diff --git a/accounts-db/src/accounts_hash.rs b/accounts-db/src/accounts_hash.rs
index 66e609dcb979d0..fe0469c33de0e5 100644
--- a/accounts-db/src/accounts_hash.rs
+++ b/accounts-db/src/accounts_hash.rs
@@ -1272,6 +1272,15 @@ const _: () = assert!(std::mem::size_of::<AccountHash>() == std::mem::size_of::<
 /// with zero lamport are treated to have the "identity" hash. add/sub such
 /// element don't impact the accumulated hash.
 ///
+/// ## Are we going to serialize/deserialize Lattice hash into DB storage?
+/// This is an open question. Unlike 32 byte Hash, Lattice hash is 2K. Too
+/// expensive to store on disk? Therefore, in current implementation, we don't
+/// Serialize Lattice hash into disk. Lattice hash is just computed from the
+/// account.
+/// In future, when we decide to serialize Lattice hash, we can add
+/// [Serialize, Deserialize, BorshSerialize, BorshDeserialize, BorshSchema]
+/// to the struct.
+///
 /// Lattice hash
 pub const LT_HASH_BYTES: usize = 2048;
 pub const LT_HASH_ELEMENT: usize = 1024;