The cloud community and anyone with an opinion if not immediately, periodically and frequently gravitate to the question of cost. Is it more? Is it less? Are we being wasteful? Are we being too frugal? Will Google and AWS drive the price of storage to zero? Can we scrunch the market forces of Cloud into a theoretical framework we already understand?

And yet just as cost of computing has been accounted for poorly when assessed in the corporate data center, when similar measures are applied to Cloud, Cloud becomes subject to the same Procrustean Bed as the corporate data center.

Questions that may highlight the incongruity:

Why do compute and datacenter costs increase each year, while AWS continuously lowers costs?

What is omitted when taking the view of application costs through a framework that has never effectively evoked the real costs of the corporate datacenter?

What is the real value of an application to the business?

Why don't corporate data centers employ significant (any?) automation if controlling costs is so critical to the business?

1. Running applications after they are built is where a lot of recurring cost can be found. These costs include the usual costs as well as surprisingly large support, coordination, and application management costs.
2. Time to Market aka Speed, and its close cousin, Agility. In economics the concept of time to market is not directly related to Opportunity Cost, and yet given that information technology increasingly determines business success, time to market is what business demands. To an extent the largely manual processes driven by month-long Mobius Strip of email chains punctuated with attached Excel spreadsheets remind me of the QWERTY keyboard layout. These processes are not designed so much for speed and agility as they are to slow the pace of change in service to stability and risk avoidance. The cost of stability at the exclusion of efficiency is a cost to be avoided in the Cloud as well as in the datacenter.
3. Value of Geographic Diversity. As hurricane Sandy taught those of us in New York, risks are often correlated, or the way I usually phrase it: risks multiply each other. A blackout and a flood are not independent, nor is an earthquake and a Tsunami. I have yet to see a rigorous analysis the of the value of the ability to build and manage applications in North America, APAC, EMEA. And yet most articles I've read do not incorporate the cost of redundant hardware, storage, network required by corporate IT for every application. While I find the outages in AWS unacceptable, even so, taking the time to design and build applications for scale and cross-region availability turns out to be worth the effort. For large and mid-market firms, applications consumed by remote offices generally suffer from poor response times due to latency and other issues. And regional datacenters conform to local standards and result in challenging and costly hardware provisioning and inconsistencies. This provides a compelling reason for these offices to build their own applications and infrastructure. I don't see much about this in the usual cloud cost debate.

What I suggest is that designing and deploying applications in the Cloud either requires an element of faith, or a moment of enlightenment. One could spend the time constructing rigorous models, yet I think the real reason we see the rapid growth in Cloud service providers like AWS is that on some level its obvious that the Cloud model makes economic sense even if it can't be easily explained through cost models that haven't ever effectivel represented where and how money is spent in the traditional datacenter. And over time I see the economic utility of applications built in the Cloud increasing, and I expect the demand curve to follow the utility curve upward and I think that's what we are seeing even this early in the game.