Hsp90 is a chaperone with over 100 identified client proteins. What makes Hsp90 especially promising as a target for anti-cancer drugs is that many of its client proteins are in signaling and chromatin-remodeling pathways, and these pathways are often disrupted in many types of cancers. Recently, it was determined that Hsp90 bound to a client protein in a co-chaperone complex has a higher ATPase activity and binds to the geldanamycin inhibitor with over 100-fold higher affinity than the low-ATPase form. Consequently, despite Hsp90 being an abundant protein in most cell types, Hsp90 inhibitors accumulate at high levels primarily in tumor cells because tumor cells are "oncogene addicted" and require especially high levels of the high-ATPase form of Hsp90. Numerous classes of Hsp90 inhibitors have recently been developed, such as the anasamysin geldanamycin and derivatives 17-AAG and 17-DMAG; the macrolide radicicol and derivatives; purine-scaffold derivatives; pyrazoles; and shepherdins that bind to the N-terminal high-affinity ATP-binding domain of Hsp90. Other inhibitors have recently been shown to bind to the C-terminal dimerization domain of Hsp90, such as cisplatin and novobiocin, or modify Hsp90 postranslationally, such as histone deacetylase or proteasome inhibitors. In this mini-review, we present hypothetical mechanisms for Hsp90 inhibitors in treating cancers, preliminary studies in early clinical trials, and potential tumor-killing and tumor-promoting activities of Hsp90 inhibitors.