ECS: Capacity Provider vs Autoscaling Group

Capacity Providers do use ASG underneath. They are just a way to make ECS "ASG aware". Prior to CP ECS would have no idea that ASG existed and viceversa. My suggestion would be to watch this re:Invent presentation about the announcement of Capacity Providers because it provides a lot of background re what they are and what their value is (Vs using raw ASGs).

As mreferre said, the CP drives the ASG capacity based on the services' task definitions which express CPU/RAM requirements for the tasks to run for that service. Instead of having to do complex logic into your CW Alarms etc. driving the ECS instances provisioning / deprovisioning, ECS deals with that for you to maintain available compute resources to satisfy services scaling.

Mind you though, for your capacity providers strategy, you can't mix-and-match FARGATE with AS capacity providers. See https://docs.aws.amazon.com/AmazonECS/latest/developerguide/cluster-capacity-providers.html#capacity-providers-considerations

Given that re:Post showed this question whilst answering another one, linking to it for posterity: https://repost.aws/questions/QU0UOJ8gczSkWlXTztwrF-ow/ecs-capacity-providers-best-practices