Background Changes in functional connectivity (FC) networks have been extensively reported in late onset Alzheimer’s Disease (AD), being the default mode network (DMN) the key system to be affected. However, it remains unclear if FC in early‐onset AD (EOAD) would show a similar pattern than late onset AD. Method We studied 48 EOAD patients (mean age=57.40±5.53 years) and 31 healthy controls (CTR, mean age=58.22±3.94 years) who underwent resting state functional magnetic resonance imaging (rs‐fMRI) in a 3T MRI scanner. We used group independent component analysis to identify the main resting state networks (RSNs). We studied group‐differences in the spatial extent of these networks, in the amplitude of their temporal oscillations and in the temporal correlation between pairs of networks, using FSLNETS. We also evaluated the discrimination capability of FC patterns by using a support vector machine (SVM) classifier. Result We identified 17 RSNs that were further classified into DMN, visual, motor, executive, salience and cerebellum systems. The network spatial maps' vertex‐wise analysis shows regional increases of some executive networks in EOAD (p<0.05, FWE corrected), suggesting increased aberrant local connectivity surrounding the main RSN nodes. In the temporal domain, we find decreases in amplitude in the posterior DMN and the salience network (p<0.05, corrected), and increases in the anterior DMN and the cerebellum networks. With the study of correlations between networks we describe a pattern of altered inter‐network connectivity that involves both increases and decreases of connectivity. Importantly, the entire pattern of network correlations discriminated AD from CTR subjects with an accuracy of 83.5%. Conclusion Our results suggest that EOAD is characterized by a complex pattern of FC alterations involving alterations in the amplitude of the main RSNs and in the way that they interconnect. Increase in connectivity in short‐range connections surrounding the main nodes might be a consequence of the decreases of larger connections between nodes. Alterations in the salience network differ from the late onset AD literature. Overall, the entire FC pattern gives a good classification rate suggesting that it might be a good biomarker for EOAD.