AWS was pretty tiny back in 2006, we reckon that the cloud computing unit had $21 million in sales and had an operating loss of $13 million. Revenues tripled in 2007 and 2008, and losses shrank until the company hit break even in 2009. Revenues and operating income both doubled every year between 2010 and 2012, inclusive and growth started to slow a bit to the 80 percent range in 2013 and slowed after that as every market and every player in any market ultimately does. During 2017, AWS sales rose by 43.2 percent to $17.49 billion, and operating profit, which at $4.33 billion, was 24.8 percent of revenue and was also up by 39.3 percent.
This is a hardware and software company that has margins like a traditional enterprise application software company like Oracle and Microsoft in their best years. Amazon has successfully turned its own IT problems as a giant online retailer into a vast profit engine – and in a way that none of the historical IT suppliers was able to do and, frankly, should have been able to do.
In the IT market at large – meaning enterprise IT – for every dollar spent on compute, storage, and networking, there is close to another two dollars spent on systems and application software above and beyond the basic operating systems. This is not the distribution at AWS, we are certain of that.
In the overall IT segment, when it comes to datacenter hardware, about half the spending is on servers, with about a quarter on storage and the remaining quarter on networking. The networking and storage shares of that hardware pie have been creeping up over the years. We think, based on the anecdotal evidence, that the portion of AWS revenue derived from compute is lower than you might expect from the ratio, and dropping over time. AWS wants to make compute cheap and easy, and get people addicted to the cheap object storage and pay a premium for the block storage and the networking to move data between clusters and across datacenters and regions.