web analytics

Bill Bennett

Menu

Unpicking AWS’ New Zealand numbers

Amazon Web Services says it will build a Region in New Zealand. AWS plans to start operating local cloud data centres by 2024.

An official press release telling us about this runs to more than 3200 words. It’s long, wordy, poorly written and hard to understand.

Despite the torrent of words, it is light on details.

Mystery

It doesn’t answer key questions.

Such as: “Where will Amazon locate its promised data centres?”

That’s not mentioned anywhere in the 3200 plus words.

It’s an important question. New Zealand is on a fault line. That has implications.

Our biggest city, Auckland, is closest to the submarine cables piping in data from around the world. That’s handy. Yet because it sits on an active volcanic field it may not be the best site for a $7.5 billion investment.

Auckland has a large workforce, although the construction skills needed to build a data centre are in short supply. Does Amazon propose to bring that expertise in?

Electricity

There’s an undersupply of electricity in the Auckland region. In winter Huntly’s dirty old coal-based power plant kicks in to top up the power supply.

Will AWS be using that power?

Meanwhile, at the other end of the country there is, or soon will be, an oversupply of low cost, clean, green power. Is that in the plan? It should be.

New Zealand’s south has a cool climate. Cooling is a major operational cost for data centres. Locating in a cooler climate not only reduces costs, it smartens up green credentials, an important part of any cloud marketing programme.

Data transit

If Amazon locates in or around Auckland, it won’t have to pay much in the way of transit charges for data travelling up and down the nation’s backbone.

Nor will it pay for the cost of bringing in the power needed to run a hyper-scale data centre.

New Zealanders subsidise moving power around the country. There is a limit to how much of that is possible without network upgrades. Who pays for that?

AWS’ numbers look as they have been polished up for maximum press release impact.

From the press release:

AWS released an economic impact study (EIS) that estimates it will create 1,000 new jobs through investment of NZ$7.5 billion (US$5.3 billion) in the new AWS Asia Pacific (Auckland) Region with an estimated economic impact on New Zealand’s GDP of NZ$10.8 billion (US$7.7 billion) over the next 15 years.

Jobs for who?

We can assume many of the 1000 jobs will be temporary roles for people building the data centres. It’s rare for a giant data centre to employ more than a handful of people and that includes security guards.

Once the ball is rolling there won’t be many data centre jobs. There could be development work piggybacking off the data centre. AWS doesn’t say. Nor does it say where the people to fill those roles will come from.

Amazon doesn’t tell us enough about its plans for any sensible analysis of its $7.5 billion investment claim.

We know AWS is a worldwide hyper scale cloud business. Any comparison with Spark is always going to look odd. Yet, Spark operates New Zealand’s largest existing data centre at Takanini. That makes it the closest we have to a benchmark.

Spark’s original project cost $60 million. Subsequent expansion means that price will be a lot higher. It won’t be anywhere near $7.5 billion. It won’t be one tenth of that amount.

Market share

IDC Research says Spark and Datacom have a 43 percent share of New Zealand’s Infrastructure as a Service market. Amazon has a 23 percent market share. Microsoft is a touch behind on 19 percent. In other words, the four are all roughly the same size.

As a rule, the large, hyper scale cloud providers AWS and Microsoft Azure take 70 to 80 percent market share in any territory.

Their lateness to build here is one reason they have less share at the moment. The lure of winning that additional market share could be part of the reason both Microsoft and AWS have made big New Zealand cloud announcements in recent months.

Sure, there is more to cloud than IaaS. Yes, the market is expanding fast. Yes, Amazon is hyper scale. Does it need to spend 20 or 50 times as much as Spark?

Datagrid

Datagrid is a $700 million project to build what may still be New Zealand’s first hyper-scale data centre. The business is headed by Remi Galasso and Callplus founder Malcolm Dick.

The pair know a thing or two: they got the Hawaiki cable off the ground.

Datagrid has chosen a Southland site.

Interestingly, Amazon has invested in Hawaiki. It’s likely the Datagrid team has talked to AWS about potential cooperation1.

Part of Datagrid’s plan is to build a new submarine cable connecting Invercargill to Australia. There will also be a domestic submarine cable linking the site to major New Zealand cities.

Will Amazon build a similar cable to distribute its data?

Scale

At first Datagrid will be a 60 megawatt, 25,000 square metre data centre. Over time it will grow to 100 megawatts and 40,000 square metres. That’s a lot of data centre.

