September 4, 2025 • 32 mins
Amazon Web Services’ highly anticipated launch of its New Zealand cloud region unravelled this week as it nonplussed journalists and the public alike with PR spin that failed to mask the cracks in its narrative.
In a special episode of The Business of Tech podcast to round out the week, Ben Kepes joins me to break down the blunders, highlighting misplaced numbers, poor communication, and missed opportunities to tell a genuinely positive story.
According to Kepes, a Christchurch-based businessman and board director who has closely followed cloud industry developments for over 20 years, AWS’s attempt to herald its new local infrastructure fell apart due to a mix of political expediency and surprising vendor immaturity.
“There are two stories here, right? The first story is one about investment, NZ Inc, infrastructure, politics, frankly. And that was a total debacle. And it was because of political expediency on the politician side and, frankly, immaturity, surprising immaturity on the vendor side,” Kepes told me.
Listen to the full conversation on this special episode of The Business of Tech powered by 2degrees Business, streaming on iHeartRadio and wherever you get your podcasts.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:12):
I'm Peter Griffin and welcome to the Business of Tech
powered by Two Degrees, a special episode I've put together
to try and unpack a really big tech related story
that played out this week, the launch of Amazon Web
Services New Zealand Data Center Region four years in the making.
It should have been a big milestone to celebrate, but
(00:34):
it turned into a train wreck for Amazon, which really
shot itself in the foot by refusing to explain why
it had seemingly pivoted from building its own shiny, expensive
data centers to leasing space in existing ones and what
that meant for the seven point five billion dollar investment
over fifteen years it promised back in twenty twenty one,
(00:55):
and the thousand jobs it expected to create AWS, well,
that was just an option on the table. Never mind,
we've done something different. Journalists called bs on that the
media coverage was wall to wall brutal. It's perhaps best
summed up by Heather Dupless Allen's news Talk ZB interview
(01:16):
with AW West New Zealand country manager Manuel Bonnett on Tuesday.
Here's a clip from it.
Speaker 2 (01:23):
Have you built any data centers?
Speaker 3 (01:25):
So a region is a cluster of data centers, and
of course we have chosen the locations that we choose
to build out our availability zones. The region consists of
multiple availability zones. We have three of these availability zones
here and of course that involves data centers.
Speaker 2 (01:42):
Cool and how many of these have you actually built
as opposed to renting or leasing.
Speaker 3 (01:47):
So we are choosing facilities based on the needs that
we have, and we have thirty eight of these regions globally.
Speaker 2 (01:54):
Exactly what you're being worried about this though, aren't you, Manuel,
because you haven't built any have you?
Speaker 3 (01:58):
I'm telling you we chose the locations that are the
best looking.
Speaker 2 (02:01):
What it happened? Why didn't you guys build it? It
wasn't that part of the announcement with just in the
back in twenty twenty one that you were going to build.
Speaker 3 (02:06):
Some Well, we always take a long term view and
as part of that we look at all the options
and I'm super happy that we found the right locations
for the region to be live as of today.
Speaker 2 (02:17):
Okay, what happened to the west Aokland one where you
flattened it? You got rid of the wetland? Are you
going to build anything there?
Speaker 4 (02:23):
Well?
Speaker 3 (02:23):
I mean, for now we are announcing that the region
is live, so the availability zones at the manual.
Speaker 2 (02:29):
I just need to tell you, mate, listen. This audience
that listens to this is really smart and they can
hear you not answering the questions. So you can keep
on with that strategy if you want to. But I
need to warn you. We know you're not answering my question.
I'm going to try it again. What are you doing
with the West Aalkland site? Are you going to build
anything there?
Speaker 3 (02:45):
I can't comment on the site. I'm basically I'm announcing
the launch of the region and the site, as you
can tell, is not part of it.
Speaker 2 (02:52):
Okay, thanks mate, appreciate it. Go well. Manuel Bonnette, New
Zealand Country Manager, Amazon Web Services. Oh my gosh, guys,
how bummed to you about this announcement. I thought when
I heard luxon on with Hosking this morning, I thought
this was awesome.
Speaker 1 (03:05):
Amazon really didn't understand what they were getting themselves into
when they dreamt up the narrative behind this launch. The
fact that the Prime Minister then jumped on it for
a political capital just made it so much worse, which
is a shame because going live with a new hyperscale
data center region is a big development. AWS has made
(03:27):
a significant investment in cloud computing and is committing to
having a significant presence in New Zealand. That's a good thing.
So I thought i'd get a true cloud computing industry
expert on to unpack what went on here, or what's
probably gone on. It's hard to know exactly because everything
is shrouded in secrecy when it comes to AWS in
(03:51):
the name of security apparently. Ben Keeps is a christ
Church businessman, board director and cloud industry analyst who was
one of the earliest New Zealand proponents of this migration
to the cloud many businesses have been on. He knows
the AWS tech stack in depth, and he knows a
lot about the economics of hyperscale data centers. So he's
(04:13):
been with his candid take on what went down this week.
Ben Keeps, Welcome to the business of tech. How are
you doing.
Speaker 4 (04:20):
I'm doing well. Thanks for inviting me.
Speaker 1 (04:22):
Well, what a debarcle I mean this week with AWS
probably one of the worst PI disasters in the tech industry, definitely,
But in terms of any big corporation in recent memory
in New Zealand, what's your take on this? What really
went wrong? This week.
Speaker 4 (04:39):
Yeah, I mean, you're right, it was absolutely abysmal. There's
two stories here, right. The first story is one about investment,
enzting infrastructure politics frankly, and that was a total debarcle
and it was because of political expediency on the politician
side and frankly immaturity, surprising immaturity on the vendor side.
(05:06):
The second part of the story, which is actually impactful
and got completely lost, is that ADS is actually setting
up or has set up a region here in New Zealand.
And that answers a bunch of questions that people have
had for going on twenty years now.
Speaker 1 (05:22):
So let's unpack these things. Let's go all the way
back actually to twenty twenty one, the day after AWS
made that announcement, the initial announcement back in the depths
of COVID that they were setting up a data center
region in New Zealand. This is what you wrote in
the New Zealand Herald. The devil is in the detail
and apart from the big number seven point five billion
(05:44):
investment and one thousand jobs, there's a little detail in
the announcement, but details aside. This is exciting and further
puts New Zealand on the map as a credible technology nation.
That was a full four years ago. And what's striking
about that is those same figures are exactly what was announced,
you know, on Monday at Amazon's Cloud Day. And I
(06:05):
think that was the root of the problem, was that
there was no advancement on this or no transparency into
the numbers. How the data centers, because there's three of
them in thew Auckland region are configured. Whether you know
Amazon is building its own, we know it's it's not
at the moment, it's it's co locating. Do you think
(06:26):
that's where the root of the problem started here before
the politics came into it. Just that lack of transparency,
which key with is you know, straight up people that's
sort of what they expect in dealing with companies.
