Rho GTPases are critical signal transducers of multiple pathways. They have been proposed to be useful anti-neoplastic targets for over two decades, especially in Ras-driven cancers. Until recently, however, few in vivo studies had been carried out to test this premise. Several recent mouse model studies have verified that Rac1, RhoA, and some of their effector proteins such as PAK and ROCK, are likely anti-cancer targets for treating K-Ras-driven tumours. Other seemingly contradictory studies have suggested that at least in certain instances inhibition of individual Rho GTPases may paradoxically result in pro-neoplastic effects. Significantly, both RhoA GTPase gain- and loss-of-function mutations have been discovered in primary leukemia/lymphoma and gastric cancer by human cancer genome sequencing efforts, suggesting both pro- and anti-neoplastic roles. In this review we summarize and integrate these unexpected findings and discuss the mechanistic implications in the design and application of Rho GTPase targeting strategies in future cancer therapies.