;00;41;12
the untapped
potential of regional data centers.
00;00;41;19 - 00;00;44;11
And AI inferencing hubs.
00;00;44;14 - 00;00;47;24
We'll look at how these emerging models
are enabling latency
00;00;47;27 - 00;00;52;04
sensitive applications and reshaping
infrastructure investment strategies.
00;00;52;19 - 00;00;53;20
And it's something that I think
00;00;53;20 - 00;00;56;28
is, really a nice bridge
from our previous episode
00;00;57;01 - 00;01;00;23
where we looked at how technology
is really driving innovation and the rise.
00;01;00;26 - 00;01;03;06
And, you know,
I was kind of quite exciting
00;01;03;06 - 00;01;06;06
at the end of that to hear about,
what the future might hold.
00;01;06;18 - 00;01;10;19
But I also think that there's a big future
that we didn't discuss, which is looking
00;01;10;24 - 00;01;14;17
at some of our regional locations,
which perhaps traditionally
00;01;14;20 - 00;01;16;17
haven't been the focus in the market.
00;01;16;18 - 00;01;18;06
The market has primarily been
00;01;18;06 - 00;01;22;04
focused in big city
locations, often for latency reasons.
00;01;22;07 - 00;01;25;07
But, you know,
I think that we are seeing a big shift
00;01;25;12 - 00;01;28;29
to the opportunities
that are available in regional locations
00;01;29;07 - 00;01;32;02
for things
like AI training, as one example.
00;01;32;05 - 00;01;34;24
But maybe to start with you, Sam,
are you able to
00;01;34;25 - 00;01;38;03
perhaps give us some insights
as to why regional locations are becoming
00;01;38;09 - 00;01;41;09
more and more viable for our data centers?
00;01;41;15 - 00;01;46;17
I guess this does link quite nicely to,
the previous discussion we had, but,
00;01;47;00 - 00;01;50;07
the regional appeal is coming
because, the resource
00;01;50;10 - 00;01;55;00
availability in the central
populated cities is diminishing.
00;01;55;14 - 00;02;00;02
So, with
I guess, lack of access to the approvals
00;02;00;05 - 00;02;03;12
for development, land,
water, energy in the cities.
00;02;03;25 - 00;02;07;25
Then, you see these,
developers are looking elsewhere
00;02;07;28 - 00;02;10;27
for solutions to satisfy that demand.
00;02;11;00 - 00;02;14;06
So that's where they cost
the net larger and broader,
00;02;14;09 - 00;02;17;25
and they find out where, the markets
that don't require
00;02;18;04 - 00;02;21;21
the immediate service of inner city,
connectivity,
00;02;22;00 - 00;02;25;25
they become a viable customer
in these regional locations.
00;02;26;15 - 00;02;27;10
Yeah. It's interesting.
00;02;27;10 - 00;02;31;03
I also think one of the other things
that I am personally keen to say
00;02;31;06 - 00;02;34;06
is that making sure we don't get like a,
a great digital divide
00;02;34;11 - 00;02;37;19
where it becomes sort of, you know,
those that live in cities have access
00;02;37;22 - 00;02;40;29
to the greatest, technology
and AI infrastructure,
00;02;41;02 - 00;02;43;19
digital infrastructure
compared to those in the region.
00;02;43;20 - 00;02;45;21
So that's the other pleasing thing,
00;02;45;21 - 00;02;49;09
is to hopefully enable
that for the regional communities.
00;02;49;21 - 00;02;52;09
And that's a great point
because, connectivity
00;02;52;09 - 00;02;56;01
in regional zones
is increasing incredibly quickly.
00;02;56;04 - 00;03;01;05
The interconnections between countries
are, increasing at an incredible rate.
00;03;01;17 - 00;03;05;01
And internally,
infrastructure developments are now
00;03;05;04 - 00;03;10;14
including high, connectivity for,
these restricted access areas.
00;03;11;05 - 00;03;15;16
But, what we are trying
to, use as our target market
00;03;15;19 - 00;03;20;25
is that regional interest
and finding a demand where we can connect
00;03;20;28 - 00;03;27;17
to landfall locations with quite a small,
latency and still provide
00;03;27;20 - 00;03;32;13
a reliable, connected service
that can satisfy that, regional demand.
00;03;32;26 - 00;03;34;09
So, okay, that's quite interesting.
