I’ve done some research years ago here Constant size fully validating node
The long story short is that it seems we already have utreexo-like equivalent just from our MMRs. The problem is that utreexo is not enough for Grin because the consensus not only checks if inputs are in the UTXO set, but also that created outputs are not (no duplicates rule). Utreexo is only dynamic accumulator meaning it supports inclusion proofs, addition and deletion, but it does not support exclusion proofs. We’d need a universal accumulator to be able to check this.