No you need to download all data first so you dont have to trust other nodes that their data is correct. Once you build your database you dont store the blocks though.
So let me get this straight, you download all data and verify it after which you can choose to store X amount of GB and relay only that data to other nodes?
Almost, you relay nothing though. Well nothing but transactions and new blocks, new as in the very latest.
Initially I thought it was going to work differently, but this does make more sense if it is the case. I haven't done enough reading recently and I'm sure that many understood the concept differently.
I also hoped that I would be able to share the 40GB I have stored, but I cant. This would allow pruned nodes to still be used for syncing. Several pruned nodes could together still have the entire blocks, each node a part. Im not sure if something like this is planed for the future.