00;03;34;12 - 00;03;38;12
So you're finding that
the regional component offers
00;03;38;19 - 00;03;42;16
the opportunity to connect closer to it,
where there's nowhere near regional area
00;03;43;01 - 00;03;46;00
to achieve the latency parameters,
because your
00;03;46;03 - 00;03;49;02
your nearby regional town
or regional area.
00;03;49;05 - 00;03;50;29
Yeah, it goes down to, physics.
00;03;50;29 - 00;03;53;16
And forgive me for,
not being too technical
00;03;53;16 - 00;03;56;20
on this topic
and asking that jumped in, but,
00;03;57;10 - 00;04;00;17
I've got the great reference
that, we're looking at Fiji and would love
00;04;00;20 - 00;04;03;23
to, have our breakout,
development there.
00;04;04;07 - 00;04;08;08
And, we've got a round trip
of 30 milliseconds from Sydney.
00;04;08;18 - 00;04;12;23
And so that's a service
that, really is quite functional
00;04;12;26 - 00;04;16;11
for quite a lot of different
consumers of data,
00;04;16;25 - 00;04;20;18
which is where we targeting
to, build up our customer base.
00;04;21;02 - 00;04;22;08
Super interesting.
00;04;22;11 - 00;04;25;15
I think it's fantastic
because also with the shift away
00;04;25;18 - 00;04;30;03
from the big city centers and also,
you know, looking to advance manufacturing
00;04;30;06 - 00;04;34;08
and build communities
and sort of the old haven't spoke model,
00;04;34;14 - 00;04;36;17
we need to be able to provide
the digital infrastructure
00;04;36;17 - 00;04;39;03
to enable those cities
and communities to thrive.
00;04;39;03 - 00;04;42;02
And so I think that's the opportunity
there to,
00;04;42;05 - 00;04;44;10
you know, not just for private investment,
but for government
00;04;44;10 - 00;04;48;06
to start to incentivize investment
in regional locations, to really make sure
00;04;48;09 - 00;04;51;12
that the industry is keeping pace
with the rest of the globe.
00;04;52;10 - 00;04;55;28
And I guess on investors tie, I'm
interested to hear from your point of view
00;04;56;01 - 00;04;59;20
and what you've been seeing in the market
around some of the opportunities
00;04;59;23 - 00;05;03;08
that investors can unlock the value
in regional locations.
00;05;04;02 - 00;05;06;09
There's so many things
you could potentially unlock,
00;05;06;09 - 00;05;06;26
but I mean, how
00;05;06;29 - 00;05;07;05
when you
00;05;07;07 - 00;05;08;20
when you think about like communities
00;05;08;20 - 00;05;11;23
and landlocked areas
or infrastructure investors,
00;05;11;26 - 00;05;14;26
like when they use the word value,
what do you think they mean
00;05;15;17 - 00;05;17;01
in terms of from an investor perspective?
00;05;17;01 - 00;05;21;21
I think it means long term
sustainable returns and having their brand
00;05;21;24 - 00;05;26;05
associated with some positive change
for society in the world.
00;05;26;25 - 00;05;28;00
So I think it's a mix of those.
00;05;28;00 - 00;05;32;26
And I think if you look at this
AI space over the next five years or so,
00;05;33;10 - 00;05;36;10
it's going to be the biggest deployment
of capital in human history.
00;05;36;19 - 00;05;40;29
And so the early investors who I think
deploy capital wisely
00;05;41;08 - 00;05;45;28
will get very good returns and be able
to sort of grow exponentially with this.
00;05;46;14 - 00;05;49;03
You know, we're just starting
with really early in the race.
00;05;49;06 - 00;05;52;18
And I think, as Andrew mentioned,
we need to make sure the digital divide
00;05;52;21 - 00;05;53;22
isn't there.
00;05;53;25 - 00;05;56;14
And if you look at the growth
of usage of ChatGPT
00;05;56;17 - 00;05;59;29
or other models we're at,
that's about a billion people using them.
00;06;00;04 - 00;06;01;29
There's 8 billion people on the planet.
00;06;02;02 - 00;06;05;09
And I think all these efficiencies
are happening are fantastic.
00;06;05;12 - 00;06;06;20
And I think we need to go further.
00;06;06;20 - 00;06;08;13
You know,
we need to reduce the cost of these models
00;06;08;13 - 00;06;12;17
from $200 a month for the high end ones,
or $20 a month for the regular
00;06;12;20 - 00;06;17;08
use ones, down to $2 or $0.20
for 8 billion people around the world.
