Over the last decades, major technological advancements have led to a better understanding of the molecular drivers of human malignancies. Nonetheless, this progress only marginally impacted the cancer therapeutic approach, probably due to the limited ability of experimental models to predict efficacy in clinical trials. In an effort to offset this limitation, there has been an increasing interest in the development of patient-derived xenograft (PDX) models where human tumors are xenotransplanted into immunocompromised mice. Considering their high resemblance to human tumors and their stability, PDX models are becoming the preferred translational tools in preclinical studies. Nonetheless, several limitations hamper a wider use of PDX models and tarnish the concept that they might represent the missing piece in the personalized medicine puzzle.