Build a cost model.
Set prices that would make them profitable based on that model.
Offer the service at that price.
See if people are willing to pay for it.
Verify that costs and profitability match the model.
If the model turns out to be inaccurate:
A) Change the architecture to reduce costs to the point where it becomes profitable at an attractive price point for customers, or
B) Move on to the next idea.
It's not that complex, and more startups should be more realistic about profitability.
I feel like there's too much focus on the price here, and the thing I am concerned with is that the original comment was about how services shut down and they didn't want to invest in the unknown. Paying for that seems doubly risky to me.
That said, I get your point that if you can create a model that gives you a good sense of future profitability, you are in a better position not to die as a company.
If the model turns out to be inaccurate: A) Change the architecture to reduce costs to the point where it becomes profitable at an attractive price point for customers, or B) Move on to the next idea.
It's not that complex, and more startups should be more realistic about profitability.