Take a closer look at the EC2 section in this post and you’ll see that I discounted the 5 year AWS cost by an overall 27% to take into account future price reductions. If I recall correctly, this was something like an 8% compound annual price reduction. And note that discounts would only apply to the usage charges since Amazon is not likely to refund any of your reserved instance fees if you go that route. If you check out the calculator linked in the most recent post, you can customize your own assumptions about future price reductions on Amazon.

I do agree that for many companies an Amazon public cloud will make more financial sense – especially those that are very small (less than a rack or 2 of servers), or have very bursty or seasonal compute requirements.

But for many companies who need more than a couple racks worth of servers, I believe that at least for the next few years a private cloud for baseline workload, with occasional use of a AWS for peak / temporary workload servicing is the best approach. If you run the numbers, you may agree.

I’ve been thinking a lot about how there are actually some fundamental inefficiencies in Amazon’s EC2 model compared to what you can achieve in a private cloud. I’ll outline some ideas in a post in the next few days.

The economics of the cloud should beat internal small data center solely on the basis of IT administration cost reduction.