@nateberkopec I’m not even sure this cost is achievable for any app doing real work. If I assume they mean 16,667 requests/second, you’d need your bill to be like 90% EC2 to hit this. But it wouldn’t be. You’d have DTO, S3, Redshift, RDS, and so on.

And anecdotally, if I compare my service to others internally, we pay the same as several others that aren’t Rails, inclusive of all stuff, and we do sometimes 100x the traffic, at 1/4th the latency.