> You should also calculate the cost of getting it up and running.

I was not doing the calculation. I was only pointing out that it was not as simple as you make it out to be.

Okay, a few other things that aren't in most calculations:

1. Looking at jobs postings in my area, the highest paid ones require experience with specific cloud vendors. The FTEs you need to "manage" the cloud are a great deal more expensive than developers.

2. You don't need to compare on-prem data center with AWS - you can rent a pretty beefy VPS or colocate for a fraction of the cost of AWS (or GCP, or Azure) services. You're comparing the most expensive alternative when avoiding cloud services, not the most typical.

3. Even if you do want to build your own on-prem rack, FTEs aren't generally paid extra for being on the standby rota. You aren't paying extra. Where you will pay extra is for hot failovers, or machine room maintenance, etc, which you don't actually need if your hot failover is a cheap beefy VPS-on-demand on Hetzner, DO, etc.

4. You are measuring the cost of absolute 0% downtime. I can't think of many businesses that have such high sensitivity to downtime. Even banks handle downtime much larger than that even while their IT systems are still up. With such strict requirements you're getting into the spot where the business itself cannot continue because of catastrophe, but the IT systems can :-/. What use is the IT systems when the business itself may be down?

The TLDR is:

1. If you have highly paid cloud-trained FTEs, and

2. Your only option other than Cloud is on-prem, and

3. Your FTEs are actually FT-contractors who get paid per hour, and

4. Your uptime requirements are moire stringent than national banks,

yeah, then cloud services are only slightly more expensive.

You know how many businesses fall into that specific narrow set of requirements?