Speaker 4 (06:38):
Yeah, so seven point five billion back in twenty twenty
one was a terrible number because power, you know, buying
smoke over the construction crew, you know, the driver that
drives the you know, the aws CEO to cut a ribbon.
I mean, it's ridiculous. It was a ridiculous number then
and it's an even more ridiculous number now. However, the
(07:00):
thing that's really stark is that if AS had in
fact built three of its own data centers, the investment
would have been would have been massive. I don't know,
you know, it wouldn't have been seven point five billion,
but it would have been massive. The investment now is
significantly smaller because for all very good reasons, they are
(07:20):
embarking on a collocation strategy, which means that they're using
data centers that already exist. Obviously higher levels of capacity
within those data centers, but putting in a few extra servers,
or even lots and lots of extra servers, is a
hell of a lot cheaper than building an entire new
data center. So frankly, and the thing that really disappoints
(07:41):
me about the AOS launch, their part of it is
that to put out a statement that said seven point
five billion dollars a thousand jobs, when everyone knows the
data center is still just a hole in the ground
or not even and they are doing COLO is not
only is it duplicitus, but it actually doesn't recognize the
(08:03):
fact that we're not stupid, you know. And I've given
that feedback to my people with an AS. It was
a terrible launch. They didn't do themselves any fabors and
the really great thing about a region got lost because
they put out the release and obviously, you know, the
government decided this is a really good news story and
(08:25):
jumped onto it without being well briefed upon what it
actually meant.
Speaker 1 (08:29):
Yeah, it is. There is a good story buried in there,
and we'll get to that. But just a bit more
under seven point five billion. I mean, you just have
to look at the AWS financial accounts over the last
few years. You know, typically a tech company, you know,
big tech, they usually spend about twelve and a half
percent off their sales, their revenue on capital expenditure. You know,
(08:52):
AWS did last year in New Zealand four hundred and
twenty five million dollars in revenue, really healthy business. But
when you look at their cap X, you know, purchases
of property and equipment, it's all in the you know,
sixteen million in twenty twenty one, eleven million in twenty
twenty two, eight point four million, and twenty twenty three,
it's clear that they're just not spending the big dollars
(09:15):
that they promised, which would have been to the tune
of five hundred million dollars a year. If you're looking
at seven point five billion over fifteen years. So they
they may have had these aspirations at one point, but
when those builds were off the table and they moved
to leasing arrangements, that probably cut the investment by half
(09:36):
at least.
Speaker 4 (09:37):
Well well even beyond that. I mean, even if you
look at the original number with the building out data
center a strategy, So four hundred million revenue a year,
let's say you know they double that, triple that, quadruple that,
you know, a couple of billion dollars revenue a year.
You don't really spend seven point five billion dollars, you know,
building out on CAPEX, building that out in a very
(10:00):
finite market. I meant the end of the day, New
Zealand is always a small market for everything. And so
you know, that's why I said in twenty two one,
the devil is in the detail. I didn't believe that
the market opportunity here justified seven point five billion dollars.
The other, the other aspect that a few commentators have
sort of talked upon is that, you know, it was
(10:22):
it was touted as sort of corporate philanthropy and magnanimity
and how amazing as is investing in New Zealand and frankly,
you know, go capitalism there. They're a for profit organization.
They invest to make a return and so therefore there's
no philanthropy in there whatsoever. They look at a capex
(10:43):
and the payback on that. It's it's about net present
value and discounted cash flows and those sorts of things.
And so you know, the I mean the stuff they're
doing around training, you know, that's that's self self interested.
But it's great, it's awesome. It's a good thing for
NTID inc. Again, like tying all the stuff together, seven
point five billion dollar investment in New Zealand, training, one
(11:05):
hundred thousand people doing all this stuff, and we're so great.
I mean, just New Zealanders aren't wired that way. And
you know, I mean frankly, you know, like I know
the AS leadership here in Apja and in Seattle, and
I don't think they've had advice from you know, internally
(11:25):
or from their external PR people around the New Zealand
psyche and what moves the needle. And unfortunately they barked
on embarked on a PR message that went down frankly
like a cup cold sick with Kiwi's and they're paying
the price. For it now. Yeah.
Speaker 1 (11:41):
Look, I had an off the record briefing with AWS
just last week, which you know I can't talk about.
I agreed to keep it off the record, but I
will say I came out of that meeting thinking just
trouble ahead here because a lot of the questions that
I've had over the years and other journalists have had,
they still weren't answering. And to go out and not
(12:03):
have credible answers to these questions, it was going to
go badly.
Speaker 4 (12:07):
The thing is there, right, So I understand that from
a corporate strategy perspective, they don't want to tell you
the truth and the whole story. I get that, but
then don't tell a story. And so if your thing
is we're secretive, we're not going to talk about this,
then then you know, when you're briefing your PR people,
you say, right, our strategy, our corporate policy is we're
(12:29):
not going to disclose this, this, and this. Then the
advice should be okay. Given that lead with we're launching
the New Zealand region. This is an amazing thing, as
we'll talk about later on. Don't say anything else. Don't
say about investment, don't say about all of those other things,
and honestly, the story would have been so positive because
(12:49):
it's a really good thing. But whether it's arrogant, whether
it's ignorant. For some reason, they felt they had to
embark on this common strategy and it's blown up in
their face.
Speaker 1 (13:01):
And then on top of that, we had the political
angle we had lux and come in and look to
be fair to the Prime minister. I don't think he
actually tried to explicitly claim credit for this. What he
did is his typical rah, rah, isn't this great? You
know anything at the moment he's trying to jump on
that shows growth of any description, and he hyped that up.
(13:25):
He was probably given all the key numbers from AWS
and his own team, which were exactly the key numbers
that justinto Ardern was talking about, you know, four years ago,
and he did his typical sort of you know, hype
machine shtick, and but it just it just politicized it
at a time when he's under pressure and there isn't
(13:47):
much good news. That really didn't help, did it, well.
Speaker 4 (13:50):
I mean it didn't. And you know, like I've talked
to Chris a few times in the past. He has
my phone number, you know, like it was obvious what
was going to happen. All the editors were gonna say, like, hey,
what's the way to position this? I want to make
some political capital, I want to get some wins. How
do I position this? And there the story is, Wow,
this is a great thing. We are such a powerhouse
(14:10):
of technology in New Zealand that AWS is investing and
is building a region here, and that's such a win.
Technology is such an opportunity for us. Success stories like
zero and vend or whatever are now going to find
it easier to get ahead because of this, you know,
this regional buildout, and no one could argue with that
(14:30):
because that's actually the truth. But instead, you know, like
I don't think I mean, I don't know like the
pundits within the National Party when they're doing their polling,
do they see this as neutral or or a loss
because I don't think it's a political win.
Speaker 1 (14:47):
It's definitely not. And look, you know the issue about
not building the data centers, right, it's disappointing. We want
to see more infrastructure come in, but the reality is
is there three data centers that a lot of Amazon
equipment will be going into, if not already in there.