00;06;17;11 - 00;06;20;11
Regardless of you know which situation
they're in to be able to use it.
00;06;20;23 - 00;06;22;16
And so we need to enable that
through efficiency.
00;06;22;16 - 00;06;25;15
And technology is
the biggest efficiency driver you'll find.
00;06;25;15 - 00;06;27;01
And we've seen this in the past.
00;06;27;01 - 00;06;29;07
You know, the cost comes down
orders of magnitude.
00;06;29;10 - 00;06;31;29
But the growth in terms
of users will go up.
00;06;32;02 - 00;06;36;15
And I think providing that level of
intelligence to every human in the world.
00;06;36;18 - 00;06;40;06
I'm really optimistic that there's
some smart young people around the world
00;06;40;09 - 00;06;42;14
that are going
to create some magic with it.
00;06;42;17 - 00;06;47;01
And I think that's our responsibility is,
you know, who's using behind the meat's,
00;06;47;06 - 00;06;50;03
sustainable power,
putting the data centers in place,
00;06;50;03 - 00;06;51;06
providing the GPUs,
00;06;51;06 - 00;06;54;06
bringing down the cost of these models,
especially on the inference side.
00;06;54;11 - 00;06;56;29
Is our responsibility
to better enable that future.
00;06;57;02 - 00;07;00;24
And so I'm really curious,
once we get to 2030, 2035
00;07;00;27 - 00;07;05;15
and hopefully we've got the cost down
to something like $0.20 a month for folks
00;07;05;18 - 00;07;06;11
to use.
00;07;06;14 - 00;07;09;09
It's going to be really interesting
to see what comes out of that.
00;07;09;12 - 00;07;11;19
I want jump in on that one as well,
because I'm quite heavy
00;07;11;19 - 00;07;13;12
in this topic currently.
00;07;13;12 - 00;07;18;14
And one of the great synergies
that we have with ocean is that,
00;07;18;24 - 00;07;23;15
we are looking at power generation
and data center provision as a bundle.
00;07;24;03 - 00;07;27;16
If we can do data center
as a single service, fantastic.
00;07;27;19 - 00;07;32;17
But for, the discussions around
infrastructure investment and development,
00;07;33;04 - 00;07;36;10
they are still locked into thinking
infrastructure
00;07;36;13 - 00;07;41;02
is energy, water
and other basic services to the community.
00;07;41;05 - 00;07;44;24
Data is not an infrastructure
kind of thing.
00;07;44;27 - 00;07;47;14
It's still just a telco investment.
00;07;47;17 - 00;07;53;09
But linking that data into development,
into something that can underwrite
00;07;53;21 - 00;07;57;00
the development of energy
as a massive development.
00;07;57;03 - 00;08;01;19
Knowing that you've got a data center
providing a guaranteed consumer
00;08;01;22 - 00;08;04;22
for X amount of years, that stability
00;08;05;02 - 00;08;07;18
breeds more investment.
00;08;07;21 - 00;08;11;11
And so, yeah, it's
kind of using the data center demand
00;08;11;14 - 00;08;16;17
as a leverage to actually do
greater energy development
00;08;16;20 - 00;08;20;17
that is going to service the community
where that development
00;08;20;20 - 00;08;23;20
may not have been financially feasible
prior to that.
00;08;24;04 - 00;08;24;25
Yeah.
00;08;24;28 - 00;08;28;02
I think my Andrew, my colleague Salmon
Hawks have helped me answer that question.
00;08;28;05 - 00;08;29;08
Yeah.
00;08;29;11 - 00;08;31;23
One of the things we haven't touched on
is that, you know, in
00;08;31;23 - 00;08;35;18
regional locations is the ability
to build closer to the generation.
00;08;35;21 - 00;08;38;19
So looking at, you know, being able
to build closer to renewable energy
00;08;38;22 - 00;08;43;14
sources, being able to build closer
to different water, supplies.
00;08;43;23 - 00;08;47;18
And, you know, that enables things
like AI training to be done
00;08;47;27 - 00;08;49;15
in regional locations.
00;08;49;18 - 00;08;51;16
So those sorts of opportunities
00;08;51;16 - 00;08;55;03
out in the regions to get investment into
AI training is huge.
