The best moves in the beginning of the game are both offensive and defensive, but I see castling as primarily defensive: the default strongest defensive position in chess is the short castling, despite the countless ways it can be attacked.
Whenever I choose not to short castle, I either have an offensive plan or the opponent has the long castle or no castle. The point is, in order to be (successively) aggressive, you need to deny counter-attacks at some level. Choosing not to castle means you think you have a firm grip of the oppoturnities you are giving. Not a beginners choise.
1. e4 e5
2. Nf3 d6
3. d4 cxd4
4. Qxd4 Nc6
5. Bb5 Bd7
6. Bxc6 Bxc6
7. Nc3 Nf6
8. Bg5 Be7
9. OOO OO
I just wanted to add a board link to this comment as I find it very interesting.
Someone had mentioned that castling brings the rooks together. It the linked line, white’s rooks are definitely linked, but black’s are not. Obviously black can link the rooks together, but the point is that castling won’t magically bring them together.
Also, white’s king is still just a little bit out there. That can be good and bad. The dark square bishop might still get a line on the white king, but, as the endgame draws near, the white king will be more prepared to move toward the center of the board.
The center is a dangerous place for the king in most situations, but it becomes more important as a moving piece in the end game.
From a strategic overview, castling often signals then end of a player’s opening and the beginning of their mid-game tactics to attack and control the center.
If your opponent has already castled and you’re still several moves away from doing so, that’s usually not good.