We use cookies and other similar technologies (Cookies) to enhance your experience and to provide you with relevant content and ads. By using our website, you are agreeing to the use of Cookies. You can change your settings at any time. Cookie Policy.

Cloud Talk

Andre Yee

McKinsey Study Darkens the Silver Lining for Cloud Computing

Vote 1 Vote

Cloud computing has taken the IT world by storm (no pun intended, really), promising faster deployment and lower TCO. So when McKinsey & Co. recently released a study indicating that the promised cost savings from cloud computing don't hold up under scrutiny, it made some waves.

It's important to note that the report did not cover SaaS or cloud based applications but focused on "hardware based services offering compute, network and storage capacity". It defined "clouds" as conforming to three parameters -

  1. Hardware management abstracted from buyer
  2. Hardware cost is covered by OPEX (operational expense)
  3. Capacity is elastic

The report actually distinguishes between "clouds" and "cloud based services" with the latter conforming to only two of the three parameters (#1 and #3). Although they make much more of the difference between clouds and cloud services, I think this distinction is actually minor point.

But let's not bury the headline here. You may not agree with all the details but the point of the study is that, despite the immense hype, "clouds" are not delivering the cost savings to larger enterprises. The total cost of assets for a typical enterprise data center is about $45/month for a CPU equivalent. Using Amazon's EC2 as an example, the report shows how most on-demand solutions are significantly more costly than that.

What's interesting is that the negative disparity in cost actually increases, the larger the EC2 deployment. In other words, the bigger the deployment, the more cost efficient the self managed data center becomes and less attractive the cloud becomes. Unless of course, you choose to migrate the entire data center to the cloud - in which case, in addition to other cost advantages, there is a 10-15% of labor savings that will kick in as well.

I think the study is interesting for a number of reasons, not least the challenges that cloud service vendors have to overcome to break into the larger enterprise market. Yet, it doesn't take away from the value of cloud services for smaller companies, who may find data center start up costs more than they want to bear.

What do you think - is the McKinsey report right? Is cloud computing overhyped in terms of its benefits? You can read the entire report and come to your own conclusions.


I have long thought that savings from using cloud computing for hardware, network and storage didnt add up. For me there does have to be a clear distinction between SaaS and actual physical hardware and storage...

SaaS for organisations, especially small businesses really does make sense. Outsourcing hardware etc, doesnt, there are just too many concerns and I am not sure about the savings to be made.

Funny enough i only blogged about this, this morning when i read this article, so had to update my blog...


Andrew - thanks for your comments. I believe your point on differentiating between SaaS and cloud service providers is true. The value proposition for SaaS isn't just about optimizing processing, storage, etc... it's about quick deployment, faster time to value, and the business logic of the app.

The value proposition for cloud services is mostly about cost savings and when those don't add up, the proposition becomes weak very quickly. Based reports like this one and other anecdotal evidence, I think it'll be a while before large enterprises sign on.

Software as a service has some definate advantages. Outsourcing the physical infrastructure has always left me with the security question. I look to NSA as the final arbitor of security requirements. They have their own physical plant and operate their own cloud. This is expensive and probably beyond the range of most small-mid sized companies. There are trade offs. If the economics don't make sense, that's a double whammy. I'm not looking to cloud anytime soon. i need more assurance and I definately would need to save money.

Richard - thanks for your perspective. I think the issue of security is important for all but particularly important in the Federal Government and certain industry segments. The reality is that for many small businesses, security offered by the SaaS app exceeds their own internal security


Just wanted to mention that EC2 now has the option to pay up front for dedicated machine. So while before it would cost $72/month, if you pay up front for an instance that will last 3 years it only costs $35/month with $500 setup fee included in that figure. You will pay for bandwidth but it probably still wouldn't bump that number above the $45 mark mentioned in the article.

The biggest advantage though for EC2 and the reason it is so attractive to small start ups is the elastic properties of it. For example if you get posted to digg and have a huge influx of visitors you can dynamically add EC2 nodes for that one day to handle the capacity. With a standard rack that isn't an option. You either have to buy or lease hardware and then you are stuck with that hardware. The elastic properties of EC2 allow startups to only incur the cost of a CPU when needed and not forever which is important when there isn't much money to go around.


You make a great point which was covered to some extent in the report - that by prepaying, the cost benefit of EC2 becomes a lot more compelling. The advantage of EC2 and other cloud services for small companies is clear - the start up costs are significantly less than self managed environment...and as you mentioned, it'll grow with your needs

I believe if we look at 'Cloud Computing' purely as hardware and operating systems resources, we will make the same mistake as we have done in the past. We are solving one part of the problem in that we can add or remove hardware resources at will thus giving us the ability to respond to demand. However,we still have the problem of interoperability. Most organizations, be they small, medium or large, require hardware, software and application resources that they would ideally like to pay for as they use them so that they can equate business value to cost of provision.

To me, a lot of this has echos of IBM's usage based pricing model from many years ago. At that time, organizations had to a) pay dearly for large mainframe systems that were sometimes idle and other times over loaded and b) large license fees to run the software with great difficulty to apportion these costs across their business. Usage based pricing promised (though never fully delivered IMHO ) the ability to pay for hardware and software as they were used again moving somewhat towards the concept of OPEX versus CAPEX.

I think this report defines Cloud Computing too narrowly as it only addresses the issue of hardware and operating system resources. I think Cloud Computing has to offer hardware, operating system and software resources (i.e. a solution) to be available on a pay as you use model to be successfull. I saw a WebEx recently where IBM and Amazon were offering something like this though I'm afraid I don't have a link. While it may not be 100% there yet, I think this is the way to go.

John - your comparison of the cloud with the mainframe, pay by usage model is interesting. This is perhaps why "private clouds" are gaining mindshare in some larger companies. I also agree with your assessment that for cloud computing as a general category to be cost efficient, it has to cover the entire "stack" up to applications.


I see you responded to my comment but I don't see any text associated with it ? I'm obviously interested in what you had to say. Do I have to do something to see the response ? I'm knew to this sort of activity so please bear with me.

Best regards,


John - sorry. Not sure what happened but my original response to your comment did appear blank. I tried to recreate it so it's now up there


Thanks for that. To take this a step further, I also believe that these "private clouds" will need to integrate with existing data and business logic - "legacy systems" to many people but it's a term I don't like :-) If the cloud doesn't integrate in some way, we will still end up with our existing "Islands of information" whereby most of the data built up over the last 40 years will exist outside the cloud.

Andre Yee blogs about cloud computing, SaaS, Web 2.0 and other emerging technologies that matter to businesses.

Andre Yee

Andre Yee is an entrepreneur and technologist with nearly 20 years of experience in the business of technology.

Recently Commented On


Monthly Archives