xargs runs the specified arguments as a command, followed by the filenames provided from standard input. If the length of a command line exceeds UNIX's capability, it runs the command again with the remaining arguments (and repeats until all arguments are consumed).

The reason you do this instead of: find . -type f -exec grep ... is because (1) you don't have to mess with the funny {} syntax and (2) it's more efficient -- fewer invocations of grep, and (3) when grep receives multiple arguments on the command line, it preceeds each match with the filename, so that you know which file grep found it in. (You can do this with find, but it's a hack).