This is not unusual around the world. I mean, there's
(15:08):
a big operator in Asia, Pacific air Trunk, huge data
center operator. It works with the big three Aws, Microsoft,
and Google. So they don't always build their own. They
do prefer to because they have complete control over the facilities.
But colocation is a legitimate thing in the data center industry.
They sort of blew that as well by not saying that, hey,
(15:31):
look we tried to build, it didn't work out, we're
co locating.
Speaker 2 (15:35):
Yeah.
Speaker 4 (15:35):
The difficulty here is that, and I do sort of
feel freight os here is that the hyperscalars are very
sensitive about the environmental impact of hyper scale data centers,
and so they've invested hugely in green energy, so green
energy in terms of input, but also low power servers
(15:55):
and really interesting ip around calling of the service because
obviously calling is a huge and when you go COLO,
you haven't got that same control. And so all of
a sudden, this strong story about you know, adel mess,
you know, green environmental credentials goes out the window. And
so it's a really tricky one. I mean, you know
the fact that they've done forward deals for green energy,
(16:18):
which as almost everyone in New Zealand knows actually is
irrelevant because you're taking from the grid and when the
grid hasn't got enough, you burn coal huntly. So again,
like you know, I don't know what the comms and
PR team within the organization are paid, but I can
imagine imagine it's quite a lot to not simply say,
(16:38):
you know, do a chat GBT, hey tell me about
the New Zealand energy sector and what happens when there's
not much power, you know, like it's not rocket science. Yeah,
they've tried. I mean, I think it's probably a situation
where you've got a big multinational is trying to have
a somewhat consistent approach in terms of regional business development,
(17:00):
and we see we saw the perils with that.
Speaker 1 (17:03):
I guess the way AWS and Microsoft, to a lesser extent,
they have a little bit more autonomy in each country,
but Google and Aws, Apple, it's the company line everywhere.
So trying to have some nuance for the New Zealand
market isn't really going to cut it when you're talking
to Singapore or Seattle. So that is a problem. But
the energy thing did get a lot of play this week,
(17:25):
and I guess the fundamental issue is Kiwis are concerned
that they've heard a lot about the AI applications in
data centers. Data center is being built specifically for AI.
Zuckerberg is building one in Louisiana, fifty billion dollar data center.
The energy needs for that are massive. So they're looking
around the world and going is that going to happen
(17:47):
to us? We're already seeing mills shut down and potential
brownouts when you know the lake levels fall. Is this
going to crowd out demand for existing industry and for consumers?
Is well? Is that an issue when you look at
the evolution of the cloud industry that you've been really
close to for a long time, that is signals real
(18:08):
trouble ahead for it.
Speaker 4 (18:09):
Yeah, it is. And I mean, you know, like both
Microsoft Nata are spending a lot of time shouting from
the rooftops about all their green energy investments, They're not
shouting quite so high from the rooftops about the number
of nuclear plants that are being built near their data centers.
I mean, I think this is a broader ends in conversation.
And you know, when the whole t Y situation happened,
(18:31):
and you know whether or not they were going to
get a good deal on their energy. We need a
conversation about what our priorities are. We need a conversation
about resilience. I mean generally speaking, generation happens in the
South Island, the load is in the North Island. Note
that data Grid is building is hoping to build a
(18:53):
large scale data center close down to Manipuri, so that
will be will be great, will's actually low and generation
capacity close by. But the conversation for us as New
Zealanders is do we want aluminium smelting to happen here?
Do we want steal smelting to happen to happen here?
Do we want pop and paper mills? And it's way
(19:15):
beyond the topic of this conversation, but I think that
no politician is given ever going to ask us those
questions because they want to put forward a win win scenario.
The reality is there is no win win scenario, and
either we need more generation or we need to accept
the nuclear is going to happen, or build some dams,
(19:36):
or we're going to do sola or wind, or we're
not going to have energy to use, or we're not
going to have large scale industrial users of energy. And
these are all leavers that we can pull, but we
can't pull all the leavers at once and expect the
stuff to work. And I sort of bemoan the fact
in New Zealand that the discourse tends to be quite
immature in terms of understanding those trade off aspects.
Speaker 1 (19:58):
And look, you know, the likes of America and Contact
Energy contacted to deal with Microsoft to take geothermal capacity.
You know, they point out, look for us to invest
in additional capacity, we need sort of anchor tenance, just
like you need an anchor tenant in a data center
if you're leasing that space out, you need commitments from
(20:19):
a couple of big companies that they're going to take
and pay for over the coming decades your output. So
how many companies are putting up their hand to do that.
The reality is the growth is in the digital economy.
So in that sense, you know, it's only natural that
the data center companies are the ones that are sort
of anchoring this.
Speaker 4 (20:40):
Yeah, that's right in I mean, you know, like I
know as you do, Microsoft has actually built out so
and data center in terms of ata WS, I don't
actually understand, like I understand that when you're at that
scale you can actually buy directly from the gentailors, and
so you know it's coming from the grid, but if
you squint, it is it is sort of directly from
(21:01):
the generators in the a US case because they're COLO.
I'm not sure how that actually works. I mean, yes,
there will be more energy consumed because AS has a
region here, but I'm not sure you can then conflate
that to ADS having a direct connection or direct direct
commercial relationship with a generator, given that they are buying capacity. Eventually,
(21:27):
essentially they are leasing space in a pre existing data center.
So yeah, yeah, truth truth, truth, statistics and damn liners
and all that good stuff.
Speaker 1 (21:37):
Yeah. I put that question to them on Monday. Are
they going to use carbon credit? So how is this
going to work, particularly if there is real constraints on
renewable energy. Didn't get an answer on that, But look, Ben,
what are we going to get here? So presumably they've
already loaded up those data centers with Amazon equipment. Amazon
makes its own really high performance silicon chips. They'll have
(22:01):
probably Nvidia and other stuff maybe in coming to those
data centers. So there's a lot of equipment in there now.
They have a big roster of New Zealand clients, most
of them will be hosted at the moment in Sydney
or further afield. So do you think there will be
a big influx of applications and data into those data
(22:23):
centers in the coming months. I guess that that is
the good story where you start talking about lower latency
because you're closer to the data center. Know what are
some of the other benefits do you think that those
clients will will now take advantage of?
Speaker 3 (22:36):
Yeah?
Speaker 4 (22:36):
I mean, I guess there are three reasons why you
choose where to cite your cloud workload. Latency, data sovereignty,
and then cost and so you know, like I don't know,
I haven't looked into ad esus pricing in New Zealand
versus in Sydney. I would imagine it's the same. I'd
(22:59):
spect latency. You know, it's it's a long time to swim,
but it's a pretty short time for the speed of
light between Sydney and New Zealand. So for sure, some
high frequency trading workloads, potentially the latency thing is going
to be attractive, But to be honest, I discount that
(23:19):
a little bit. So really that the core proposition is
the data sovereignty one, and in a world where geopolitics
are becoming ever more fraught, that is even more compelling.
I mean, Australia is a very good friend of New Zealand.
