In fact, the last company I worked for closed due to a disastrous switch to the cloud. Track record matters...
> The hard truth is that self-hosting only brings in meaningful improvements in cost if your operation grows beyond a certain scale
What nonsense, I’ve seen many small projects with ~500/month in hosting costs including manpower lose tons of money by trying to go with cloud services. Self hosting scales down ridiculously far because you need talent but your server guy can do other things when they don’t need to mess with servers for months on end.
You don’t always need a server, you could also just go serverless, get charged 10x while you make your architecture a distributed, slow, hard to debug mess.
The more practical day to day reason for the top management to do it is that they manage to remove a significant amount of the specific knowhow their team has and replaces it with a more general skillset which they can hire from at any point and fire any of their team without a second thought if they idk, dare to ask for a raise or something.
It's about fucking over the workers and having all the power, as always. The cost doesn't even matter.
It took half a day to setup. 1/2 hour if you'd done it a few times before.
If you were being fancy you bought two VMs, one for the webserver and one for SQL.
When you got bigger, you bought a bigger VM. Then dedicated servers. Then a web farm with load balancers.
For most companies, all the cloud did is get rid of the entirely minor hurdle of learning how to setup a server. Which these days in bigger companies the same guys who were the infra team are now just called the DevOps team and do exactly the same job, just inside AWS or Azure.
It's just quite a bit more convenient and easy to use a cloud than do the boring job of setting up your own server.
Every time you use a VM instead of some special cloud doodad thingy bell, you can get it much cheaper doing it yourself. But then you got to setup backups. And updates. And firewalls. And DNS. And install your runtime. And install your dB engine.
It doesn't take long, it's just tedious and worth throwing a couple of hundred $$$ at a cloud to forget about it.
What it is not is anything expensive or complicated.
These are things you still need to think about and setup in the cloud as well. I wouldn't even say it's less work compared to just maintaining your own one or two servers. Except for the backups, that's the only solid convenience win for the cloud in my experience.
There's some ignorance in this comment, which turns your comment into a pointless jab at pet peeves. I'll explain you why.
The value proposition of cloud providers for business perspectives is a) turning capex into open, b) lowering upfront costs infrastructure and colocation by paying someone else to use their own infrastructure and managed services, c) be able to scale up instantly to meet demand, even internationally.
The hard truth is that self-hosting only brings in meaningful improvements in cost if your operation grows beyond a certain scale and can afford to have on the payroll a dedicated infrastructure team to manage and administrate your compute infrastructure. We are talking yearly payroll expenses that are in the six or even seven figure range.
How big does your operation need to be to amortize that volume of expenses by migrating out of the cloud?
I think you should pause for a second and think really hard on why the whole world opts to pay cloud providers instead of going bare metal. If your conclusion is that all cloud engineers are oblivious to cost control, you should go to square one and try again.