I really love the Twilio model personally. I've got a system that is pretty seasonal, and it costs $10/mo for the DigitalOcean VM and pennies/mo (+$1 for the nu

I really love the Twilio model personally. I've got a system that is pretty seasonal, and it costs $10/mo for the DigitalOcean VM and pennies/mo (+$1 for the number) most months for Twilio. Around holidays and long weekends, the Twilio costs ramp up quite a bit (I think it hit $100-200 in a weekend once), but it's fine because that's when it makes money. Only having to pay when we use it helps keep the overhead down dramatically. I've never considered looking at other options because it's been great. For developers, too, I think the pay-per-usage or fixed-monthly model works well when it's something that they can tangibly understand. Looking at my example above, I'm fine with the fixed monthly DO cost because I know that there's some fraction of some real computer "out in the cloud" that is reserved for me and that is running my code continually. I also know that Twilio has a finite number of phone numbers available, and $1/mo to reserve one of those makes sense to me. But I also know that the way most telecom agreements work is that they bill based on usage, so me getting charged based on usage works fine for me. If they wanted to charge me $20/mo for a number even if I had not sent any messages at all, that'd be harder to swallow. Edit: also, it probably depends on whether you're selling to small businesses or enterprises. My experiences with enterprises is that they seem to prefer having fixed costs even if it ends up costing more. Seems like it has to do with how budgets and funds get allocated. "Our SMS sending costs went up by a factor of 100 because it's Christmas" doesn't seem like something that flies very well, even if the pay-per-usage model averages out lower over time.