But still, however, I would say, you know, I have
Don Christie ringing in my ear. Don Christy from the Catalyst,
(23:40):
and if he was here, he would say, as at
the end of the day is a US corporation. If
we are concerned about geopolitics, then jumping into bed with
the US corporation, notwithstanding that it has some presence here
in New Zealand, isn't a good thing. He would probably
back that up by saying, of that four hundred and
(24:02):
whatever million dollars with the revenue that AWS received here,
how much do they pay tax on? So all of
those questions are valid, I guess, and I have massive
amounts of respect for Don. I guess. My response to
him is the horse has already bolted. You know, the
hyperscalers Google, Microsoft, As and Ali Baba are quite just that.
(24:26):
They're hyperscalers. They are the ones to beat. Frankly, you
can't beat them. You're playing around the edges. And what
Ketalyst Cloud does is really really valuable, and there's a
strong value proposition. But it's a niche player, and I
think will only be a niche player. So it's a
great thing that AWS is here. There will be a
(24:47):
bunch of organizations that are quite excited by it. Frankly,
if I ran AWS, I would absolutely go down a
road of co location, because I'm not convinced that the
scale or of the workloads there are going to be
run in the New Zealand region as significant enough to
justify actually building the head infrastructure here. So I think
(25:09):
it's the right strategy. It's just been badly articulated.
Speaker 1 (25:12):
Yeah. Well, maybe they came to that conclusion sort of
as we came out of the boom of the pandemic
area where there was a lot of cheap cash around
and frankly a massive surge in the digital economy as
people moved online. Maybe the thinking was in twenty twenty one,
this is going to go exponential, where it's going to
(25:34):
get more and more business in this part of the world,
and it didn't turn out that way. So difficulties with
the site in Auckland as side. Something obviously changed along
the line.
Speaker 4 (25:44):
Yeah, And the thing is that if you're spending or
let's believe the number spending seven point five billion dollars.
Consenting issues are not a barrier. You can afford really
good lawyers. You can pay for drainage or whatever. So
I actually don't think that it was a you know,
an issue with the consenting regulatory framework in New Zealand.
(26:07):
I think frankly that said a lot. Actually the prize
isn't that big. Let's go co low and we can
always change if we need to.
Speaker 1 (26:14):
And just finally being you know, they're talking obviously you
can put your apps and shift all your data there now,
but they talked on Monday about the AI tools such
as the Bedrock platform. This is where you can choose
what large language model you want to use. Sage Maker
(26:35):
their AI sort of tool sets. So that stuff is coming.
They need dedicated hardware to run that. That will be
switched on, they said later this year. We're running out
a year, so that won't be far away. Is that
an exciting thing to you? I mean, that's what this
is all about, is not simply just putting your applications
in the cloud and the efficiencies you get from that,
(26:55):
but actually transforming your business using IF intelligence. Is that
genuinely exciting to you?
Speaker 2 (27:03):
Yeah?
Speaker 4 (27:03):
That's right. So so for some context there, without being
being too geeky, you know your general sort of computer
or storage basically taking the server that was in your
cupboard and moving that into the cloud. That's pretty generic
commodity hardware that is easy to come by that has
done at scale. When you're do an AI workloads, that's
(27:25):
done generally on what's called GPU's graphical processing units that
come from the likes of n Video. But as you
pointed out, a WS has its own trainingum silicon and
that stuff is a little bit harder. The supply chain
for that's a little bit longer, to be honest. The
other reason is that everyone uses computing storage, not everyone
is training large language models and needs GPUs. So I
(27:50):
think there's there's there's a couple of answers to that.
Generally speaking, training l l MS is something that isn't
particularly critical in terms of latency or even location. So
if I was running a SaaS company and latency was
a huge issue, and I was also doing some AI stuff,
(28:12):
I would probably strategically say putting data sovereignty aside for
a moment, I'm going to train my llms wherever it's
cheapest and easiest and most accessible, but the actual application
I'm going to run as close to the demand as possible.
So I guess the question that comes out of it is,
(28:32):
if you're doing AI type workloads, how concerned are you
about whether those happen here or elsewhere? Given that latency
generally isn't an issue. Again, it comes back to the
sovereignty thing. The other thing I would say, which is
really interesting, and it's going to be a bit of
a voyage on gequery, sorry, but there's a lot of
(28:54):
stuff happening around AI at the edge. So basically, you
chuck a bunch of data into a big hyperscale data
center in order to train a model, but then you
deploy that model at the edge so that you know,
let's say, for example, you're doing a connected camera for
home security, right, and you want to be able to
(29:16):
identify when someone comes, you know, different types of behavior
sentiment analysis, look at their face, see if they're happy
or sad or aggressive or whatever. Probably the approach is
you're going to train a bunch of models to understand happy, sad, aggressive, whatever,
and then you're going to get a really small, relatively
(29:38):
small model, and you're going to implement that on your camera,
so you don't need to send data back to some
data center wherever to say, oh, hey, something's happening, tell
me what it is. You'll do that at the edge.
So it's really interesting what's happening around distributed cloud distributed
data processing. And the interesting thing with the hyperscale is
(29:59):
is they are very much sort of predicated on a
quite a centralized model, but that's becoming a lot more granular,
and all of the hyper scalers are offering also sort
of models at the edge or you know, compute at
the edge or l l ms at the edge. But
it's really interesting in terms of this bifurcation around mass
(30:21):
processing at scale in a centralized manner and edge processing
where you need it, where there's zero latency and you're
not sending a bunch of data backwards and forward.
Speaker 1 (30:33):
But just finally, Ben, as you know, how big a
blow is this for AWS? I mean for the likes
of Kiwi Bank and those who are buying capacity in
that data center. You know, they'll probably just shrug and go, well,
that was a bad week for AWS, but they're still
providing great services to us. Do you think it damages
the company's credibility in any sustainable way.
Speaker 4 (30:54):
It's a really great question because you know, people like
you and I the tech sector are spending a huge,
huge amount of cycles sort of second guessing what as
did and didn't do and should have done. The reality
is they're very smart individuals. They've built a massive, massive
business and in a year's time isn't going to make
(31:15):
any difference to the growth they have seen or would
have seen in New Zealand. Absolutely not. So it's absolutely
a mistake and it's kind of an amateur screw up,
but it really really doesn't matter from the scheme of things,
Probably that the biggest damage, to be fair is political damage.
As I say, I think Prime Minister should have got
(31:36):
better advice, whereas advisors should have sought better advice. You know,
like again, that impact is not going to be you know,
the election isn't going to be one or lost on
a gas around an a OS launch. But yeah, I
don't think it's.
Speaker 1 (31:49):
Helped thanks to being keeps for his analysis of aws's
bumpy launch, He's right, a WS will happily roll on
picking up business for its co located data centers and
power to them. There's a lot of great innovation that's
going to be available right on our doorstep. But there's
also a lesson for big tech. People are concerned about
(32:11):
the data center boom, what it means for energy prices,
for data sovereignty, for our ability to steer our own destiny.
