Rice is the most important food crop in the developing world. For rice production systems to address the challenges of increasing demand and climate change, potential and on-farm yield increases must be increased. Breeding is one of the main strategies toward such aim. Here, we hypothesize that climatic and atmospheric changes for the upland rice growing period in central Brazil are likely to alter environment groupings and drought stress patterns by 2050, leading to changing breeding targets during the 21st century. As a result of changes in drought stress frequency and intensity, we found reductions in productivity in the range of 200–600 kg/ha (up to 20%) and reductions in yield stability throughout virtually the entire upland rice growing area (except for the southeast). In the face of these changes, our crop simulation analysis suggests that the current strategy of the breeding program, which aims at achieving wide adaptation, should be adjusted. Based on the results for current and future climates, a weighted selection strategy for the three environmental groups that characterize the region is suggested. For the highly favorable environment (HFE, 36%–41% growing area, depending on RCP), selection should be done under both stress-free and terminal stress conditions; for the favorable environment (FE, 27%–40%), selection should aim at testing under reproductive and terminal stress, and for the least favorable environment (LFE, 23%–27%), selection should be conducted for response to reproductive stress only and for the joint occurrence of reproductive and terminal stress. Even though there are differences in timing, it is noteworthy that stress levels are similar across environments, with 40%–60% of crop water demand unsatisfied. Efficient crop improvement targeted toward adaptive traits for drought tolerance will enhance upland rice crop system resilience under climate change.