Embodied cognition posits that processing concepts requires sensorimotor activation. Previous research has shown that perceived power is spatially embodied along the vertical axis. However, it is unclear whether such mapping applies equally in the two languages of bilinguals. Here, using event-related potentials, we compared spatial embodiment correlates in participants reporting the source of auditory words as being presented from above or below their sitting position. Chinese-English bilinguals responded faster for congruent presentations of high-power words (presented above) but not of low-power words (presented below) in both languages. Low-power words together also failed to modulate N400 amplitude or interact with language. However, follow-up analyses on high-power words showed congruency effects on N400 amplitude in Chinese, but not in English. Finally, English controls showed no effect. This suggests that spatial embodiment differs across languages in bilinguals, but the roles of culture and proficiency require further research.