It is generally accepted that the presence of a second language (L2) has an impact on the neuronal substrates build up and used for language processing; the influence of the age of L2 exposure, however, is not established. We tested the hypothesis that the age of L2 acquisition has an effect on the cortical representation of a multilingual repertoire in 44 multilinguals with different age of exposure to a L2 (simultaneous or covert simultaneous exposure to L1 and L2, sequential acquisition of L1 and L2 between 1 and 5 years, late learning of L2 after 9 years of age) and all fluent in a late learned L3. Regional activation in a language production task showed a high in-between-subject variability, which was higher than within-subject variability between L1, L2, and L3. We, therefore, performed a single subject analysis and calculated the within-subject variance in the numbers of activated voxels in Broca's and Wernicke's area. Subjects with early exposure to L2 showed low variability in brain activation in all three languages, in the two early as well as the late learned language. In contrast, late multilinguals exhibited higher variability. Thus, cerebral representation of languages is linked to the age of L2 acquisition: early exposure to more than one language gives rise to a language processing network that is activated homogeneously by early and late learned languages, while the inhomogeneous activation in late multilinguals indicates more independent access to the multilingual repertoire. Early passive exposure to L2 results in the same low variance as active bilingual upbringing. Variability in local brain activity increases progressively from the simultaneous to late L2 exposure, indicating a gradual transition from the mode of early bilingual language representation to that of late ones.