We went through something similar as chumby industries was dying where a lot of effort near the end (and even past it) went into network traffic optimization aimed at "keeping the lights on" as long as possible past the retail death of the product. Significant dependencies on Amazon's AWS/S3 infrastructure were built into our backend at that time, which limited our effective ability to host on something cheaper without making changes much larger than were warranted to make on a volunteer effort for maintaining a dead service.
TL;DR/Lesson learned -- It is a good idea to implement your backend services in such a way that they could, in the future, be deployed on vanilla Linux hosting if need be, even if you currently rely on some "cloud" provider with its own APIs.