As a tiny handful of companies become central to everything
we do in the digital world, people have questions. Trust
is fragile, and it definitely took a dent this week.
(32:31):
Hope you enjoyed this special episode. Let me know what
you think about the AWS Data Center region launch back
to normal next week. I'll catch you next Thursday with
another episode of the Business of Tech.
I'm Peter Griffin and welcome to the Business of Tech
powered by Two Degrees, a special episode I've put together
to try and unpack a really big tech related story
that played out this week, the launch of Amazon Web
Services New Zealand Data Center Region four years in the making.
It should have been a big milestone to celebrate, but
(00:34):
it turned into a train wreck for Amazon, which really
shot itself in the foot by refusing to explain why
it had seemingly pivoted from building its own shiny, expensive
data centers to leasing space in existing ones and what
that meant for the seven point five billion dollar investment
over fifteen years it promised back in twenty twenty one,
(00:55):
and the thousand jobs it expected to create AWS, well,
that was just an option on the table. Never mind,
we've done something different. Journalists called bs on that the
media coverage was wall to wall brutal. It's perhaps best
summed up by Heather Dupless Allen's news Talk ZB interview
(01:16):
with AW West New Zealand country manager Manuel Bonnett on Tuesday.
Here's a clip from it.
Speaker 2 (01:23):
Have you built any data centers?
Speaker 3 (01:25):
So a region is a cluster of data centers, and
of course we have chosen the locations that we choose
to build out our availability zones. The region consists of
multiple availability zones. We have three of these availability zones
here and of course that involves data centers.
Speaker 2 (01:42):
Cool and how many of these have you actually built
as opposed to renting or leasing.
Speaker 3 (01:47):
So we are choosing facilities based on the needs that
we have, and we have thirty eight of these regions globally.
Speaker 2 (01:54):
Exactly what you're being worried about this though, aren't you, Manuel,
because you haven't built any have you?
Speaker 3 (01:58):
I'm telling you we chose the locations that are the
best looking.
Speaker 2 (02:01):
What it happened? Why didn't you guys build it? It
wasn't that part of the announcement with just in the
back in twenty twenty one that you were going to build.
Speaker 3 (02:06):
Some Well, we always take a long term view and
as part of that we look at all the options
and I'm super happy that we found the right locations
for the region to be live as of today.
Speaker 2 (02:17):
Okay, what happened to the west Aokland one where you
flattened it? You got rid of the wetland? Are you
going to build anything there?
Speaker 4 (02:23):
Well?
Speaker 3 (02:23):
I mean, for now we are announcing that the region
is live, so the availability zones at the manual.
Speaker 2 (02:29):
I just need to tell you, mate, listen. This audience
that listens to this is really smart and they can
hear you not answering the questions. So you can keep
on with that strategy if you want to. But I
need to warn you. We know you're not answering my question.
I'm going to try it again. What are you doing
with the West Aalkland site? Are you going to build
anything there?
Speaker 3 (02:45):
I can't comment on the site. I'm basically I'm announcing
the launch of the region and the site, as you
can tell, is not part of it.
Speaker 2 (02:52):
Okay, thanks mate, appreciate it. Go well. Manuel Bonnette, New
Zealand Country Manager, Amazon Web Services. Oh my gosh, guys,
how bummed to you about this announcement. I thought when
I heard luxon on with Hosking this morning, I thought
this was awesome.
Speaker 1 (03:05):
Amazon really didn't understand what they were getting themselves into
when they dreamt up the narrative behind this launch. The
fact that the Prime Minister then jumped on it for
a political capital just made it so much worse, which
is a shame because going live with a new hyperscale
data center region is a big development. AWS has made
(03:27):
a significant investment in cloud computing and is committing to
having a significant presence in New Zealand. That's a good thing.
So I thought i'd get a true cloud computing industry
expert on to unpack what went on here, or what's
probably gone on. It's hard to know exactly because everything
is shrouded in secrecy when it comes to AWS in
(03:51):
the name of security apparently. Ben Keeps is a christ
Church businessman, board director and cloud industry analyst who was
one of the earliest New Zealand proponents of this migration
to the cloud many businesses have been on. He knows
the AWS tech stack in depth, and he knows a
lot about the economics of hyperscale data centers. So he's
(04:13):
been with his candid take on what went down this week.
Ben Keeps, Welcome to the business of tech. How are
you doing.
Speaker 4 (04:20):
I'm doing well. Thanks for inviting me.
Speaker 1 (04:22):
Well, what a debarcle I mean this week with AWS
probably one of the worst PI disasters in the tech industry, definitely,
But in terms of any big corporation in recent memory
in New Zealand, what's your take on this? What really
went wrong? This week.
Speaker 4 (04:39):
Yeah, I mean, you're right, it was absolutely abysmal. There's
two stories here, right. The first story is one about investment,
enzting infrastructure politics frankly, and that was a total debarcle
and it was because of political expediency on the politician
side and frankly immaturity, surprising immaturity on the vendor side.
(05:06):
The second part of the story, which is actually impactful
and got completely lost, is that ADS is actually setting
up or has set up a region here in New Zealand.
And that answers a bunch of questions that people have
had for going on twenty years now.
Speaker 1 (05:22):
So let's unpack these things. Let's go all the way
back actually to twenty twenty one, the day after AWS
made that announcement, the initial announcement back in the depths
of COVID that they were setting up a data center
region in New Zealand. This is what you wrote in
the New Zealand Herald. The devil is in the detail
and apart from the big number seven point five billion
(05:44):
investment and one thousand jobs, there's a little detail in
the announcement, but details aside. This is exciting and further
puts New Zealand on the map as a credible technology nation.
That was a full four years ago. And what's striking
about that is those same figures are exactly what was announced,
you know, on Monday at Amazon's Cloud Day. And I
(06:05):
think that was the root of the problem, was that
there was no advancement on this or no transparency into
the numbers. How the data centers, because there's three of
them in thew Auckland region are configured. Whether you know
Amazon is building its own, we know it's it's not
at the moment, it's it's co locating. Do you think
(06:26):
that's where the root of the problem started here before
the politics came into it. Just that lack of transparency,
which key with is you know, straight up people that's
sort of what they expect in dealing with companies.
Speaker 4 (06:38):
Yeah, so seven point five billion back in twenty twenty
one was a terrible number because power, you know, buying
smoke over the construction crew, you know, the driver that
drives the you know, the aws CEO to cut a ribbon.
