I think people have covered the quality of the hash and the chances of collisions. People briefly touched upon performance, but I feel that one thing was overlooked here: intrinsics. Based on posting this with the title that you use it for Asset ID it is likely you will not be hashing this at runtime, and for a pipeline this shouldn't really matter. But never-the-less, this could be of use

Most CPUs on market today (since ~2008 if I remember correctly) have intrinsics for calculating CRC32 hashes quite efficiently (SSE4.2 and up): https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=crc.