There are servers, but you don't manage them. You provide your code, and the runtime environment packages it, deploys it into containers and runs them when they are invoked. All of this is managed by your service provider (ie you don't have to manage servers or containers).
In fact, there's a small delay (around 5s) when the lambda is 'cold' on AWS while the container is spun up. The container will remain running for a while to service subsequent requests (but you only pay for the time your application is servicing requests). It is terminated when idle for a while.