In accordance with an embodiment, described herein is a system and method for optimizing cloud infrastructure through client request classification in a cloud computing environment. The cloud infrastructure can include one or more high-compute infrastructure instances, and one or more standard-compute infrastructure instances. Each request received at a load balancer can be checked against a request classification cache, to determine whether the request has been classified, and if it has been classified, whether the request can be routed to a high-compute infrastructure instance or a standard-compute infrastructure instance based on a classification of the request. An unclassified request can be classified based on a plurality of request parameters, and threshold values derived from the cloud infrastructure; and can be stored in the request classification cache. Classified requests in the classification cache can be provided to a cloud vendor for use in optimizing the cloud infrastructure.