I think the long-term hypothesis of many of these companies is that you can hit a critical mass where your logistics can be made significantly more efficient. The current efficiency bottleneck, both in terms of cost and time, results from the fact that deliveries are infrequent and relatively distant. If enough people were ordering in an area, the theory goes, you could efficiently route delivery drivers to minimize the amount of time spent between deliveries and restaurants. Whether this works in practice remains to be seen.