Background Identifying the brain bases for phenotypic heterogeneity in Autism Spectrum Disorder (ASD) will advance understanding of its pathogenesis and improve its clinical management. Methods We compared Diffusion Tensor Imaging (DTI) indices and connectome measures between 77 ASD and 88 Typically Developing (TD) control participants. We also assessed voxel‐wise associations of DTI indices with measures of regional cerebral blood flow (rCBF) and N‐acetylaspartate (NAA) to understand how tissue microstructure associates with cellular metabolism and neuronal density, respectively. Results Autism Spectrum Disorder participants had significantly lower fractional anisotropy (FA) and higher diffusivity values in deep white matter tracts, likely representing ether reduced myelination by oligodendrocytes or a reduced density of myelinated axons. Greater abnormalities in these measures and regions were associated with higher ASD symptom scores. Participant age, sex and IQ significantly moderated these group differences. Path analyses showed that reduced NAA levels accounted significantly for higher diffusivity and higher rCBF values in ASD compared with TD participants. Conclusions Reduced neuronal density (reduced NAA) likely underlies abnormalities in DTI indices of white matter microstructure in ASD, which in turn are major determinants of elevated blood flow. Together, these findings suggest the presence of reduced axonal density and axonal pathology in ASD white matter. Greater pathology in turn accounts for more severe symptoms, lower intellectual ability, and reduced global efficiency for measures of white matter connectivity in ASD.