00;08;55;26 - 00;08;59;15
Also, just due to geopolitical challenges
at the moment and,
00;08;59;18 - 00;09;02;09
where things are being built
and being invested,
00;09;02;12 - 00;09;05;09
there's sovereign capability
with a lot of uncertainty in the globe
00;09;05;09 - 00;09;09;13
at the moment,
being able to train your own capability
00;09;09;16 - 00;09;12;19
and knowledge of data
and the ability to control data and to get
00;09;12;21 - 00;09;16;01
smarter is is kind of become
a sovereign capability, if not already.
00;09;16;04 - 00;09;20;05
So I think that, you know, being able
to do AI training on, on home
00;09;20;08 - 00;09;21;16
soil in regional locations
00;09;21;16 - 00;09;24;14
is going to be become more
and more prevalent across the globe
00;09;24;17 - 00;09;27;19
and will not just be limited
to a select few nations.
00;09;28;19 - 00;09;31;11
And look, we did just mentioned
latency there.
00;09;31;11 - 00;09;35;21
And some of the, I guess, challenges,
but also opportunities in the regions.
00;09;35;24 - 00;09;37;28
But one of the challenges
with regions is latency.
00;09;37;28 - 00;09;42;15
But we might actually zoom back in
to the opposite end of the spectrum.
00;09;42;23 - 00;09;46;05
And perhaps, Hawkes,
I might start with you just to talk around
00;09;46;08 - 00;09;49;18
what ultra low latency is
and why it's important
00;09;49;21 - 00;09;53;00
for data centers and what what's
the sort of some of the types of uses
00;09;53;03 - 00;09;56;16
where ultra low latency
is going to be important in data centers?
00;09;57;05 - 00;09;58;00
Yeah, absolutely. Andrew.
00;09;58;00 - 00;10;01;04
So, I think as you mentioned,
there will be some regional
00;10;01;07 - 00;10;03;06
and there'll be
some national capabilities.
00;10;03;06 - 00;10;03;15
You know,
00;10;03;18 - 00;10;06;20
I think there's general intelligence
and then there's sovereign models,
00;10;06;23 - 00;10;07;12
as you mentioned.
00;10;07;12 - 00;10;09;15
And geopolitics is providing
00;10;09;18 - 00;10;12;25
another nuance on top of that, especially
with the rise of sort of nationalism
00;10;13;12 - 00;10;16;01
in terms of ultra low latency,
where it becomes
00;10;16;01 - 00;10;18;13
really important is on the
AI inferencing side.
00;10;18;16 - 00;10;22;03
So inferences where AI is monetized,
that's where you deploy the models
00;10;22;06 - 00;10;24;12
and people prompt them,
and that's where you pay the subscription.
00;10;24;12 - 00;10;26;08
Whichever model you choose to use
00;10;26;11 - 00;10;28;14
to be able
to get the replies to your prompt
00;10;28;14 - 00;10;30;15
and the intelligence and tokens
that you need.
00;10;30;18 - 00;10;33;16
So inference to scale
is going to be distributed.
00;10;33;19 - 00;10;37;04
Being close to
I say users is really important
00;10;37;07 - 00;10;41;02
in that sort of ultra low latency
and uses today I think are humans,
00;10;41;17 - 00;10;44;15
but users in the future might be machines
and so really depends.
00;10;44;18 - 00;10;45;25
Although we think about it as cities,
00;10;45;26 - 00;10;48;03
it also depends on
what entities users are.
00;10;48;06 - 00;10;51;05
So I think as we unpack that,
as we go forward as an industry,
00;10;51;08 - 00;10;53;29
it's about being close
to whoever that user is.
00;10;54;02 - 00;10;56;08
And we've found to be able
to deliver distributed
00;10;56;08 - 00;11;00;06
AI and inference, you need a really
a standard rapid deployment model.
00;11;00;09 - 00;11;02;22
You need something that's standardized
that you can play in multiple ways.
00;11;02;22 - 00;11;05;21
And this is where the prefab model
does really well.
00;11;05;24 - 00;11;09;20
And so the inference side to your point
is a data security piece.
00;11;09;23 - 00;11;10;03
So you know,
00;11;10;03 - 00;11;11;22
if you think about the sovereign models,
00;11;11;25 - 00;11;14;14
the sovereign models
are going to be for national security.
00;11;14;14 - 00;11;17;02
But there's also going to be civilian
capabilities as well.
00;11;17;02 - 00;11;20;13
So a country will make available
government services
00;11;20;24 - 00;11;23;03
to all of its citizens, essentially.
00;11;23;06 - 00;11;26;13
And so that will need to be inference
for citizens.
