The vast majority of humans are right-handed, but how and when this bias emerges during human ontogenesis is still unclear. We propose an approach that explains postnatal handedness starting from 18 gestational weeks using a kinematic analysis of different fetal arm movements recorded during ultrasonography. Based on the hand dominance reported postnatally at age 9, the fetuses were classified as right-handed (86%) or left-handed, in line with population data. We revealed that both right-handed and left-handed fetuses were faster to reach to targets requiring greater precision (i.e., eye and mouth), with their dominant (vs. non-dominant) hand. By using either movement times or deceleration estimates, handedness can be inferred with a classification accuracy ranging from 89 to 100% from gestational week 18. The reliability of this inference hints to the yet unexplored potential of standard ultrasonography to advance our understanding of prenatal life.