Prostate cancer is one of the major causes of cancer-related death in the western world. Androgen-deprivation therapy (ADT) for the suppression of androgens binding to the androgen receptor (AR) has been the norm of prostate cancer treatment. Despite early success to suppress prostate tumor growth, ADT eventually fails leading to recurrent tumor growth in a hormone-refractory manner, even though AR remains to function in hormone-refractory prostate cancer. Interestingly, some prostate cancer survivors who received androgen replacement therapy had improved quality of life without adverse effect on their cancer progression. These contrasting clinical data suggest that differential androgen/AR signals in individual cells of prostate tumors can exist in the same or different patients, and may be used to explain why ADT of prostate cancer fails. Such a hypothesis is supported by the results obtained from transgenic mice with selective knockout of AR in prostatic stromal vs epithelial cells and orthotopic transplants of various human prostate cancer cell lines with AR over-expression or knockout. These studies concluded that AR functions as a stimulator for prostate cancer proliferation and metastasis in stromal cells, as a survival factor of prostatic cancer epithelial luminal cells, and as a suppressor for prostate cancer basal intermediate cell growth and metastasis. These dual yet opposite functions of the stromal and epithelial AR may challenge the current ADT to battle prostate cancer and should be taken into consideration when developing new AR-targeting therapies in selective prostate cancer cells.