A method includes receiving a request at a media server from a computing device for a portion of a particular rendition of content that is available for adaptive streaming. The request received while the media server has access to a single rendition of the content. When the portion is part of the single rendition, the media server sends the portion retrieved from the single rendition to the computing device in response to the request. When the portion is not part of the single rendition, the media server generates the portion by transcoding a corresponding portion of the single rendition of the content, stores the generated portion in a cache of the media server, and sends the generated portion to the computing device in response to the request. The method further includes removing portions of one or more outdated renditions from the cache while maintaining the generated portion in the cache.