> The scale of the engineering effort required to compete with AWS is enormous, and no-one's really trying
I would suggest that you're wildly misinformed. Not only AWS has plenty of competitors who're silently kicking their butt one area at a time, but even as pure cloud environments go, the value of virtualization continues to evaporate as hardware continues to grow faster than the population.
In other words, theoretically you'll be just able to purchase a single $5K DELL server in a few years and run an Airbnb-sized business on it. As the Moore's law keeps going, it's getting more and more insane for a start-up to invest so much of software fault-tolerance and scaling around AWS-style farms of barely alive virtualized servers from the 90s.
For a lot of businesses even today, a single Dell R720 with a Chef script makes a lot more sense (in terms of uptime, hosting bill and engineering efforts) than a small army of VPS boxes.
If you're not CPU-bound, if you don't need x100 spikes of compute, then you don't need to overpay for AWS-style elasticity, especially in terms of engineering that's required to get stable on it.
One server will never be enough. You need redundancy in case of a fire. Don't forget to optimize kernels, load balance, ensure security patches get applied, etc. I know I don't want that world, so I pay Google a measly amount to run App Engine for me and spin up new instances as my service needs them. I will be very surprised if the future is not bigger and better clouds.
I would suggest that you're wildly misinformed. Not only AWS has plenty of competitors who're silently kicking their butt one area at a time, but even as pure cloud environments go, the value of virtualization continues to evaporate as hardware continues to grow faster than the population.
In other words, theoretically you'll be just able to purchase a single $5K DELL server in a few years and run an Airbnb-sized business on it. As the Moore's law keeps going, it's getting more and more insane for a start-up to invest so much of software fault-tolerance and scaling around AWS-style farms of barely alive virtualized servers from the 90s.
For a lot of businesses even today, a single Dell R720 with a Chef script makes a lot more sense (in terms of uptime, hosting bill and engineering efforts) than a small army of VPS boxes.
If you're not CPU-bound, if you don't need x100 spikes of compute, then you don't need to overpay for AWS-style elasticity, especially in terms of engineering that's required to get stable on it.