Amazon Web Services, by the numbers v

Table of Contents

  1. Summary


When Amazon released its third-quarter earnings recently, some stock analysts were dismayed by the amount of spending that took the shine off the company’s otherwise fine overall revenue. At least when it comes to infrastructure, though, Amazon, like all web-based businesses, has to spend money now to ensure it can make even more later on. That rule might be doubly true for Amazon’s position as a cloud-computing provider, because that business requires having enough capacity to serve untold numbers of users at any given time. Fortunately, we’re already starting to see the fruits of Amazon’s labor on that front.

Since launching Amazon Web Services in Aug. 2006, Amazon has seen consistent — sometimes exponential — growth on its earnings statement in the “Other” category that encompasses the AWS business and “miscellaneous marketing and promotional agreements, other seller sites, and co-branded credit card agreements.” Amazon doesn’t break out specific revenue within the “Other” category, but UBS financial analysts have conducted their own analysis and they project that AWS accounts for a large and growing majority of that revenue, which currently amounts to about 3.7 percent of Amazon’s total revenue.

Last year at this time, I looked at AWS’ revenue and concluded that it was outpacing analysts’ projections to hit the $500 million mark for 2010. Last week, I revisited Amazon’s earnings and suggested that AWS is far outpacing UBS projections to earn $750 million in 2011. If it has fourth-quarter growth on par with years past, AWS might do $1 billion in revenue this year, which seemingly puts it leaps and bounds ahead of its cloud competitors.

How does a business grow from nothing to a billion dollars in five years? It looks like this:


Lest anyone question whether AWS is really responsible for all the “Other” growth, one might consider the oft-cited statistic that bandwidth into and out of AWS surpassed that to as early as 2008. Or one might look at how growth in the amount of objects stored in AWS’ S3 storage service aligns with growth in “Other” revenue. Since the fourth quarter of 2009, the number of objects stored in S3 grew from 102 billion to 566 billion while “Other” revenue grew from $239 million to $407 million per quarter, including an average year-over-year increase of almost 71 percent for the past three quarters. This suggests that sharp increases in AWS usage directly affect revenue.


In March, Cloudscaling’s Randy Bias did a count of significant feature releases by Amazon Web Services since its inception. Bias noted marked increases in the number of new features each year. As of March, he found that, based on his somewhat-subjective definition of “significant feature,” AWS is on pace for 66 releases in 2011. Not surprisingly, the annual increase in features tracks closely with annual “Other” revenue growth, although which one is more a product of the other is something of a chicken-and-egg problem. Are new features bringing in more money (either from new users or existing users doing more), or does more revenue allow AWS to keep investing in more features to pad its lead?


One thing that doesn’t appear up for debate anymore — if it ever was — is that AWS and its customers are all benefiting from economies of scale. Using some more data that Bias pulled and finishing up 2011 on my own, I did a count of all of AWS’ price reductions by year. For the purposes of this comparison, “price reductions” include lowered prices on existing services, free tiers, new instance types that offer lower-cost options for a service and the addition of services into new pricing models such as Spot Pricing. AWS probably can’t keep up a breakneck pace of price reductions like it does with feature upgrades (as the 2011 price-reduction count indicates), but there’s no denying that as AWS does more business and can drive lower costs on its end, customers are reaping the rewards, too.


Some might ask why AWS gets so much attention and why so many analysts speculate about its revenue. There are two likely reasons for this: 1) Amazon doesn’t specifically disclose the portion of its revenue attributable to cloud computing, and 2) if the projections are correct, it’s a much larger business than any of its competitor’s cloud businesses. Look at how Amazon’s “Other” revenue — the majority of which, again, is thought to be derived from AWS — compares to that of presumed cloud runner-up Rackspace, which does disclose its cloud computing revenue. Rackspace hasn’t announced third-quarter earnings for this year, although one should probably expect cloud revenue to reach about $50 million.


Looking forward, though, Rackspace — by way of the OpenStack project that it launched — might be the biggest inhibitor to AWS’ reaching the lofty goals that UBS and others predict for the business. UBS projected $2.5 billion in AWS revenue in 2014 — representing 3.69 percent of Amazon’s total revenue — but Amazon is already on pace to outdo that prediction. However, most projections of AWS revenue didn’t take into account the emergence of OpenStack, which just last week was able to claim its first publicly available compute cloud courtesy of Internap. In the near future, Rackspace, Dell, HP and smaller providers such as DreamHost will have their own OpenStack clouds available, and other providers will surely join them in the years to come.

If these OpenStack providers execute well, they should be able to take a bite out of AWS’ growth with the promise of an ecosystem of clouds that share common APIs and core technologies. Up until now, AWS has been the king of an Infrastructure-as-a-Service space where it’s every provider for itself and sharing anything among providers was only a dream for customers. Plus, OpenStack public clouds should be compatible with the slew of OpenStack-based private-cloud software that’s now hitting the market. Private clouds alone won’t likely have any impact on AWS’ business, but the desire for better-connected hybrid environments could affect the public clouds with which businesses choose to complement their private clouds (VMware would also be an option in this scenario).

However, the reality remains that AWS is growing like mad, and with a large untapped market of cloud customers, it’s not likely to stop growing by leaps and bounds anytime soon. A broad OpenStack ecosystem — and perhaps a surge in popularity for Microsoft Windows Azure and some VMware-based clouds — will mean that AWS doesn’t have the IaaS market to itself, competing only against a contingent of much-smaller providers. But it’s hard to see anyone doing anything other than slowing AWS’ meteoric climb.

Access Report

Available to GigaOm Research Subscribers

Subscribe to
GigaOm Research