Why McKinsey's Cloud Report Missed the Mark


Sixty years ago, the first Kinsey Report was released, challenging conventional beliefs and causing consternation in the community. Last week, a McKinsey report did more or less the same thing, showing that for most enterprise customers, moving applications to cloud infrastructure would more than double their total costs. McKinsey’s conclusion: moving to the cloud may be a mistake for most enterprises.

A major software, systems, and services vendor in this space retorted that the McKinsey analysis “failed to consider that most companies would only use cloud services for a small portion of their technology needs.”  Unfortunately, that sounds a little bit like the argument that burning hundred-dollar bills is fine as long as you do it for only a portion of your portfolio. But there’s another reasons the McKinsey report missed the mark: Unit cost doesn’t matter — total cost is what counts.

The McKinsey analysis was very thorough: benchmarking a variety of Amazon EC2 (s amzn) options against a do-it-yourself approach, considering operations, administration and maintenance cost savings, and even tax implications and equipment depreciation schedules. However, one doesn’t need to conduct this level of analysis to realize that if an enterprise buys or leases a server, it’s probably going to cost less than if a service provider buys that server and adds in SG&A and margin. I’ve argued the same thing before: It isn’t clear that alleged economies of scale exist for cloud service providers, since they use essentially the same building blocks as enterprises. The conclusion McKinsey draws from this fact is this that while cloud computing and services may make sense for smaller companies, enterprises should hold off until the financial disparity is resolved.

But wait a minute.  If McKinsey had studied, say, automotive transportation, the conclusion they might have come to would be that, since enterprises and “cloud service providers” — rental car companies — use the same infrastructure components — cars — there really aren’t any economies of scale and the service provider premium then makes rental cars financially unattractive.  In fact, the rental car industry may have been “getting away” with an even higher premium than the cloud computing industry; a Chevy that one can finance or own for $10 a day might rent for four or five times that.

So wouldn’t the analogous conclusion then be that enterprises should be wary of using rental cars, until that industry comes up with a better value proposition?

Tell that to Hertz, which made roughly half of its over 6 billion dollars in car rental revenues last year from enterprises. Rental cars and other “meatspace” utility services create value for customers by cost-effectively addressing unpredictable or variable demand with dynamically provisioned resources on a pay-per-use basis.  So do cloud services.  And, as enterprises consolidate data centers, they’d best leverage content delivery clouds and other dispersed infrastructure services available from cloud service providers to meet interactive application requirements.

In other words, sure, service providers typically cost more for infrastructure on a unit-time basis.  So?  That’s not the relevant comparison, total cost is.

Joe Weinman is Strategy and Business Development VP for AT&T Business Solutions.



I’d love to see the report before I comment but I’d have to believe that Enterprise customers MUST have tons of data that hasn’t been touched in years, will never be touched for the next 10 years, but needs to be stored anyway…

I bet that Enterprises would like to control the ‘here and now’ while archiving the rest… I gotta fiigure the ‘here and now’ is about 20% of their total data and the rest is the 80%. At some point, some Enterprise CIO will realize that yesterday’s stuff isn’t worth today’s talent and investment

Nick Wright

It may be a jaundiced view, but I can see the Cloud as providing a number of exceptional benefits that must have a value, though it is true that these will depend on this existence of weak, slow IT departments and CIOs who aren’t paying attention.

As an example, the keen-to-get-on end-users may be thwarted by dullards in the IT department. Amazon Services, plus a contract or two, and end-users can get their application whatever the blue meanies say. Very bad for the corporate IT policy of course, but if the users have the political clout they will get their way and the CIO will lose face.

But the view is not clear for other reasons. In the UK at least, government splits its money between Capex and Opex (Resource) in an utterly divided way. There’s always plenty of Capital (I exaggerate) but never enough Resource money in the annual budget. Resource will now have to stretch to paying for the Cloud – you certainly can’t pay-as-you-go out of Capital. It might be easier for the government people to buy a few more servers rather than wipe out the staff bonus.

Ian Bell

The economics of scale that benefit cloud computing emerge from the sharing of unused compute cycles, storage, RAM, etc. and to a lesser extent from shared OAM&P costs — but mostly from elasticity. I agree that this is really a TCO debate but the car rental analogy, while pretty cool, only helps the argument stretch so far.

McKinsey, rather typically, has it all wrong… but I doubt there are many of us in the technology industry who are surprised by that. I find it ironic that the author works for AT&T: a company once advised by McKinsey that the mobile phone market would be no greater than 100,000 users nationwide. AT&T later had to buy their way into the wireless business, creating billionaires in the McCaw family.

Next time you bump into one of McKinsey’s freshly-minted MBA’s in the hallway, Joe, maybe throw in a little elbow. :)


I am in all support of cloud computing in general and its potential for enterprises. However, I got to say this article is very weak. I see that Roman Geyzer has done better job of trying to break down the several costs involved, the article should have been more in that lines.

Roman Geyzer

It’s easy to knock cloud computing as being more expensive than outright ownership when you’re comparing a clear-cut cost for c-computing vs a hard-to-quantify in-house TCO.

