Are there any tools that allow to "diff" text files of terabyte size?
You can use
bdiff. It is used to compare files that are too large to be loaded in memory.
NAME bdiff - big diff SYNOPSIS bdiff filename1 filename2 [n] [-s] DESCRIPTION bdiff is used in a manner analogous to diff to find which lines in filename1 and filename2 must be changed to bring the files into agreement. Its purpose is to allow processing of files too large for diff. If filename1 (filename2) is -, the standard input is read.
Are the files sortable? (e.g. via "sort" command)
If they are, and depending on the kind of difference you are looking for and if this is a once-off exercise, why don't you sort them each, split into the letters of the alphabet (or whatever, as long as you get multiple files) and then compare each letter..
Primitive, but might work neat and easily for you?