The sponge class Demospongiae is the most speciose and morphologically diverse in the phylum Porifera, and the species within it are vital components of a range of ecosystems worldwide. Despite their ubiquity, a number of recalcitrant problems still remain to be solved regarding their phylogenetic inter-relationships, the timing of their appearance, and their mitochondrial biology, the latter of which is only beginning to be investigated. Here we generated 14 new demosponge mitochondrial genomes which, alongside previously published mitochondrial resources, were used to address these issues. In addition to phylogenomic analysis, we have used syntenic data and analysis of coding regions to forge a framework for understanding the inter-relationships between Demospongiae sub-classes and orders. We have also leveraged our new resources to study the mitochondrial biology of these clades in terms of codon usage, optimisation and gene expression, to understand how these vital cellular components may have contributed to the success of the Porifera. Our results strongly support a sister relationship between Keratosa and (Verongimorpha + Heteroscleromorpha), contradicting previous studies using nuclear markers. Our study includes one species of Clionaida, and show for the first time support for a grouping of Suberitida+(Clionaida+(Tethyida + Poecilosclerida). The findings of our phylogenetic analyses are supported by in-depth examination of structural and coding-level evidence from our mitochondrial data. A time-calibrated phylogeny estimated the origin of Demospongiae in the Cambrian (~529 Mya), and suggests that most demosponge order crown-groups emerged in the Mesozoic. This work therefore provides a robust basis for considering demosponge phylogenetic relationships, as well as essential mitochondrial data for understanding the biological basis for their success and diversity.