It takes almost no work at all. If some bean counter at Netflix tells their board next year that they the current cost/benefit ratio for EC2 now makes no sense

It takes almost no work at all. If some bean counter at Netflix tells their board next year that they the current cost/benefit ratio for EC2 now makes no sense it would take Netflix less than a month to move to another cloud provider and less than six months to move everything in-house again. if this were true, netflix would have never moved to the cloud in the first place, because like I said, even right now, at scale that is ridiculously small compared to neflix, it's pretty easy to beat amazon on price. Now, first, I mostly agree with your assessment, if we limit the field to smaller companies where the people who own it are still involved with running it, and where people in positions of power have experience running cheap data centers. But, I think the rules are different for "the enterprise." many of the large enterprises I've worked for ended up paying many times over what a company with more data center experience on the board (or in the beancounters) would would pay for compute infrastructure. You've seen the enterprise sales process... the whole system is set up to extract as much money as possible from the enterprise in question. I saw one company that was paying a thousand bucks a month for mid-spec servers to a well-known infrastructure outsourcing company; servers that would cost me close to $2000 total capital cost, and then $100/month in power/bandwidth. According to this company's bean counters, paying that huge premium was actually cheaper for the company than buying and hosting in-house. As for moving from amazon to another outsourced infrastructure provider... well, the problem is one of sales. Amazon is the only company in it's space that I know (with the exception of rackspace, maybe?) that has the credibility to be used by the enterprise without sending a sales guy over. This is incredibly valuable.