I had investigated if this was possible to do but I couldn't find a way to do it myself. It's possible that pd.Timestamp might be intricately tied interally to vectorization optimizations using numpy or others. Therefore separating out just the timestamp stuff from pandas would be a lot of work.