Assuming data centre costs scale better than linearly and Amazon can call on its worldwide economies of scale, its project will build more than ten times Datagrid’s capacity.

Which brings us to another question. How big is New Zealand’s cloud market?

In March Spark told Computerworld’s Sarah Putt it estimates total cloud revenue at around $730 million.

That figure doesn’t square with AWS’s $7.5 billion build budget over 15 years unless AWS anticipates the market continuing to grow at a fast rate.

Assuming AWS doesn’t capture the entire market and intends its New Zealand operation to be profitable, either the local market would need to grow at about 40 percent a year for the next decade or the company expects to host a huge amount of international business here.

Unknowns

There are plenty of unknowns. Too many unknowns to make a careful analysis of AWS’s plans. Yet there are four possible conclusions one could make about the $7.5 billion announcement.

The first is the most cynical: that it is pure public relations hype.

Mentioning a big enough number and promising lots of jobs is a sure fire way of seeing off any resistance and buying-in political good will. AWS can rest assured no-one is going to look back in 15 years to check it spent $7.5 billion.

Big technology companies like AWS have plenty of form when it comes to talking things up.

Efficient

A second conclusion, is that Amazon throws money around like water and is hugely inefficient. It overpays for everything.

This is implausible. It doesn’t square with anything we know about Amazon which is famous for trimming costs to the bone. Operational efficiency is key to making money from the cloud.

The third possibility is that AWS expects to scythe through the local cloud market. It has done this before. It’s possible, but wise cloud customers are wary of dealing with a single international ecosystem. Many will seek alternative service providers as a back-up.

That’s going to limit AWS’ potential market share. And even 100 percent share of a $730 million market doesn’t justify spending $7.5 billion even with heroic growth rates.

A more likely story is that AWS has bigger plans for New Zealand that serving local markets. It has hinted at this without explicitly saying anything. New Zealand gives, say, Australian AWS users a viable alternative location with, if not always similar, at least readily understood local conditions.

One last point. Until now, the big global cloud companies stayed away from New Zealand. They didn’t like it when there was only one submarine cable network. They didn’t like what they saw as a hostile and monopolistic telecommunications market. It took ten years of industry reform.

As far as technology is concerned, no-one thinks of us as a smug hermit kingdom.


  1. A speculating person might wonder if Datagrid will become part of AWS. ↩︎

«

»


6 thoughts on “Unpicking AWS’ New Zealand numbers

  1. The $7.5b numbers make no sense. The datagrid project sounds more feasible and given time and scale that seems like a more useful way to calibrate the size of AWS project. I read somewhere that NZ needs more capacity on the cable for moving electricity north ( also more capacity for data ) up the country too. The smelter wind down has to be an obvious source of cheap electricity but only if some new cables are added to the mix. Maybe that accounts for some serious inflation of the numbers.

  2. True. But the cost of upgrading power links South to North will, ultimately, fall on all power customers. AWS will probably get a free or a subsidised ride on power if it locates in Auckland. If it locates near the cheaper power, it will either need to build its own cable or spend a lot on buying connectivity.

  3. Thanks for this analysis Bill. I wonder if our breathless government and business leaders have considered the implications of the US Cloud Act or Patriot Acts prior to pronouncing this development (and Microsoft’s similarly sensational announcement of its intentions to build an ‘Azure’ zone in NZ) ‘great for NZ’… I suppose it creates a bit more onshore competition… but $trillion dollar corporations have deep pockets to loss-lead and kill off domestic competition… (see the Bunnings business model, for example). I’m distressed by the degree to which pundits (and our PM!) have largely ignored the fact that we’ve had viable clouds in NZ for well over a decade… Our government has snubbed those local providers with on-shore cloud infrastructure at almost every opportunity, instead cavalierly committing government data (and data about NZers) to offshore clouds owned by these two US-based foreign providers, using very flimsy justifications. I wonder why. It’s not a cost thing (the domestic clouds are cost-competitive). It will be interesting to see how Amazon plays this – and a) where they will go (Auckland, or where the power is?), b)) whether any US money will actually appear in NZ, c) how much (we all need to keep track to see if they’re just blowing smoke right now!), and d) how much tax they’ll pay in NZ.

  4. Interesting post re the Prod Com… Yes, I suspect you’re right. Which is why I say we deserve better gov’t tech policy and decision makers. The ones we’ve got are clearly unfit for purpose.

Comments are closed.

%d bloggers like this: