According to the cross-stressor adaptation hypothesis, physically trained individuals show lower physiological and psychological responses to stressors other than exercise, e.g. psychosocial stress. Reduced stress reactivity may constitute a mechanism of action for the beneficial effects of exercise in maintaining mental health. With regard to neural and psychoneuroendocrine stress responses, the acute stress-buffering effects of exercise have not been investigated yet. A sample of highly trained (HT) and sedentary (SED) young men was randomized to either exercise on a treadmill at moderate intensity (60-70% VO2max; AER) for 30 min, or to perform 30 min of "placebo" exercise (PLAC). 90 min later, an fMRI experiment was conducted using an adapted version of the Montreal Imaging Stress Task (MIST). The subjective and psychoneuroendocrine (cortisol and α-amylase) changes induced by the exercise intervention and the MIST were assessed, as well as neural activations during the MIST. Finally, associations between the different stress responses were analysed. Participants of the AER group showed a significantly reduced cortisol response to the MIST, which was inversely related to the previous exercise-induced α-amylase and cortisol fluctuations. With regard to the sustained BOLD signal, we found higher bilateral hippocampus (Hipp) activity and lower prefrontal cortex (PFC) activity in the AER group. Participants with a higher aerobic fitness showed lower cortisol responses to the MIST. As the Hipp and PFC are brain structures prominently involved in the regulation of the hypothalamus-pituitary-adrenal (HPA) axis, these findings indicate that the acute stress-buffering effect of exercise relies on negative feedback mechanisms. Positive affective changes after exercise appear as important moderators largely accounting for the effects related to physical fitness.