How the brain deals with more than one language and whether we need different or extra brain language sub-networks to support more than one language is unanswered question. Here, we investigate structural brain network differences between early bilinguals and monolinguals. Using diffusion-weighted MRI (DW-MRI) tractography techniques and a network-based statistic (NBS) procedure (Zalesky et al., 2010), we found two structural sub-networks more connected by white matter (WM) tracts in bilinguals than in monolingual; confirming WM brain plasticity in bilinguals (Luk et al., 2011; Mohades et al., 2012; Schlegel et al., 2012). One of these sub-networks comprises left frontal and parietal/temporal regions, while the other comprises left occipital and parietal/temporal regions and also the right superior frontal gyrus. Most of these regions have been related to language processing and monitoring (Abutalebi and Green, 2007); suggesting that bilinguals developed specialized language sub-networks to deal with the two languages. Additionally, a complex network analysis showed that these sub-networks are more graph-efficient in bilinguals than monolinguals and these increase seems to be at the expense of a decrease in whole network graph-efficiency.