Well, this time I tend to agree with your points. Personally, I don’t believe an abstracted black box system can assure you it’s going to work as intended without any issues. To me, that’s more like an illusion, since you know no internals of that black box system. Theoretically you can sue the company if their promises are broken somewhere (e.g. you were offline for longer than expected), but suing Amazon just does not sound real to me. Also, I don’t like isolating the whole system into some ever-changing black box instead of learning how to maintain a real server, but possibly that’s just YAGNI.
But really, if DIY, someone's got to actually have it meet SLOs and SLAs. So you need a person or two, which is when those hours add up.
These days housing and benefiting an employee can cost 50% to 100% overhead, depending on firm efficiency. So, $400/hr means $800k/yr (because 40 hrs x 50 weeks = rate x 2000) but half that can be considered overhead (recruiting, real estate, benefits, training, vacation, "management" when some number of headcount requires adding a lead or manager who is expensive overhead), so it's really 400k a year which is not out of line at firms with regulatory requirements.
Anyway, if your workload is critical, you can't have only one, so call it 2 at 200k. Point is, when all these things matter, GCP/Azure/AWS isn't the thing that stands out.
---
> As the rest of your comment, personally, I see it more like a pitch to use AWS
Re AWS, I thought I was clear:
If YAGNI, don't choose it.