Buried in today’s news of Microsoft’s pricing scheme for their cloud service, aka Azures, is some amazing insight into their own IT infrastructure costs.
Before we continue, I suggest you go here and become vaguely familiar with Azure’s pricing structure.
Now that you are back, the most important number to look for in their pricing scheme is the overall cost of a running a CPU for a full hour (commonly known as cpu/hr) which, in the cloud age in we live in, is the common currency of all providers. Microsoft clocks in at 12 cents per cpu/hr. That is very important since it gives you a fair ratio by which to compare providers and how efficiently they are at running their IT departments.
For instance, Amazon Web Services (AWS) – undoubtedly the market leader and Microsoft’s real competition – is able charge $0.10 per cpu/hr (http://aws.amazon.com/ec2/#pricing) which, besides being 20% cheaper, also conveys that they are able to run their IT infrastructure 20% more efficiently than Microsoft. But why should you care?
Amazon is not in the business of selling enterprise IT software (at least not their own software) and services, much less operating systems expertise. However, buried in their cost structure is the idea that they are able to deliver an internal (and external) IT infrastructure that is 20% cheaper running Linux than the best Windows IT infrastructure in the world – assuming that Microsoft employs the best Windows engineers in the world. That is truly mind blowing especially if you factor in that in both cases you can cross out the cost of licensing (assuming that Microsoft’s IT doesn’t pay Microsoft for licensing of their software and that Amazon does not pay to run Linux).
An astute reader at this point of the post is either asleep or noticing that I’m overlooking a basic rule of pricing: it’s not just cost but cost + margin and maybe Microsoft’s margins are just much bigger? This is really unlikely, especially if you keep in mind that to Amazon, AWS was always intended to be a revenue generating engagement (otherwise why bother?) meanwhile to Microsoft, which feels pressure to embark in the cloud hype, this is more of an attempt to remain relevant which would require a much lower return, maybe even revenue neutral.
Again, why does this matter? I believe it matters a lot, especially if you are in charge of running a large “enterprise” IT infrastructure in which you benchmark your “cpu/hr” cost against others to assess overall company efficiency. If you are building your infrastructure around Microsoft products you can expect to hit a theoretical limit of $0.12 cpu/hr, meanwhile, on a Linux deployment modeled by Amazon, the bar is much lower at $0.10 or less (since we don’t know Amazon margins).
Lastly, the question that remains to be answered is how many MBAs worked on coming up with the proper pricing structure for Azure without foreseeing its market implications?