00;11;26;16 - 00;11;27;14
And,
00;11;27;15 - 00;11;32;01
you know, returns will be significant
here in the sense of lowest risk returns.
00;11;32;08 - 00;11;34;13
The sovereign models
and the sovereign deployments,
00;11;34;13 - 00;11;37;17
I think will be most attractive
to investors, and then it will go down
00;11;37;20 - 00;11;41;04
to investment grade, and then it's a more
of the more speculative stuff.
00;11;41;07 - 00;11;41;22
Essentially.
00;11;41;22 - 00;11;46;10
So ultra low latency data security
and investors is really all comes together
00;11;46;23 - 00;11;49;28
with our inferencing around
general intelligence and sovereign models.
00;11;50;25 - 00;11;51;02
Yeah.
00;11;51;02 - 00;11;55;28
Just, as one example of the machines,
it's things like self-driving cars.
00;11;56;01 - 00;11;58;14
You know, I certainly hope
that it doesn't take too long
00;11;58;14 - 00;12;01;09
to make a decision
when something runs in front of the car.
00;12;01;09 - 00;12;01;27
You certainly want
00;12;01;27 - 00;12;06;04
the ultra low latency there
so that, a decision can be made by AI.
00;12;06;07 - 00;12;07;12
Yeah.
00;12;07;15 - 00;12;10;12
So I guess the other thing to,
00;12;10;12 - 00;12;13;14
with you on the line heart is, you know,
what do you think?
00;12;13;17 - 00;12;17;08
You know, sort of next wave of data
center projects with, like
00;12;17;11 - 00;12;21;11
you said, speed to market, ability
to get projects online quickly.
00;12;21;21 - 00;12;22;05
What are those
00;12;22;05 - 00;12;25;10
sort of investors looking for
in the next wave of data center projects?
00;12;26;09 - 00;12;28;02
Yeah, I think as we look at the next wave
00;12;28;03 - 00;12;30;00
and as I mentioned,
we're early in this space,
00;12;30;00 - 00;12;32;29
but it is going to be the biggest
deployment of capital in human history,
00;12;33;02 - 00;12;33;20
as I mentioned.
00;12;33;20 - 00;12;38;08
And so the early investors who get in
and pick the right bets, I would say,
00;12;38;20 - 00;12;39;23
are going to do really well
00;12;39;25 - 00;12;42;26
and probably have really
attractive returns in the longer term.
00;12;43;09 - 00;12;46;00
It's almost like
we're at the start of the internet era,
00;12;46;03 - 00;12;47;25
but it's going to accelerate
a lot quicker.
00;12;47;25 - 00;12;48;27
What happened in 20 years
00;12;48;27 - 00;12;51;29
and the internet is going to happen
in five years with AI essentially.
00;12;52;02 - 00;12;56;06
So there's huge opportunities here
if investors pick the right places
00;12;56;09 - 00;12;57;16
to back, essentially.
00;12;57;19 - 00;13;01;14
And I think, as you mentioned
earlier, power is the biggest constraint.
00;13;01;20 - 00;13;03;29
And so it's about
how do you get the most out of that power.
00;13;04;02 - 00;13;07;01
And the way to get the most out of
the power is to be able to deploy the most
00;13;07;04 - 00;13;10;03
tokens, deploy the latest Nvidia chips,
00;13;10;06 - 00;13;12;28
and have data centers
that evolve with the speed of technology.
00;13;12;28 - 00;13;15;06
And that's what we're focused on
is to be able to do that.
00;13;15;06 - 00;13;18;24
And so whether it's behind the meter,
the power that Sam generating
00;13;19;05 - 00;13;21;14
or this grid power,
you know, it's about saying, okay,
00;13;21;14 - 00;13;24;07
we have a limited amount of power,
how do we get the most amount of tokens?
00;13;24;07 - 00;13;24;28
And that's how you grow.
00;13;24;28 - 00;13;25;10
Revenue
00;13;25;10 - 00;13;27;07
essentially, is the number of tokens
that you're able to deliver
00;13;27;07 - 00;13;28;29
and the intelligence
you're able to deliver.
00;13;29;02 - 00;13;31;07
So I think you need to make sure
you look across the whole chain.
00;13;31;07 - 00;13;34;10
As we've mentioned before,
the whole stack, the whole ecosystem.
00;13;34;20 - 00;13;37;21
And then also make sure
you find the right sort of models
00;13;37;24 - 00;13;41;02
and the right risk
adjusted, projects to get involved in.
