Cloud Computing vs Build

There are always 2 main reasons why IT introduce changes in system/technology:

1. Reduction in costs (efficiency, technology, etc); or/and
   
2. Reduction in risks (outdated technology, servers, user buy-in, etc)
   

So far, the general "sales talk" I've heard so far on cloud computing is on reduction in costs. However, I'm still not totally convinced that there is a true reduction in costs, looking 3 to 5 years ahead.

Why? Maybe I'm a pessimistic but based on my experience talking to users, they always want to have changes made to the systems fast and cheap. There will always be change requests after the system is built. Cloud computing I doubt can fulfill this criteria.

Why not fast? Because cloud computing services are essentially shared services. They are like public transport. It gets you from point a to point b at a reasonable price. That's it. If you want to watch DVDs, get your own DVD player. The services that you are using are being used by many organisations/people. They can't change the software just for you. Even if they can, because of the scale of the system, it will not be fast as they need to worry about compatibility with other user requirements.

Why not cheap? Because I believe cloud computing services are built with a certain roadmap in mind. If your user requirements vary from this roadmap (so far based on my experience, it has always been the case), it will take some effort to customise the software, if it can be done in the first place without affecting the other users from other organisations. Even if it is possible, the software is not meant for such customisations and response time may be affected.

I've always felt that the word scalability is over-used. Yes, a software can be designed to be scalable, but it all depends on the user. The designer can envision how the software design can grow, if the user's present and future requirements were communicated. If not, customisations can still be made but it will not be optimal in relation to the design. This will most often result in slower response time or even more defects appearing in the system every now and then.

So over the long run, the users may request more change requests, which the cloud computing vendor may or may not be able to fulfill. This will neither be cheap nor fast in my opinion. There will also be an increase in risks if the system is unable to cater for urgent change requests. Looking at a 3 to 5 years horizon, I don't see how cloud computing can be better than building a system on-premise.

Having said that, this does not mean that all systems are not suitable for the cloud. As I've shared previously, I feel that low risk systems or general operational systems (like e-mail) may still be suitable. They are generally low risk systems that do not change much over the years.

Cloud computing in my opinion cannot support requirements scalability, which is what ultimately the users are most concerned. Not everyone can be like Apple. Or can they?