Tables in HBase are initially created with one region by default. Then for bulk imports, all clients will write to the same region until it is large enough to split and become distributed across the cluster. So empty regions are created to make this process faster.