Each node that stores the blockchain can decide when to prune old data, but there will be 'archival' nodes, that store all the data, all the time.
Which incentive have archival nodes to store all the data?
Hard drive space is cheap, you need only very little incentive to just keep the data laying around. Doing any sort of statistics may be fun enough for many.
To provide a better service to its users. Or just to spy on them.