I mean, it's ridiculous. It was a ridiculous number then
and it's an even more ridiculous number now. However, the
(07:00):
thing that's really stark is that if AS had in
fact built three of its own data centers, the investment
would have been would have been massive. I don't know,
you know, it wouldn't have been seven point five billion,
but it would have been massive. The investment now is
significantly smaller because for all very good reasons, they are
(07:20):
embarking on a collocation strategy, which means that they're using
data centers that already exist. Obviously higher levels of capacity
within those data centers, but putting in a few extra servers,
or even lots and lots of extra servers, is a
hell of a lot cheaper than building an entire new
data center. So frankly, and the thing that really disappoints
(07:41):
me about the AOS launch, their part of it is
that to put out a statement that said seven point
five billion dollars a thousand jobs, when everyone knows the
data center is still just a hole in the ground
or not even and they are doing COLO is not
only is it duplicitus, but it actually doesn't recognize the
(08:03):
fact that we're not stupid, you know. And I've given
that feedback to my people with an AS. It was
a terrible launch. They didn't do themselves any fabors and
the really great thing about a region got lost because
they put out the release and obviously, you know, the
government decided this is a really good news story and
(08:25):
jumped onto it without being well briefed upon what it
actually meant.
Speaker 1 (08:29):
Yeah, it is. There is a good story buried in there,
and we'll get to that. But just a bit more
under seven point five billion. I mean, you just have
to look at the AWS financial accounts over the last
few years. You know, typically a tech company, you know,
big tech, they usually spend about twelve and a half
percent off their sales, their revenue on capital expenditure. You know,
(08:52):
AWS did last year in New Zealand four hundred and
twenty five million dollars in revenue, really healthy business. But
when you look at their cap X, you know, purchases
of property and equipment, it's all in the you know,
sixteen million in twenty twenty one, eleven million in twenty
twenty two, eight point four million, and twenty twenty three,
it's clear that they're just not spending the big dollars
(09:15):
that they promised, which would have been to the tune
of five hundred million dollars a year. If you're looking
at seven point five billion over fifteen years. So they
they may have had these aspirations at one point, but
when those builds were off the table and they moved
to leasing arrangements, that probably cut the investment by half
(09:36):
at least.
Speaker 4 (09:37):
Well well even beyond that. I mean, even if you
look at the original number with the building out data
center a strategy, So four hundred million revenue a year,
let's say you know they double that, triple that, quadruple that,
you know, a couple of billion dollars revenue a year.
You don't really spend seven point five billion dollars, you know,
building out on CAPEX, building that out in a very
(10:00):
finite market. I meant the end of the day, New
Zealand is always a small market for everything. And so
you know, that's why I said in twenty two one,
the devil is in the detail. I didn't believe that
the market opportunity here justified seven point five billion dollars.
The other, the other aspect that a few commentators have
sort of talked upon is that, you know, it was
(10:22):
it was touted as sort of corporate philanthropy and magnanimity
and how amazing as is investing in New Zealand and frankly,
you know, go capitalism there. They're a for profit organization.
They invest to make a return and so therefore there's
no philanthropy in there whatsoever. They look at a capex
(10:43):
and the payback on that. It's it's about net present
value and discounted cash flows and those sorts of things.
And so you know, the I mean the stuff they're
doing around training, you know, that's that's self self interested.
But it's great, it's awesome. It's a good thing for
NTID inc. Again, like tying all the stuff together, seven
point five billion dollar investment in New Zealand, training, one
(11:05):
hundred thousand people doing all this stuff, and we're so great.
I mean, just New Zealanders aren't wired that way. And
you know, I mean frankly, you know, like I know
the AS leadership here in Apja and in Seattle, and
I don't think they've had advice from you know, internally
(11:25):
or from their external PR people around the New Zealand
psyche and what moves the needle. And unfortunately they barked
on embarked on a PR message that went down frankly
like a cup cold sick with Kiwi's and they're paying
the price. For it now. Yeah.
Speaker 1 (11:41):
Look, I had an off the record briefing with AWS
just last week, which you know I can't talk about.
I agreed to keep it off the record, but I
will say I came out of that meeting thinking just
trouble ahead here because a lot of the questions that
I've had over the years and other journalists have had,
they still weren't answering. And to go out and not
(12:03):
have credible answers to these questions, it was going to
go badly.
Speaker 4 (12:07):
The thing is there, right, So I understand that from
a corporate strategy perspective, they don't want to tell you
the truth and the whole story. I get that, but
then don't tell a story. And so if your thing
is we're secretive, we're not going to talk about this,
then then you know, when you're briefing your PR people,
you say, right, our strategy, our corporate policy is we're
(12:29):
not going to disclose this, this, and this. Then the
advice should be okay. Given that lead with we're launching
the New Zealand region. This is an amazing thing, as
we'll talk about later on. Don't say anything else. Don't
say about investment, don't say about all of those other things,
and honestly, the story would have been so positive because
(12:49):
it's a really good thing. But whether it's arrogant, whether
it's ignorant. For some reason, they felt they had to
embark on this common strategy and it's blown up in
their face.
Speaker 1 (13:01):
And then on top of that, we had the political
angle we had lux and come in and look to
be fair to the Prime minister. I don't think he
actually tried to explicitly claim credit for this. What he
did is his typical rah, rah, isn't this great? You
know anything at the moment he's trying to jump on
that shows growth of any description, and he hyped that up.
(13:25):
He was probably given all the key numbers from AWS
and his own team, which were exactly the key numbers
that justinto Ardern was talking about, you know, four years ago,
and he did his typical sort of you know, hype
machine shtick, and but it just it just politicized it
at a time when he's under pressure and there isn't
(13:47):
much good news. That really didn't help, did it, well.
Speaker 4 (13:50):
I mean it didn't. And you know, like I've talked
to Chris a few times in the past. He has
my phone number, you know, like it was obvious what
was going to happen. All the editors were gonna say, like, hey,
what's the way to position this? I want to make
some political capital, I want to get some wins. How
do I position this? And there the story is, Wow,
this is a great thing. We are such a powerhouse
(14:10):
of technology in New Zealand that AWS is investing and
is building a region here, and that's such a win.
Technology is such an opportunity for us. Success stories like
zero and vend or whatever are now going to find
it easier to get ahead because of this, you know,
this regional buildout, and no one could argue with that
(14:30):
because that's actually the truth. But instead, you know, like
I don't think I mean, I don't know like the
pundits within the National Party when they're doing their polling,
do they see this as neutral or or a loss
because I don't think it's a political win.
Speaker 1 (14:47):
It's definitely not. And look, you know the issue about
not building the data centers, right, it's disappointing. We want
to see more infrastructure come in, but the reality is
is there three data centers that a lot of Amazon
equipment will be going into, if not already in there.
This is not unusual around the world. I mean, there's
(15:08):
a big operator in Asia, Pacific air Trunk, huge data
center operator. It works with the big three Aws, Microsoft,
and Google. So they don't always build their own. They
do prefer to because they have complete control over the facilities.
But colocation is a legitimate thing in the data center industry.
They sort of blew that as well by not saying that, hey,
(15:31):
look we tried to build, it didn't work out, we're
co locating.
Speaker 2 (15:35):
Yeah.
