Caching is a process of saving a copy of something (a web page in the case of ContentManagement) in a fast server closer to the requesting agency than the originating source server.

For example, a web page may be created dynamically (generated) by middleware on a Web Server or an Application Server. This dynamic generation may be a relatively slow process, limiting the number of requests (Hits?) that can be responded to per second.

A Caching Server? keeps a copy of the dynamic page, either in a huge RAM memory or in a very-fast-access hard disk enviroment capable of serving millions of hits per minute.

Note that the cached version is not easily customized or personalized. Some fast delivery servers may integrate personalization into a small part of a page whose content has largely been cached, after deployment/publishing from the authoring/creation server environment.

This blurs the distinction between Caching Server? and Application Server. Adding a Content Element at the delivery server is "late binding" of the Content Object?. Multiple application servers may be involved in an N Tier? architecture, and these last minute content objects may be application componenets coming in as portlets via Web Services.

References:
Wikipedia