Having been hit by AWS's CPU cap before (on the t2.micro instance type) and having used many other providers, I can tell you firsthand that AWS's CPU limits are MUCH lower than their competitors'. In our case, a sustained usage of about 15% CPU caused our VM to eventually be starved of CPU time, which in turn crashed the software running on it.

I've never dealt with stricter CPU limits than AWS's. Most providers will not be happy if you peg an entire core to 100% (after all, the physical cores are oversold), but they usually don't mind if the percentage is even as big as 50%.

You've outgrown the micro. The strict CPU accounting doesn't exist beyond the micro and the nano. I regularly pegged every core offered on 4xls and what not and never heard a peep from Amazon over several years.

My point is this is an instance type thing, not a provider thing as you've extrapolated. They're experimenting with vast overselling on those instance types and it's not across the board.

Moreover, and this is something that I would think should be bleedingly obvious: this is exactly what those instance types are for. Everything up-front discusses CPU in terms of low usage and brief bursts. You're getting a discount because you're not using it in the same way as you would an M or a C or whatever.

> Most providers will not be happy if you peg an entire core to 100% (after all, the physical cores are oversold)

On anything that isn't a t1/t2 instance, AWS doesn't care a bit if you're pegged at 100% all month long.