Columnar DBs can optimize a lot of space as they keep homogenous data in a single block, so they can apply strategies to compress the data in a block. Similarly, over a huge number of columns and rows, they will have considerably less fragmentation as compared to row wise.
This is one of the reasons that Columnar DBs are popular as big data DBs because less storage means fewer costs, faster reads (read fewer blocks). And since OLAP systems generally work with big data and aggregations to produce small columns as output, Columnar DBs hit the sweet spots there