Setting up a load balanced system and front-end web cache brings "zero value" to end users? You may as well make the same argument over design or for that matter development.

Learning how to program a loop is a total waste of time which brings zero value to the end user. If I can hand that drudgery off to someone else, I'll be adding features and stealing your customers while you're giddy about nesting your loops.

Learning how to use photoshop is a total waste of time which brings zero value to the end user. If I can hand that drudgery off to someone else, I'll be adding features and stealing your customers while you're giddy about reducing your PNGs.

And both those statements are true. If there's a third party that you can pay to do your photoshopping, and they deliver high-quality product, damn right I'm going to hand my photoshop jobs off to them rather than do it myself. Likewise if there were a third party development shop _that delivered higher quality than doing it in house_. Concentrate on your USP; for anything else, if you can get it done competently externally, hand it off.

Learning how to admin a server is a simple enough task for a small-ish company.

We set up on Rackspace Cloud in a couple of hours, and we're running just fine. We're lucky in that our core market doesn't give us traffic spikes.

If and when we run into scaling problems, we'll have a learning experience but that's the same whether you're on App Engine, Amazon or just a cheap web-host running PHP/cPanel/MySQL for £20/year.

Yeah, good luck stealing customers from someone else by developing for a third party's arbitrary stack, that they manage and update as they see fit, lacks tons of critical components, often sees unreported downtimes, and isn't flexible when your business needs change.

As someone who's been building major production applications on GAE for the last three years (and on traditional stacks for the decade prior), it's obvious that you haven't even cracked the manual.

GAE is not perfect but it's a much faster environment to develop on than IaaS services like EC2. I spend 100% of my time developing features and 0% of my time developing infrastructure. "DevOps" isn't even a job description in my present company.

It is true that there are occasionally some components (eg large memory indexes) that need to be "outsourced" to other parts of the cloud. It's trivial to do, but we judge each very carefully because of the added operational load.