McKinsey & Co.'s conclusion that cloud computing is twice as expensive as do-it-yourself data centers is welcome news. Even though McKinsey brings some shaky assumptions to its analysis, it's the kind of cold slap of reality that IT departments need if they're to avoid betting too much, too soon on the cloud.Everyone agrees that the hype around cloud computing is at an all-time high. In this environment of inflated expectations, a good dose of skepticism--and some hard ROI data--are much-needed, so kudos to McKinsey for both. The industry needs more independent, critical assessments of the cloud model, including hands-on testing of cloud security and service reliability. (Along these lines, InformationWeek is preparing an analysis of storage-as-a-service offerings, based on a publicly issued Request For Information. The publication date for our report is May 11.)

McKinsey's "report" (it's really just a long PowerPoint presentation), called "Clearing The Air On Cloud Computing," isn't entirely negative. McKinsey finds that cloud computing "already makes sense for many small and medium-size businesses."

It's the wholesale replacement of corporate data centers with compute clouds that McKinsey warns against. Its key data point is that public cloud services (using Amazon Web Service as the point of reference) cost $366 per CPU per month, while on-premises systems can do the same job for $150/CPU per month. McKinsey advises that IT efforts are better spent virtualizing corporate data centers than replacing them.

Here's the flaw in McKinsey's argument: Few if any major corporations are looking to replace their data centers with a cloud. In all the conversations I've had with people, the concept of the "server-less company" is one that's only feasible for startups and SMBs. I haven't heard any CIOs talk of fork-lifting their entire data centers in favor of cloud services.

The more likely scenario, and the one that's playing out, is that companies of all sizes will use cloud computing to supplement their own IT infrastructures. Cloud bursting for temporary server requirements and special projects, storage-as-a-service to back up and augment on-premises storage, and software-as-a-service in lieu of certain business apps can be great alternatives to throwing more money and resources at internal systems. Indeed, McKinsey acknowledges that the cloud computing may be economical in more limited scenarios.

And McKinsey is right about one thing. It's a mistake to assume that cloud computing will be cheaper than do-it-yourself IT. Microsoft senior VP Chris Capossela made that point to me in an interview a year ago, in pointing out that SaaS and cloud services are recurring expenses. He called expectations of automatic savings "one of the most overblown things" about cloud services. "You're going to pay forever," Capossela said. "It's a subscription for goodness sake."

McKinsey paints cloud computing too much as an either/or decision, and that's the wrong way to look at it. IT pros need to do both--virtualize internal systems like crazy and investigate cloud services as a fast, flexible, and cost effective (if not always cheaper) option to capital investment in on-premises software and hardware.

Nevertheless, McKinsey has done the industry a service in trying to diffuse faulty assumptions about the new cloud model. I have no doubt that cloud computing will be a viable, long-term market, so the sooner we expose it for what it really is, the better.

For another take on McKinsey, see Michael Hickens' "Anti-Cloud Hype Is Hype."

InformationWeek has published an in-depth report on private cloud computing. Download the report here (registration required).