What is the cost of:
1) In-house servers’ inability to scale with demand or meet an unexpected service level? (SLA not met)
2) In-house server’ inability to be quickly re appropriated due to reduction in demand? (Idle equipment)
3) Downtime associated with in-house servers’ lack of redundancy? (Downtime)
4) Retention/development of skilled in-house resources to manage the platform? (OAM)
5) In-house servers’ procurement time on project scheduled? (Project Cost)
6) Additional development overhead performance tuning applications due to hardware restrictions? (Dev Cost)
7) Provisioning remote access to in-house server applications (Infrastructure Costs)

Clouds these days do not simply represent on-demand hardware. They are as much a software platform as they are a hardware resource. Companies don’t just roll-out Linux and Windows Servers and tell their users to just “do their jobs”. They actually run applications on these servers. To forget about the considerable value gained from the software-side of the cloud computing analysis does not fairly account for the true cost.

This is not to say that c-computing is appropriate in all cases. Many times, there are good reasons to stay away or where it is not feasible (disconnected environments), but the analysis must look at it more holistically than just a bean-counter review of the hardware economics.


Obviously you only rent cars when you need to in an area you aren’t normally in, and only for a small portion of time. This doesn’t equate to cloud computing at all.

Why rent something that you need everyday and everyhour? It just isn’t financially justifiable. The cloud folks will argue otherwise, but I think the proof will be when they see very few to no large enterprises moving to the cloud.

The cloud makes sense if you can’t predict your growth rate, or if you need the ability to handle a surge of traffic. That’s it’s real value. These guys have gotten off track and assume that it makes sense for enterprises, or for mature websites to use them. They are wrong. Time will show that it’s just not cost effective to rent when you can own.

A better analogy would be housing. You can rent or you can own. They both use the same basic item, a house. But as any financial expert will tell you that for various reasons (current economy notwithstanding), you are better to buy then to rent. Not so with the cloud.

Wake up and Smell the Coffeee….

Jeffrey McManus

>>> It isn’t clear that alleged economies of scale exist for cloud service providers, since they use essentially the same building blocks as enterprises

But the cloud infrastructure is shared among many customers and priced on a pay-as-you-go basis. That’s where the economy comes from.

There’s an awfully strange disconnect happening on this topic. The “experts” and analysts seem to be working pretty hard to discredit the economies provided by virtualization, while VCs and startup CTOs (i.c., people who are putting their money where their mouths are) are saying that the cloud is not only necessary but vital. They can’t both be right.


Well said Joe. Peak vs. base load is one big factor in driving adoption of any utility service and no enterprise application hits any decent level of concurrent base load utlizazion compared to, say, electricity. Therefore the key comparison metric is TCO, and it better include all the elements, not just the server farm McKinsey covers. Private network infrastructures, security infrastructres, endpoint management, mobility infrastructure etc are critical cost elements that must be included and they, too, end-up having a very low utilization ratio for on premise enterprise infrastructures…tons’ of money and power consumption thrown in the trash bin……..They totally missed the point.


There are economies of scale and better resource allocation optimizations when moving from Enterprise to Cloud services…

since the Cloud platforms are typically virtualized… in fact the trend is to move to platforms that offer virtualized data centers.. (e.g. SAN/Switches/etc..)


I would see it as follows:

– If you are a frequent driver, you should own your own car/ your own fleet of cars. This is especially true, if you have a lot of cars. Why should you rent a 1000 cars, if all of the get used for several hours each day?
– If you are a very infrequent driver, then you should rather rent. Why should you buy a car, if you only need it for a few days each year?

The same is true for cloud computing. If you use a lot of server capacity and bandwidth all the time, you better own your own kit. However, if you do something that requires a lot of computing power, but you only need it once a month for one day, then you are probably better of renting the power in the cloud, rather than buying your own kit.

You own the ‘base’ and rent the ‘spikes’.


The analogy from the car rental business is quite useless.

Cars are a utility with obvious resource scarcity through portability limitations: business travelers can’t take their cars with them when flying. Thus, a local player can rent you a car, even though you already have access to a car back home. This is against the very idea of cloud computing: access to your data & services anywhere anytime. In cloud computing, there are no local / temporary scarcities, based on which you could charge a premium.


Ah, I feel the car analogy is very relevant. Don’t make the mistake of comparing the analogy to the individual car owner. Instead think in Enterprise terms.

For an Enterprise, the alternative to the car rental is similar to the alternative to the cloud. An Enterprise needing cars in other cities would simply add them to their fleet, for use by corporate travelers. Thus they would have CAPX for cars, maintenance for the cars, as well as workers to manage the fleet locally. Car rental, while more expensive than owning a car (as an individual) is much cheaper than owning the car as an Enterprise.

Don’t forget that today, most Enterprises have servers in local plants and branch offices, so this is not as far fetched as you might think.

Dave Michels

I’ve always been suspicious of those guys at Hertz.

I think this is a valid point. The keys to cloud will be scalability and speed. Those are pretty important factors when you need them and worthless when you don’t.

Comments are closed.