For that reason, the usual trick is to use asymmetric encryption to protect a symmetric key (a fresh one per payload) that is then used for the actual payload. That way you get the speed of symmetric encryption with the benefits of asymmetric ones. (Though evidently that's not what the PS5 uses here, otherwise it would be useless to read out one symmetric key).

First of all, we are talking about on-device decryption, so there is no difference in terms of security whether people extract a symmetric secret key or an asymmetric secret key.

Secondly, yes, asymmetric encryption is slower, so you almost always use hybrid scheme: asymmetric scheme (ElGamal, for instance) for encrypting the symmetric key, and then a symmetric scheme (e.g. AES) for the bulk of the data.

Only in very exotic protocols such as digicash blind signatures or anonymous credentials with "algebraic MACs", you want your actual content encrypted directly via ElGamal or RSA, so that you can do some operations or proofs about the ciphertext — in those cases you can't encrypt symmetrically as that would erase all the algebraic relations.

AES can only encrypt a limited size too

This isn't a good answer for "Why don't they use asymmetric keys for everything?" as you basically just explained that it would be more secure to use asymmetric keys.