Amazon's Cloud Is One of the Fastest-Growing Software Businesses in History

http://www.businessweek.com/articles/2014-07-15/amazons-cloud-is-one-of-the-fastest-growing-software-businesses-in-history#r=read

by adventured • 12 years ago

142 30 12 years ago

businessweek.com

beachstartup• 12 years ago

a couple of points about aws:

1. it's not really a software business. their marginal cost of products sold is nowhere near that of software. the software they develop facilitiates the sales/leasing of hardware and networks.

2. it's a division of a much larger parent company that poured billions into it. it's a "business" but not a company, which is what they're comparing it against.

dba7dba• 12 years ago

I agree with the department versus a business. No other startup could've done what they did, providing the tech knowhow, logistics (always a big issue at that scale), and the brand.

EVERYONE willingly gives their credit card info to Amazon.

dictum• 12 years ago

Amazon's cloud services are a software business in the same way a utility company that wrote software to automate some management tasks is a software business.

teej• 12 years ago

I had to call a human being and use a fax machine to get my power turned on with the local utility. I can spin up an EC2 cluster with an API call.

opendais2• 12 years ago

Heh. I signed up on a website and put in a requests to turn my utilities on.

Mandelbug• 12 years ago

Have you actually used and looked into AWS? There are dozens of services that do a lot more than just what people generally think of as "the cloud".

flyt• 12 years ago

One of those rare examples where the best product is winning in an entire category.

meddlepal• 12 years ago

I don't know if they're the best at everything. I've been using Google Compute Engine extensively lately and I think it is a lot more polished feeling at times than EC2.

tszming2• 12 years ago

I am interested to know which part of the GCE is a lot more polished than EC2.

padobson• 12 years ago

I'm not surprised at all. When I design a web app, I'll design it around AWS infrastructure so that they're the "Ops" in my "DevOps" title. They handle everything about web app development that I don't care to.

tluyben2• 12 years ago

I wanted to come and type that up here; when we do apps, web, scaling, cdn, file storage, trivial SSL etc we just prep for AWS. All the scripts are there as most people ask for it and it always works. With their solid offerings and constant price drops I would see no reason to ever move. It's all the stuff so painful once and now is all such a warm bath.

yry4345• 12 years ago

"I would see no reason to ever move."

(Obligatory) Apart from the single political point of control. AWS is to web apps what GoDaddy is to DNS and Verisign is to SSL.

ChuckMcM3• 12 years ago

It isn't particularly surprising, but it would be much more interesting to look at margins vs revenue. What is the margin on that $5B vs the $3.1B? I'm guessing that given the recent price cuts in between Amazon, Google, Softlayer, and others that the margins are starting to get squeezed down a bit.

I find that particularly interesting when looking at the relative success of Docker. The reason being that IBM's Z1 mainframe seems to have better operational economics near full utilization than rack and stack servers do. You can see how that works by looking at where the OpenCompute stuff takes out cost, no more sheet metal around servers, shares power supplies etc.

Makes me think that 'blade' computing should have been 'building sized' from the beginning :-)

opendais2• 12 years ago

Aren't those like $75,000?

I'm curious what you are basing the operational economics statement on.

alttab• 12 years ago

Probably the lifetime of the mainframe versus the long running cost of hosting it on AWS? 8 years, you know what you need it for, might be worth it?

Not if you don't have very large POs in hand, however. This includes most start ups.

ChuckMcM3• 12 years ago

I currently have about a thousand servers that do Blekko's web crawl in a colo facility, they are pretty modernish (westmere vs Ivy bridge but similar configuration), the operation economics for running those is a couple million a year fully loaded. My particular application is a 'bad fit' for services like AWS due to its large storage foot print and the fact that it constantly churns through petabytes of data but were we to host web servers w/ Docker or the equivalent we could probably provide something like a 100K web hosts. Doing that $5 'droplet' style with Digital ocean would be say a half a million a month so pretty good margins (caveat all the things like racks and drive replacement and floor space) but when those margins got squeezed something like an equivalently sized Z10 could shave operational costs from both a operator perspective and a power/cooling perspective. (total watts per delivered core-gigabyte-megabit of compute delivered to the Internet)

At web scale, costs shift in interesting ways, operator time is much more valuable than say the cost of hardware power, and cooling overwhelm floorspace costs etc. Luiz Barrosa is an engineer who gave a talk titled Data Center sized computers which alluded to some of those shifts, Urs and Luiz updated it a couple of times.

As I've gone from NetApp to Google and now to Blekko it has been interesting to me to watch how the multi-core, large memory, SSD changes have pushed around system design like this. And I was thinking that back in the day the "big problem" with blade computing was that using 20 blade chassis an install with 20 blades was cost effective but one with 21 blades was not. The virtualization wave made that moot since you only install full racks/chassis as demand calls for it, and current container wave adds to that single OS type installs for even better cost savings. Fun times.

wmf• 12 years ago

I think you're missing a zero or so.

deeths• 12 years ago

AWS is doing some great stuff, but this is a really misleading article.

Ignoring the obvious issue that AWS is a service and not a software business (and they have very different growth characteristics), the data on the graph is incorrect. Perhaps calling it a "tech" company would have been more clear.

AWS's $3.1B revenue in 2013 would be year 3 of the graph. Everything after that is speculation, but is compared to actual historic data from other companies (without explanation).

Several of the other companies lines seem significantly off also. For instance MSFT's first billion dollar year was 1990. By year 3 they were making $2.76B, not the just under $2B shown. By year 4 they were showing $3.75B, not the under $3B shown. By year 6, MSFT was making $5.94B, not the $4.5B shown. Similarly, the lines for CRM and VMW are showing lower revenue numbers than the actual ones for the later years. For instance CRM should be $3.72B in year 4, not the $1.75B shown.

If you stop the graph at year 3, where actual AWS data stops, AWS has grown only a little bit faster than MSFT. Also, the periods graphed for MSFT, CRM, and VMW all include recessions, whereas AWS is growing during a boom.

eevilspock• 12 years ago

Furthermore, if you inflation adjust all of the numbers, AWS's graph will deflate further.

Amazon's Cloud Is One of the Fastest-Growing Software Businesses in History

Related Stories