My go to 'cloud' path is just a digital ocean droplet, using docker containers to spool up whatever you need and connect it all on a docker network.
A single server (2-4 cores) running quite literally ANY modern backend framework (node, go, C#) should be able to handle _thousands_ of requests per second, I'm not sure where or when this idea has disappeared, it seems like everyone automatically assumes their small SaaS or webshop needs an autoscaling kubernetes 20 rack workhorse of a server. Not the case at all!!!!
Scale when you need to - if you're getting the kind of traffic where you need to, by then you won't need to worry about the added cost to do the actual scaling / upgrading.
Sorry for the winded / ranty answer, I've done this like 20+ times at this point and always had to battle against the "let's put it on AWS with kubectl and 2349023 redundant instances!", when in the long run it was never needed...