00;13;41;23 - 00;13;42;03
Yeah.
00;13;42;03 - 00;13;46;11
And so I'm just interested in your views
as well, in terms of, where you see
00;13;46;18 - 00;13;50;00
the next wave of projects coming,
particularly perhaps
00;13;50;03 - 00;13;53;04
in some of those regional locations
and what it might mean for ocean Bay.
00;13;53;29 - 00;13;56;14
Certainly a portion of the market
that I'm watching
00;13;56;14 - 00;13;59;19
closely is the sustainable development
of these data centers.
00;13;59;23 - 00;14;03;18
So we're seeing constriction
in all major cities.
00;14;03;21 - 00;14;07;10
The bottlenecks are stretching,
and that constriction
00;14;07;13 - 00;14;09;21
isn't going to go away
until the demand goes away.
00;14;09;21 - 00;14;14;06
And, I mean, Hawks is pretty confident
this demand is here to stay, and I agree.
00;14;14;20 - 00;14;20;08
And so the next wave is moving
to the tiered structure of deployments
00;14;20;11 - 00;14;25;11
of data centers that, not reducing value
for the communities around them,
00;14;25;13 - 00;14;29;28
that they're not impacting the resources
required for everyday life.
00;14;30;11 - 00;14;33;00
And, adding that extra value.
00;14;33;03 - 00;14;36;02
And really,
that's that's what we are targeting.
00;14;36;02 - 00;14;38;05
We are trying to target,
00;14;38;05 - 00;14;41;02
the demand that is growing
at these regional data centers.
00;14;41;05 - 00;14;42;27
They're being pushed
out of the major cities.
00;14;42;27 - 00;14;46;00
They're they're being sent
to locations that
00;14;46;03 - 00;14;49;07
people kind of go, well,
we don't have the power available yet.
00;14;49;10 - 00;14;51;24
So we got, you know,
what are we going to do?
00;14;51;24 - 00;14;53;22
We want the data,
but we don't have the space.
00;14;53;22 - 00;14;56;13
We don't have the energy,
we don't have the water.
00;14;56;16 - 00;15;00;25
And, Ocean Beat wants to be there saying,
well, we've got a solution for that.
00;15;00;28 - 00;15;05;00
And it's baseload renewable energy
and we're not going to take any land.
00;15;06;01 - 00;15;08;14
I think
I'll just add to what Sam mentioned there.
00;15;08;14 - 00;15;12;10
From an investor perspective,
there's about $70 trillion,
00;15;12;13 - 00;15;17;08
which is going to pass from baby boomers
to Gen Z in the next 20 years.
00;15;17;24 - 00;15;21;24
And the idea is that that $70 trillion
as it pass to the next generation, that
00;15;21;27 - 00;15;24;26
that would be invested
in more of the ESG, sustainable,
00;15;24;29 - 00;15;28;10
renewable
kind of opportunities, projects and funds.
00;15;28;13 - 00;15;31;13
And so there's a huge amount of capital
which is going to transfer
00;15;31;21 - 00;15;32;29
and be invested in these areas.
00;15;32;29 - 00;15;35;29
So I think you will see
that capital will flow over time as well
00;15;36;02 - 00;15;39;04
into this spaces,
and it'll become more of the norm now.
00;15;39;06 - 00;15;40;20
I think one of the other things that we
00;15;40;23 - 00;15;42;25
perhaps haven't touched on,
and just before we move to
00;15;42;25 - 00;15;46;04
finalize, is the reason why
there's a demand for data centers.
00;15;46;07 - 00;15;47;26
And you know why the data centers exist.
00;15;47;26 - 00;15;51;19
It's because it's going to enable progress
of the human race.
00;15;51;22 - 00;15;54;06
And it's a really important
it's the next evolution.
00;15;54;06 - 00;15;56;10
And it's really exciting to be part of.
00;15;56;10 - 00;15;59;15
And I think the challenge
and the opportunity at the same time
00;15;59;18 - 00;16;03;18
for all of us is what is the role
that we can play to enable
00;16;03;21 - 00;16;06;24
that change and the legacy that we leave
for the next generation.
00;16;07;03 - 00;16;07;21
As you mentioned,
00;16;07;21 - 00;16;11;12
hock $70 trillion worth of funds
being passed from the baby boomers
00;16;11;15 - 00;16;14;27
to the next generation, and certainly
a significant portion of that will be
00;16;15;07 - 00;16;18;24
on digital infrastructure to enable
that productivity and efficiency.