Speaker 4 (15:35):
The difficulty here is that, and I do sort of
feel freight os here is that the hyperscalars are very
sensitive about the environmental impact of hyper scale data centers,
and so they've invested hugely in green energy, so green
energy in terms of input, but also low power servers
(15:55):
and really interesting ip around calling of the service because
obviously calling is a huge and when you go COLO,
you haven't got that same control. And so all of
a sudden, this strong story about you know, adel mess,
you know, green environmental credentials goes out the window. And
so it's a really tricky one. I mean, you know
the fact that they've done forward deals for green energy,
(16:18):
which as almost everyone in New Zealand knows actually is
irrelevant because you're taking from the grid and when the
grid hasn't got enough, you burn coal huntly. So again,
like you know, I don't know what the comms and
PR team within the organization are paid, but I can
imagine imagine it's quite a lot to not simply say,
(16:38):
you know, do a chat GBT, hey tell me about
the New Zealand energy sector and what happens when there's
not much power, you know, like it's not rocket science. Yeah,
they've tried. I mean, I think it's probably a situation
where you've got a big multinational is trying to have
a somewhat consistent approach in terms of regional business development,
(17:00):
and we see we saw the perils with that.
Speaker 1 (17:03):
I guess the way AWS and Microsoft, to a lesser extent,
they have a little bit more autonomy in each country,
but Google and Aws, Apple, it's the company line everywhere.
So trying to have some nuance for the New Zealand
market isn't really going to cut it when you're talking
to Singapore or Seattle. So that is a problem. But
the energy thing did get a lot of play this week,
(17:25):
and I guess the fundamental issue is Kiwis are concerned
that they've heard a lot about the AI applications in
data centers. Data center is being built specifically for AI.
Zuckerberg is building one in Louisiana, fifty billion dollar data center.
The energy needs for that are massive. So they're looking
around the world and going is that going to happen
(17:47):
to us? We're already seeing mills shut down and potential
brownouts when you know the lake levels fall. Is this
going to crowd out demand for existing industry and for consumers?
Is well? Is that an issue when you look at
the evolution of the cloud industry that you've been really
close to for a long time, that is signals real
(18:08):
trouble ahead for it.
Speaker 4 (18:09):
Yeah, it is. And I mean, you know, like both
Microsoft Nata are spending a lot of time shouting from
the rooftops about all their green energy investments, They're not
shouting quite so high from the rooftops about the number
of nuclear plants that are being built near their data centers.
I mean, I think this is a broader ends in conversation.
And you know, when the whole t Y situation happened,
(18:31):
and you know whether or not they were going to
get a good deal on their energy. We need a
conversation about what our priorities are. We need a conversation
about resilience. I mean generally speaking, generation happens in the
South Island, the load is in the North Island. Note
that data Grid is building is hoping to build a
(18:53):
large scale data center close down to Manipuri, so that
will be will be great, will's actually low and generation
capacity close by. But the conversation for us as New
Zealanders is do we want aluminium smelting to happen here?
Do we want steal smelting to happen to happen here?
Do we want pop and paper mills? And it's way
(19:15):
beyond the topic of this conversation, but I think that
no politician is given ever going to ask us those
questions because they want to put forward a win win scenario.
The reality is there is no win win scenario, and
either we need more generation or we need to accept
the nuclear is going to happen, or build some dams,
(19:36):
or we're going to do sola or wind, or we're
not going to have energy to use, or we're not
going to have large scale industrial users of energy. And
these are all leavers that we can pull, but we
can't pull all the leavers at once and expect the
stuff to work. And I sort of bemoan the fact
in New Zealand that the discourse tends to be quite
immature in terms of understanding those trade off aspects.
Speaker 1 (19:58):
And look, you know, the likes of America and Contact
Energy contacted to deal with Microsoft to take geothermal capacity.
You know, they point out, look for us to invest
in additional capacity, we need sort of anchor tenance, just
like you need an anchor tenant in a data center
if you're leasing that space out, you need commitments from
(20:19):
a couple of big companies that they're going to take
and pay for over the coming decades your output. So
how many companies are putting up their hand to do that.
The reality is the growth is in the digital economy.
So in that sense, you know, it's only natural that
the data center companies are the ones that are sort
of anchoring this.
Speaker 4 (20:40):
Yeah, that's right in I mean, you know, like I
know as you do, Microsoft has actually built out so
and data center in terms of ata WS, I don't
actually understand, like I understand that when you're at that
scale you can actually buy directly from the gentailors, and
so you know it's coming from the grid, but if
you squint, it is it is sort of directly from
(21:01):
the generators in the a US case because they're COLO.
I'm not sure how that actually works. I mean, yes,
there will be more energy consumed because AS has a
region here, but I'm not sure you can then conflate
that to ADS having a direct connection or direct direct
commercial relationship with a generator, given that they are buying capacity. Eventually,
(21:27):
essentially they are leasing space in a pre existing data center.
So yeah, yeah, truth truth, truth, statistics and damn liners
and all that good stuff.
Speaker 1 (21:37):
Yeah. I put that question to them on Monday. Are
they going to use carbon credit? So how is this
going to work, particularly if there is real constraints on
renewable energy. Didn't get an answer on that, But look, Ben,
what are we going to get here? So presumably they've
already loaded up those data centers with Amazon equipment. Amazon
makes its own really high performance silicon chips. They'll have
(22:01):
probably Nvidia and other stuff maybe in coming to those
data centers. So there's a lot of equipment in there now.
They have a big roster of New Zealand clients, most
of them will be hosted at the moment in Sydney
or further afield. So do you think there will be
a big influx of applications and data into those data
(22:23):
centers in the coming months. I guess that that is
the good story where you start talking about lower latency
because you're closer to the data center. Know what are
some of the other benefits do you think that those
clients will will now take advantage of?
Speaker 3 (22:36):
Yeah?
Speaker 4 (22:36):
I mean, I guess there are three reasons why you
choose where to cite your cloud workload. Latency, data sovereignty,
and then cost and so you know, like I don't know,
I haven't looked into ad esus pricing in New Zealand
versus in Sydney. I would imagine it's the same. I'd
(22:59):
spect latency. You know, it's it's a long time to swim,
but it's a pretty short time for the speed of
light between Sydney and New Zealand. So for sure, some
high frequency trading workloads, potentially the latency thing is going
to be attractive, But to be honest, I discount that
(23:19):
a little bit. So really that the core proposition is
the data sovereignty one, and in a world where geopolitics
are becoming ever more fraught, that is even more compelling.
I mean, Australia is a very good friend of New Zealand.
But still, however, I would say, you know, I have
Don Christie ringing in my ear. Don Christy from the Catalyst,
(23:40):
and if he was here, he would say, as at
the end of the day is a US corporation. If
we are concerned about geopolitics, then jumping into bed with
the US corporation, notwithstanding that it has some presence here
in New Zealand, isn't a good thing. He would probably
back that up by saying, of that four hundred and
(24:02):
whatever million dollars with the revenue that AWS received here,
how much do they pay tax on? So all of
those questions are valid, I guess, and I have massive
amounts of respect for Don. I guess. My response to
him is the horse has already bolted. You know, the
hyperscalers Google, Microsoft, As and Ali Baba are quite just that.
