Objective: While there is converging evidence of abnormal brain connectivity in autism, there has been little examination of how this connectivity varies across time, a measure known as dynamic functional connectivity (DFC). A recent study reported altered DFC of specific networks in autism, but no study has examined DFC at the whole-brain level. In this study, we examine how whole-brain connectivity varies across time in autism. Method: Resting-state fMRI data was obtained from Autism Brain Imaging Data Exchange (ABIDE). Analyses included 556 control participants (17.53±7.63 years) and 432 participants with autism (17.27±7.82 years). Data were preprocessed using Data Processing Assistant for Resting-State fMRI (DPARSF); global signal regression and deconvolution were performed. Time courses were extracted from 200 regions. DFC analyses were performed using a variable, sliding-window technique, and variance of DFC was computed. Results: Participants with autism showed increased variability in DFC compared to controls (alpha = .05, Bonferroni correction) in 52 connections. The three most significant connections were: left medial orbital frontal–left middle temporal, right anterior cingulum–right precentral, right superior temporal–left superior temporal. Of the significant connections, 54.72% were cross-hemispheric, and 41.5% included the prefrontal cortex. Conclusion: Our study provides evidence that functional connectivity is increasingly variable across time in participants with autism, suggesting increased noise and intra-subject neural variability. This is consistent with findings from previous EEG studies. Increased neural noise could create an unpredictable perceptual environment, resulting in abnormal social responses characteristic of autism. This has implications for developing autism treatments that stabilize neural variability.