It's an interesting question. In my experience people often have this idea that AWS can scale infinitely, and get shocked when they are faced with the AWS limits ("but I thought we could just spin up 200 instances, what do you mean we have to ask them to increase limits?"), and I've actually had errors from the EC2 api telling us there were no instances of a given instance type available (one of the larger/more unusual instance types, in one of the smaller regions).
But their margins should be large enough, and have been for so many years, that they should have been able to take that into account and adjust expansion accordingly.
I'm more inclined to think that they have simply decided to milk it while they can. I rarely meet people who have any clue what their infrastructure choices actually cost them and what they would pay elsewhere, and I regularly have conversations about how much people believe they'd save by moving to/from provider X where it transpires that said person hasn't even tried costing it out but are certain they'll save money. I also regularly talk to IT departments that have no kind of budget or forecasting in place for their hosting requirements. The state of budgeting for server infrastructure is simply shocking. In that kind of environment, if you price based on actual cost, rather than based on whether or not people believe you're expensive, you're leaving money on the table.
Interesting observation, though it raises the obvious responses:
1. What's constraining AWS's growth capacity? Isn't it simply a matter of stamping out AZs and DCs?
2. Why would Amazone create a pricing structure which encourages new entrants into the field? Contrast local retail in which defunct firms continue to sit on, and extend, real-estate leases, strictly to prevent other companies from growing into the market they've abandoned. See Dominicks / Anderson grocery stores in the Chicago area, in which seventy-two stores are being leased, at a cost of ~$1m/year, each, to prevent competitors from occupying the space:
(This works in part because rent is only a small part of the operational cost of a grocery store: electric power, labour, maintenance, and of course, groceries, are required if you're actually utilising the space.)
> Why would Amazone create a pricing structure which encourages new entrants into the field?
Because the primary entrant threats were not going to be dissuaded by low price competition. Specifically Microsoft and Google. Both are hyper cash and income rich and under no circumstances did they want to cede the cloud computing universe to Amazon. Fundamentally it's the exact same reason AWS stopped trying to compete so intensely on price cuts quite a while back now.
Put another way, if you're going up against Google, Microsoft, IBM, Oracle - speaking hypothetically it's better to own 1/3 of the market and print significant operating income, than to own 1/2 of the market and barely make money. That operating income is responsible for roughly doubling the value of Amazon's share price (the stock took off like a rocket after it became clear AWS was going to be a cash-cow).
But their margins should be large enough, and have been for so many years, that they should have been able to take that into account and adjust expansion accordingly.
I'm more inclined to think that they have simply decided to milk it while they can. I rarely meet people who have any clue what their infrastructure choices actually cost them and what they would pay elsewhere, and I regularly have conversations about how much people believe they'd save by moving to/from provider X where it transpires that said person hasn't even tried costing it out but are certain they'll save money. I also regularly talk to IT departments that have no kind of budget or forecasting in place for their hosting requirements. The state of budgeting for server infrastructure is simply shocking. In that kind of environment, if you price based on actual cost, rather than based on whether or not people believe you're expensive, you're leaving money on the table.