A diaphragm for an audio speaker includes an outer shell and an inner shell. Each shell has a face portion and an edge portion that is formed to be substantially perpendicular to the face portion. The inner shell is inserted into the outer shell such that at least a part of the edge portion of each shell is in contact with at least a part of the edge portion of the other shell and a space is formed between the face portions of each shell. A cellular core fills the space formed between the face portions of each shell and is bonded to the face portion of each shell. The outer and inner shells may have identical sizes and shapes such that there is an interference fit between the contacting edge portions. A voice coil may be supported by the edge portions.