AWS sorta has something like this already with CloudWatch, but it'd be nice if it was simpler and immediate instead of reactive. I just run a little reserved EC2 instance so my main billing risks are excessive data out or forgetting to renew a reservation and reverting to by-the-hour billing. So that I don't worry about it much I have three alarms that notify me if either "estimated charges > $x for 1 datapoints within 6 hours", or there's "anomalous" NetworkOut over a day, or "there's more than X total NetworkOut over a day", and another alarm that's "NetworkOut > Y for 1 datapoint within 15 minutes" that notifies me and shuts the instance down. I'd like to have a hard cap of "my instantaneous billing running total for the month, not my 'estimate', has exceeded $x, shut everything down" but what's there is something.