We are in the middle of deracking and throwing out (no kidding) some 3+ year old machines at my work. Care to take a guess at what a 3+ year old machine's specs are? Cause I know you don't have a chance at getting it right.
Dell C1100 1U (almost 4 years old)
2x 6core with HT, 24 active threads
144 GB RAM
10x 300 15k SAS
I think there are very few use cases where Amazon makes sense. Their VMs are very expensive, and under powered for what you pay.
And yet, very rational people, with full understanding of the costs of both environments, and pretty detailed spreadsheets (I've spoken with them) continue to buy into PaaS/IaaS offerings aggressively.
For the most part, it's because the cost of hardware isn't the only factor, but the flexibility, ability to rapidly scale (and descale), and, most importantly, the fact that Amazon takes care of all the dirty network engineering/system administration work required to keep the plumbing working.
But, hey, that's the great thing about the free market - every company/individual gets to make the choices that are most advantageous for themselves. I wouldn't be surprised that in regions where network engineers/sysadmins make less than $125-$150K/year, and there isn't a need to turn up a dozen servers overnight, (and, turn them off the night after that) - that AWS/Azure/GCS isn't attractive. But, clearly, for others, it is.
Well, I always see it this way: With Amazon you never pay for the actual hardware, because that would be a rip-off. You pay for the service of not having to "throw out old machines", you pay for being able to start a machine with one click, you pay for the services around (S3, Dynamo, ELB, Route 53 etc.) and ultimately, you pay for not having to care about 80% of infrastructure problems.
Dell C1100 1U (almost 4 years old) 2x 6core with HT, 24 active threads 144 GB RAM 10x 300 15k SAS
I think there are very few use cases where Amazon makes sense. Their VMs are very expensive, and under powered for what you pay.