Public Clouds Like Cheap Iron, Private Clouds Like the Expensive Stuff

July 11, 2011 Timothy Prickett Morgan

If there is a bright lining to the advent of cloud computing–virtualized and highly automated server, storage, and networking capacity–it is that not every application can or should be on a public cloud and that the kind of machine embodied by a Power Systems server running the IBM i operating system is, in fact, what one might build a private cloud upon.

The wizards at IDC have been polishing their crystal balls lately, as we report elsewhere in this issue referring to IT spending growth projections in the United States. But the people who live and breathe servers also put on their pointy hats and have done some prognosticating about how many servers will be deployed in private and public clouds over the coming years and how much money this will represent.

On the public cloud front, IDC expects that X64-based machines will dominate, with shipments worldwide reaching 1.5 million units by 2015, rising at a compound annual growth rate of 21.1 percent between 2011 and 2015, inclusive. Those servers will represent about $3.6 billion in revenues, or about $3,000 per server. But those building internal clouds have beefier applications, and more importantly, a mix of platforms, applications, and databases. IDC reckons that companies will deploy only 570,000 servers in private clouds by 2015, and that this is after growing at 22.4 percent CAGR between 2011 and 2015. While that is a relatively small amount of machinery, IDC reckons that given the preference for larger systems for private clouds (in part because of the redundant components inside the servers and redundancy built into networks of machines running mission-critical applications and in part because of a mix of architectures, including Power, Itanium, Sparc, and X64 processors), revenues for private clouds will nonetheless be a lot larger at $5.8 billion by 2015. That works out to an average selling price of $10,175 per machine.

Overall, server spending will probably hit $51.5 billion by 2015, if present trends persist, so there are two interesting conclusions to draw from this. First, true public and private cloud computing will represent about a fifth of new server spending by 2015. And second, the remaining four-fifths of server spending will be for less cloudy and quite possibly unvirtualized server instances to support monolithic, mission-critical workloads. Everything may get to a cloud in the long run, but it is going to be a very long run indeed.