00;16;18;27 - 00;16;22;02
So it's a really exciting time,
for all of us.
00;16;22;05 - 00;16;26;28
And I guess with that, I'm keen to hear
if you had $10 billion,
00;16;27;01 - 00;16;31;04
so $10 billion, pick a currency,
let's call it US dollars in the industry.
00;16;31;07 - 00;16;34;01
Where would you be investing?
So you've got $10 billion.
00;16;34;01 - 00;16;35;04
Where would you invest?
00;16;35;07 - 00;16;37;17
I'll start with you, Sam.
00;16;37;20 - 00;16;41;10
So I did introduce myself
as a co-founder of a, startup.
00;16;41;15 - 00;16;44;14
So, no, I want to take that out.
00;16;44;17 - 00;16;48;12
To be fair,
I would be looking at the strategy
00;16;48;15 - 00;16;51;27
of the standardization
and prefabrication and,
00;16;52;26 - 00;16;53;13
maybe I'll.
00;16;53;15 - 00;16;57;05
If I had that 10 billion,
maybe I'd try and, come across the pond.
00;16;57;19 - 00;17;03;08
But, I think that is increasingly
the required strategy to be
00;17;03;11 - 00;17;06;23
the best in field is ensuring
that you've got a standardized,
00;17;06;26 - 00;17;11;13
repeatable process that ensures
that you can get the best efficiency
00;17;11;16 - 00;17;15;24
out of that construction
phase, reduce lost time,
00;17;16;11 - 00;17;19;18
and really just ensuring
that you've got best in class, fast
00;17;19;21 - 00;17;24;27
development, capable,
tied in with a solution that brings
00;17;25;00 - 00;17;27;29
in that social license
to continue to operate because,
00;17;28;03 - 00;17;29;24
you know, fast development is great.
00;17;29;24 - 00;17;33;16
But without that continued social losses,
that $10 billion is not coming back.
00;17;33;28 - 00;17;35;09
Well said.
00;17;35;12 - 00;17;38;11
And tie,
where would you invest $10 billion.
00;17;38;14 - 00;17;40;29
Yeah, $10 billion is a lot of money.
00;17;40;29 - 00;17;44;07
But it sounds not like a lot
compared to the 70 trillion
00;17;44;10 - 00;17;46;05
that's about to be transferred
from one generation
00;17;46;05 - 00;17;47;25
to another, or indeed the multi-trillion
00;17;47;25 - 00;17;50;00
that's going to be spent in the next
ten years.
00;17;50;03 - 00;17;53;03
Can I cheat
and give five to Sam and five to Hawks?
00;17;53;22 - 00;17;55;24
But that's cheating for me.
00;17;55;27 - 00;17;58;03
I do, you shorten. We're sorry.
00;17;58;06 - 00;17;58;17
You're right.
00;17;58;20 - 00;18;01;11
No, no. But, that's cheat answer. But,
you know, maybe.
00;18;01;11 - 00;18;05;07
Is there something in,
like setting up a, a global competition
00;18;05;18 - 00;18;10;16
where the person who gets to a P of X
Hawks in number?
00;18;10;21 - 00;18;14;11
What's a magic number
that you'd love to get to in terms of P or
00;18;14;13 - 00;18;16;04
in terms of investment? P.
00;18;17;08 - 00;18;19;10
It needs to be less sub 1.1.
00;18;19;13 - 00;18;20;04
Yeah. Right.
00;18;20;04 - 00;18;25;12
So the first person that can get to 1.05
gets $10 billion.
00;18;25;27 - 00;18;28;04
Sam, what's the P of I should be.
00;18;28;07 - 00;18;28;15
Yeah.
00;18;28;16 - 00;18;32;06
Well, I was just going to say
when the power that we're generating is,
00;18;32;09 - 00;18;38;06
for moving a lot of seawater that is,
coming out at the system at 17°C,
00;18;38;09 - 00;18;42;05
we have excess cooling that requires
no energy to move it
00;18;42;08 - 00;18;45;08
because we've consumed it
in the power generation cycle.
00;18;45;17 - 00;18;47;28
I believe we are very close to one.
00;18;48;01 - 00;18;49;21
Therefore, you might win the 10 million.
00;18;49;21 - 00;18;51;20
Congratulations
00;18;51;23 - 00;18;53;23
with great competition to.
