As far as I know, nobody has counted how many Indian MSS have been catalogued.  However, it should be possible to do some kind of back-of-an-envelope calculation for this, as follows.

There's a publication by Madras University called the New Catalogus Catalogorum (NCC)Its current editor is the energetic Professor Siniruddha Dash <dash_sans@yahoo.co.uk>.  The NCC is a digest of all published MS catalogues.  Well, not all, but most.  At least, up to the late 70s, and some later ones.  So, in NCC you can look up an author or the title of a Sanskrit or Prakrit work, and you'll get a list of the known MSS of that work, culled from the published manuscript catalogues. 

The NCC isn't finished.  Only nineteen volumes have been published, bringing it up to the end of ma (म), 37th letter of the alphabet.  There are 8 more letters of the alphabet to go, so NCC is about 37/45x100=82% done.   Each volume is about 350 pages.  Each page has about 50 MSS mentioned (this is *very* rough! - per-page counts vary wildly).  So each volume mentions 17,500 catalogued MSS, and there are 19 vols, so that comes out at 332,500 MSS mentioned so far.  And that's 82%.  So the total would be 405,487.  Say half a million.

There are *lots* of rough edges to this figure.  It's very, very crude.  But it does give one at least something to hold on to.  Half a million catalogued manuscripts out of a minimum total of 7,000,000.  That's 7%. 

But if the Koba people have put their MSS into a database - which they're doing at quite a rate, that could quite soon add 250,000 MSS to the total catalogued.  And there are other projects like that (though none so big, or well-funded).   So the total catalogued could be higher.  Say it's double.  A million.  That's 14% of the seven-million figure.  But the seven-million figure is probably very conservative.  So we're still hovering in the 5%-15% range, I'd say.

Improvements to the above argument and result are welcomed!

Best,
Dominik Wujastyk