> If an engineer needs that much time you have a serious technical debt for setup of your software or a inexperienced or inept engineer.

Everyone does something for the first time once. Just because someone has not set up a hypervisor before doesn't mean they're inexperienced.

> I can set up Proxmox VE as hyper visor, some container for each DBs, load balancer in front and some app in about an hour max from scratch,

And I can spin up containers + load balancer on AWS in less than five minutes. That doesn't mean that it's just an easy thing to do. (although, this specific example is).

> upfront setup and frequent maintenance is naturally something you need to price in. <...> the cloud offer may not really be your salvation and just delay the fall while costing all the more.

Agreed 100% on both counts.

The difference with a PaaS/serverless system is that I don't need to hire you, or have someone on my team learn to be you.

I'm sorry, but all that stuff you describe doesn't bring any business value. My customers don't care what hypervisor I'm running, so I don't care either. PaaS means someone else deals with it, forever. The last time I had to employ an ops (or devops) person was 2007.

That's true - if you want to make a fair comparison between the two you need to consider the costs of the setup on AWS vs the cost of setting it up on whatever your platform of choice is. For a small team with no/development only loads, then a $5 digital ocean droplet would likely work for them, maybe even 10 of them. It's not worth managing a VPS for deploying 5 containers when you can have DO do it for $25 behind a load balancer. For a small team with moderate load, the question is "is it worth spending X on setup to save Y but potentially spend Z on maintenance of the systems on Hetzner/whoever, vs spending A on setup, B on compute and C on maintenance". If the difference is < 6 months salary, you go with whatever your current team is comfortable with and reevaluate in a year.

For a large company, it's not about $ cost, it's about risk management and avoiding cost centers.

Good point. If you are using Linux on a daily basis, it's easier to set up a server than configuring AWS.

> You'd have to figure out the setting up stuff just once though and then automate it.

You're assuming that this is for recurringly set up infrastructure. Sometimes infra is set up once and maintained, othertimes it's set up and spun down. It's also not always automated. The time spent automating something like that might not be worth it in the medium to even long term.

> Also, it doesn't take a week.

The actual amount of time it takes doesn't matter; if it's a day or a month. what matters is costing the time spent on setting it up and maintaining it, and pricing that against AWS costs.