00;18;53;24 - 00;18;55;29
I love it.
00;18;56;02 - 00;18;59;24
And Huck's interested from your point of
view where you would invest, $10 billion.
00;19;00;15 - 00;19;02;26
Yes. Andrew,
we did some maths on investment
00;19;02;26 - 00;19;05;23
required in the data center space
and in the AI space
00;19;05;26 - 00;19;10;04
generally, and to build a one gigawatt
data center, to just build a data
00;19;10;07 - 00;19;14;09
center is about $12.5 billion to fit it
full of chips
00;19;14;16 - 00;19;18;11
and GPUs
is probably another 30, $35 billion.
00;19;18;29 - 00;19;21;21
And so to have a gigawatt of GPU cloud
00;19;21;21 - 00;19;24;20
is in the order
of about a $50 billion check size.
00;19;24;23 - 00;19;28;27
So I'd come back to you and ask for five x
what you're giving me and say I would,
00;19;29;06 - 00;19;33;26
I would build a gigawatt
GPU cloud powered by renewable,
00;19;33;29 - 00;19;38;08
sustainable energy
and ideally supporting some sovereign
00;19;38;14 - 00;19;42;10
AI workloads, which is the lowest
risk from an investment perspective.
00;19;42;21 - 00;19;45;21
And hopefully,
like you said, giving everybody
00;19;45;26 - 00;19;48;18
Einstein
level of intelligence in their pocket
00;19;48;21 - 00;19;52;19
and then seeing what does that enable
and what does that give to humans
00;19;52;22 - 00;19;54;24
around the world. Nice.
00;19;54;27 - 00;19;59;14
For what it's worth, I'd be picking
some communities in need and investing
00;19;59;17 - 00;20;03;21
co-locating data centers
with some community benefits alongside it,
00;20;03;23 - 00;20;06;09
so that we really change
some of the narrative around
00;20;06;12 - 00;20;09;04
data centers at the moment,
because I think the industry needs it.
00;20;09;07 - 00;20;12;26
And I think that being able to demonstrate
that data centers can give back
00;20;12;28 - 00;20;15;10
to the community
and be a really fundamental part
00;20;15;10 - 00;20;18;13
of the future for the next generation
and the legacy that it leaves
00;20;18;16 - 00;20;19;23
is is really important
00;20;19;23 - 00;20;20;26
because we're not going to change
00;20;20;26 - 00;20;22;19
the world of data centers with 10 billion,
00;20;22;19 - 00;20;25;20
but we will change the lives
of a few communities that really need it.
00;20;25;22 - 00;20;27;29
So that's where I've been visiting
AI 10 billion.
00;20;27;29 - 00;20;30;18
But,
look, we're unfortunately out of time,
00;20;30;18 - 00;20;33;18
but it has been an insightful
and amazing discussion
00;20;33;29 - 00;20;35;05
and we could have talked for hours.
00;20;35;05 - 00;20;36;15
I think it's been fantastic.
00;20;36;18 - 00;20;41;14
And we started with some of
the opportunities in regional locations,
00;20;41;24 - 00;20;43;04
and then we've moved to
00;20;43;07 - 00;20;47;06
some of the specialist examples
where speed to market is really important,
00;20;47;09 - 00;20;52;08
which is things like AI inference,
low latency locations and the benefits
00;20;52;11 - 00;20;56;29
that prefabrication
can bring to data centers with also
00;20;57;06 - 00;21;00;14
the license ocean bid with PPAs close
to one
00;21;00;22 - 00;21;03;19
that bring very sustainable solutions
to those regional areas.
00;21;03;22 - 00;21;06;08
And it's been a fascinating discussion.
00;21;06;11 - 00;21;07;06
Thank you Ty.
00;21;07;09 - 00;21;10;07
Thank you Hawks. Thank you Sam. You Andrew.
00;21;10;10 - 00;21;12;02
See ya, bye. Thanks, Andrew.
00;21;16;18 - 00;21;19;18
Brought to you by Nexus, Publish By GHD.
00;21;19;26 - 00;21;21;00
Where ideas connect
Short, sharp points of view on:
- What makes regional sites viable for AI workloads
- How prefab builds and bundled services reduce setup time
- Why investors are backing sustainable infrastructure outside cities
- How regional access supports equity and resilience
This conversation shows how regional data centres are driving new forms of community value and inclusion.
Smarter insights, tailored to your industry
Cut through the noise with ideas that speak to your world.