(24:26):
They're hyperscalers. They are the ones to beat. Frankly, you
can't beat them. You're playing around the edges. And what
Ketalyst Cloud does is really really valuable, and there's a
strong value proposition. But it's a niche player, and I
think will only be a niche player. So it's a
great thing that AWS is here. There will be a
(24:47):
bunch of organizations that are quite excited by it. Frankly,
if I ran AWS, I would absolutely go down a
road of co location, because I'm not convinced that the
scale or of the workloads there are going to be
run in the New Zealand region as significant enough to
justify actually building the head infrastructure here. So I think
(25:09):
it's the right strategy. It's just been badly articulated.
Speaker 1 (25:12):
Yeah. Well, maybe they came to that conclusion sort of
as we came out of the boom of the pandemic
area where there was a lot of cheap cash around
and frankly a massive surge in the digital economy as
people moved online. Maybe the thinking was in twenty twenty one,
this is going to go exponential, where it's going to
(25:34):
get more and more business in this part of the world,
and it didn't turn out that way. So difficulties with
the site in Auckland as side. Something obviously changed along
the line.
Speaker 4 (25:44):
Yeah, And the thing is that if you're spending or
let's believe the number spending seven point five billion dollars.
Consenting issues are not a barrier. You can afford really
good lawyers. You can pay for drainage or whatever. So
I actually don't think that it was a you know,
an issue with the consenting regulatory framework in New Zealand.
(26:07):
I think frankly that said a lot. Actually the prize
isn't that big. Let's go co low and we can
always change if we need to.
Speaker 1 (26:14):
And just finally being you know, they're talking obviously you
can put your apps and shift all your data there now,
but they talked on Monday about the AI tools such
as the Bedrock platform. This is where you can choose
what large language model you want to use. Sage Maker
(26:35):
their AI sort of tool sets. So that stuff is coming.
They need dedicated hardware to run that. That will be
switched on, they said later this year. We're running out
a year, so that won't be far away. Is that
an exciting thing to you? I mean, that's what this
is all about, is not simply just putting your applications
in the cloud and the efficiencies you get from that,
(26:55):
but actually transforming your business using IF intelligence. Is that
genuinely exciting to you?
Speaker 2 (27:03):
Yeah?
Speaker 4 (27:03):
That's right. So so for some context there, without being
being too geeky, you know your general sort of computer
or storage basically taking the server that was in your
cupboard and moving that into the cloud. That's pretty generic
commodity hardware that is easy to come by that has
done at scale. When you're do an AI workloads, that's
(27:25):
done generally on what's called GPU's graphical processing units that
come from the likes of n Video. But as you
pointed out, a WS has its own trainingum silicon and
that stuff is a little bit harder. The supply chain
for that's a little bit longer, to be honest. The
other reason is that everyone uses computing storage, not everyone
is training large language models and needs GPUs. So I
(27:50):
think there's there's there's a couple of answers to that.
Generally speaking, training l l MS is something that isn't
particularly critical in terms of latency or even location. So
if I was running a SaaS company and latency was
a huge issue, and I was also doing some AI stuff,
(28:12):
I would probably strategically say putting data sovereignty aside for
a moment, I'm going to train my llms wherever it's
cheapest and easiest and most accessible, but the actual application
I'm going to run as close to the demand as possible.
So I guess the question that comes out of it is,
(28:32):
if you're doing AI type workloads, how concerned are you
about whether those happen here or elsewhere? Given that latency
generally isn't an issue. Again, it comes back to the
sovereignty thing. The other thing I would say, which is
really interesting, and it's going to be a bit of
a voyage on gequery, sorry, but there's a lot of
(28:54):
stuff happening around AI at the edge. So basically, you
chuck a bunch of data into a big hyperscale data
center in order to train a model, but then you
deploy that model at the edge so that you know,
let's say, for example, you're doing a connected camera for
home security, right, and you want to be able to
(29:16):
identify when someone comes, you know, different types of behavior
sentiment analysis, look at their face, see if they're happy
or sad or aggressive or whatever. Probably the approach is
you're going to train a bunch of models to understand happy, sad, aggressive, whatever,
and then you're going to get a really small, relatively
(29:38):
small model, and you're going to implement that on your camera,
so you don't need to send data back to some
data center wherever to say, oh, hey, something's happening, tell
me what it is. You'll do that at the edge.
So it's really interesting what's happening around distributed cloud distributed
data processing. And the interesting thing with the hyperscale is
(29:59):
is they are very much sort of predicated on a
quite a centralized model, but that's becoming a lot more granular,
and all of the hyper scalers are offering also sort
of models at the edge or you know, compute at
the edge or l l ms at the edge. But
it's really interesting in terms of this bifurcation around mass
(30:21):
processing at scale in a centralized manner and edge processing
where you need it, where there's zero latency and you're
not sending a bunch of data backwards and forward.
Speaker 1 (30:33):
But just finally, Ben, as you know, how big a
blow is this for AWS? I mean for the likes
of Kiwi Bank and those who are buying capacity in
that data center. You know, they'll probably just shrug and go, well,
that was a bad week for AWS, but they're still
providing great services to us. Do you think it damages
the company's credibility in any sustainable way.
Speaker 4 (30:54):
It's a really great question because you know, people like
you and I the tech sector are spending a huge,
huge amount of cycles sort of second guessing what as
did and didn't do and should have done. The reality
is they're very smart individuals. They've built a massive, massive
business and in a year's time isn't going to make
(31:15):
any difference to the growth they have seen or would
have seen in New Zealand. Absolutely not. So it's absolutely
a mistake and it's kind of an amateur screw up,
but it really really doesn't matter from the scheme of things,
Probably that the biggest damage, to be fair is political damage.
As I say, I think Prime Minister should have got
(31:36):
better advice, whereas advisors should have sought better advice. You know,
like again, that impact is not going to be you know,
the election isn't going to be one or lost on
a gas around an a OS launch. But yeah, I
don't think it's.
Speaker 1 (31:49):
Helped thanks to being keeps for his analysis of aws's
bumpy launch, He's right, a WS will happily roll on
picking up business for its co located data centers and
power to them. There's a lot of great innovation that's
going to be available right on our doorstep. But there's
also a lesson for big tech. People are concerned about
(32:11):
the data center boom, what it means for energy prices,
for data sovereignty, for our ability to steer our own destiny.
As a tiny handful of companies become central to everything
we do in the digital world, people have questions. Trust
is fragile, and it definitely took a dent this week.
(32:31):
Hope you enjoyed this special episode. Let me know what
you think about the AWS Data Center region launch back
to normal next week. I'll catch you next Thursday with
another episode of the Business of Tech.
Popular Podcasts
NFL Daily with Gregg Rosenthal
Gregg Rosenthal and a rotating crew of elite NFL Media co-hosts, including Patrick Claybon, Colleen Wolfe, Steve Wyche, Nick Shook and Jourdan Rodrigue of The Athletic get you caught up daily on all the NFL news and analysis you need to be smarter and funnier than your friends.
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com