Objective: The activation of different physiopathological pathways (neuroinflammation, apoptosis, and oxidation) can lead to secondary brain injury in ischemic stroke, and in animal models the administration of melatonin has reduced that secondary injury. Lower levels of serum melatonin were found at the time of admission of cerebral infarction in surviving patients than in non-surviving patients. Thus, we carried out this prospective and observational study with the aim of exploring serum melatonin levels in the first week of a malignant middle cerebral artery infarction (MMCAI) in surviving and non-surviving patients, and to explore the capacity of those levels to predict mortality. Methods: Patients with severe MMCAI, defined as computed tomography showing acute infarction in more than 50% of the territory and Glasgow Coma Scale (GCS) lower than 9, were included in the study. We measured serum melatonin concentrations at days 1, 4, and 8 of MMCAI. Mortality at 30 days was the endpoint of our study. Results: Non-surviving patients (n = 34) compared to surviving patients (n = 34) showed higher serum melatonin levels at days 1 (p < 0.001), 4 (p < 0.001), and 8 (p = 0.001) of MMCAI. Serum melatonin concentrations at days 1, 4, and 8 of MMCAI had an area under the curve (AUC) (95% confidence interval (CI)) in the prediction of mortality of 0.89 (0.80–0.96; p < 0.001), 0.81 (0.68–0.91; p < 0.001), and 0.82 (0.68–0.92; p < 0.001), respectively. Conclusions: The novel findings of our study were that serum melatonin levels in the first week of MMCAI were higher in non-surviving patients, and were able to predict mortality.