i feel the same way!
(I've said this before so sorry if you have read it many times already - this is not a copy/paste job though)
From an infrastructure perspective using EC2 cloud is like going back to the 90s(in terms of capacity provisioning, there is no pooling of resources). If they fixed their shit where basic things like resource sharing, persistent network configuration, automatic migration of workloads so maintenance can be performed, etc. Basically get rid of built to fail for those customers that want it. Then I'd see some value in it.
Having worked closely with a bunch of developers across multiple companies who built their apps from the ground up in the EC2 cloud and not spending any time taking into account the built to fail model I feel pretty convinced that this is not a priority for most of the devs out there (of course there are exceptions but the #s are relatively minuscule). From a business strategy perspective it makes sense as well, if you spend a bunch of time making the app resilient you can't make the features your customers want (or your costs go way up if you try to do both - never mind the fact that managing EC2 instances is significantly more expensive then doing it yourself with really the only exception being very highly variable workloads that are very efficiently managed).
I've worked for the past 11 years with software companies with home grown internet-facing software, and none of them spent much time focused on making their apps capable of handling built to fail models whether they were in a cloud or operating on their own bare metal.
Sadly enough the problem seems to be incompetent IT groups in these organizations, they are not able to respond effectively to their users needs so the users decide it is simpler to bypass them and go to a cloud, and of course Amazon has the biggest name in the space at this point.
One of my friends is at a big company where he has done the math and says they are spending 700% more by using a cloud managed service provider than if they did it themselves (they are not an EC2 customer). But the company doesn't want to bring things in house -- yet at the same time they want to cut costs... all I can say is I'm glad I don't work there. There is so much poor management (at all levels even down to the operations folks, I phone screened a bunch last year) throughout the world really, and the influx of startups has put an even greater burden on those which can do the job - which means a massive shortage of good talent, leading to people making decisions like that.
I've said time and again that it takes significantly more expertise to effectively operate inside EC2 than it does on enterprise equipment (primarily because of support, training, no built to fail models etc). Unfortunately most management do not realize that or don't care all they see is puffy white stuff and think they are cool by using it.
It pains me to no end to see folks using EC2, or even other cloud players, EC2 is the worst offender though. It can be like a roach motel for those who buy into it hook line and sinker.
I have absolutely no doubt that there are a lot of investors out there that have invested in the cloud stuff and in turn try to drive the companies they invest in to use the cloud as a result. It's sad to see.
My last dealing with EC2 was about 8 months ago when my company moved the last of their stuff out of there. It's been all green meadows since then. Now I'm looking at doubling the memory capacity of the servers (hey, saves on vSphere licensing costs because like most people we are not CPU bound) so we can continue to add more VMs (the number of VMs has probably doubled since we left EC2 - and yet our costs have remained constant -- I believe the migration has already paid for itself at this point and we're barely a year from when we moved the first set of systems out). If we become CPU bound then we can upgrade the CPUs and get another ~30% capacity again without any license hit. Though we'd have to probably quadruple our CPU usage to need that.
Now if your in that fraction of a fraction of 1% of organizations who have a highly variable workload and can spin up and down things on demand for short periods of time - that is something EC2 is good for. If you nabbed yourself a password file and want to crack it on a cluster of GPU systems or something and not have to buy it - that is something EC2 is good for.
Running something 24/7 though - is a waste of money and waste of resources. Not that most organizations aren't wasting money and resources already, so I suppose it comes down to relative waste in those cases.