Why is this so slow?

I have this script;

#!/bin/sh

find . -type f -exec sed -i -e "s/comes in/gets out/g" {} \;
find . -type f -exec sed -i -e "s/come in/get out/g" {} \;

It takes about 30 seconds to run while a simple [code]sed 's/comes in/gets out/g' -i *[code] takes about two seconds. Why's that?

why not just use s/come(s)? in/get\1 out/g?
works for me in perl