Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder which can result from a range of observed symptoms pertaining to social, emotional, and cognitive processing difficulties especially regarding communication of their own feelings, opinions, and thoughts with others. Although a diagnosis received by 1 in 68 children in the United States, a cause for autism is still unknown. Both genetic and environmental factors during prenatal and postnatal development are believed to account for the emergence of ASD. It is known that early treatment improves language and cognitive abilities later. However, currently, the earliest diagnosis of ASD can be made only around 18 months of age. It is critical to be aware of an altered initial brain anatomy in ASD for better predicting what aspects of brain function will likely preferentially and precociously deteriorate, which would allow for potentially earlier/better prevention and treatment. We hypothesized that prospective ASD patients will show region‐specific abnormalities in developmental stages. We have used retrospective clinical brain MRI of the fetuses who were diagnosed as ASD in later age to obtain earliest MRI‐based regional volumetric biomarkers. Thirty‐nine fetuses (mean age 24.87 ± 5.05 gestational week [GW]) were included in this study (9 prospective ASD patients, 20 neurotypical controls, 6 non‐ASD controls with neurological comorbidities observed in the prospective ASD patients, and 4 non‐ASD controls with non‐neurological comorbidities observed in the prospective ASD patients. After preprocessing, the images were segmented automatically with an atlas‐based automated anatomical labeling method. A set of three‐factor ANOVA was performed for lobar and regional comparisons of the segmented brain regions. The insula/insular lobe showed statistically significantly larger volume in ASD than that in all three control groups in the lobar comparison. In the regional comparison, the ASD group had statistically significantly larger amygdala, hippocampal commissure, and insula compared to the non‐ASD controls with neurological and non‐neurological comorbidities. Our results suggest that an increased volume of the insula/insular lobe may be a strong prenatal MRI biomarker that could predict the emergence of ASD later in life. The larger amygdala and hippocampal commissure could be additional biomarkers for identifying such prospective ASD patients.