The cloud is over-engineered and over-priced for personal projects and small groups. If you're a larger company and want high-availability and speed globally, then you're probably wanting a cloud provider nowadays unless you're really wanting to manage hardware yourself.
Setting up your own website for fun or something for your local business obviously doesn't need a fancy Kubernetes setup in EKS. Hell, even a moderate business could be fine if you're not expecting usage spikes or latency issues (although you'd probably want more than a repurposed desktop).
I'd have a slightly different take: managing things in-house is going to be cheaper if you have a competent team to do it. The existence of the cloud as a crucial infrastructure is because it is hard to come up with competent IT and sysadmin people. The market is offer-driven now. IT staff could help the company save money on AWS hosting but it could also be used in more crucial and profitable endeavour and this is what is happening.
I see it at the 2 organization I am working at: one is a startup which does have a single, overworked "hardware guy" who sets up the critical infra of the company. His highest priority is to maintain the machine with private information that we want to host internally for strategic reasons. We calculated that having him install a few machines for hosting our dev team data was the cheapest but after 3 months of wait, we opted out for a more expensive, but immediately available, cloud option. We could have hired a second one but our HR department is already having a hard time finding candidates for out crucial missions.
On the non-profits I am working on, there is a strong openness/open-hardware spirit. Yet I am basically the only IT guy there. I often joke they should ditch their Microsoft, Office and Google based tools, and I could help them do it, but I prefer to work on the actual open hardware research projects they are funding. And I think I am right in my priorities.
So yes, the Cloud is overpriced, but it is a convenience. Know what you pay for, know you could save money there and it may at some point be reasonable to do so. In the end that's a resource allocation problem: human time vs money.
How many companies need such a scale, but are not able to provide it inhouse for less money?
Everyone wants to be Netflix, but 99% of companies don't even need close to that amount of scalability. I'd argue, a significant part of projects could be run on a raspberry pi, if they'd be engineered properly.
