I don't think you can build an interesting public cloud without raising money, unfortunately. At least, not without jumping back in time 25 years and starting then.
Just because a startup might raise prices in the future shouldn't stop you from using their products - with a caveat, that is, how easy it is to shift to another operator. I haven't used fly.io personally, so I don't know about that - could you elaborate on shifting to another provider just in case?
Only if your condition is that you want to do it in the next 3 years.
Given 15, I imagine quite a lot is possible.
There are a couple of things working against a boostrapped public cloud:
First, you gotta buy expensive kit. And then hope you can make your money back over the next 18 months.
Funding this is hard. You could try to borrow money, but that (a) increases your underlying cost and (b) dictates how you sell. You can't borrow money against developer usage / traction because banks don't know how to value that. So you have to do top down sales and land some big, committed customers before you can use debt.
The sales model constrains the product. There is no big, committed customer on the planet that will buy global infrastructure from a company who hasn't built it out yet. So you won't be building a global cloud, you'll end up in one region.
And, no big committed customer is going to buy something novel. They do not care that "fly launch" makes it easy for a dev to launch a new project. They care that they got the best possible pricing when they were shopping for their hardware. They might care that they can run k8s on it.
This is all fine, though. I'm not opposed to it. But I think it leaves you with something that's not as good as AWS, even though it's cheaper.
Do you provide a seamless offboarding experience?
Or do you lock in customers?
It's pretty easy to move off, though. `fly launch` generates a Dockerfile you can run pretty much anywhere.