RSA in CMOS?

Many people have been talking about cryptographic applications in
hardware on the list.  I suspect these will be largely untapped until a
chip that does general-purpose RSA encryption comes along. There are
many DES implementations that could be integrated with such a chip and
then worked into computer cards, telephones, cryptography kits, etc. 
Does anyone know of the existence of an RSA chip? To the best of my
knowledge they do not exist. Are there any plans for an RSA chip? The
chip would ideally *not* be constrained to work with DES, or at least
the DES encryption would be an option in addition to just pure RSA
encryption of arbitrary data, for reasons of potential insecurity in
DES and emergence of other superior algorithms (which upon appearance
would not automatically make a general-purpose chip obsolete).

>Does anyone know of the existence of an RSA chip? To the best of my
>knowledge they do not exist. 

Cylink makes one, as well as Mykotronx.

I don't have data sheets here, but the Cylink chips are a fairly old
design, do modular exponentiation, multiplication, and addition.  One
is 512 bits wide (roughly), the other is 1024; these sizes are
inexact--the actual width differ by a few bits.  They run at 16 Mhz
(or at least one of them does).  They're implemented in an old design
process; just reimplementing them in .8 micron could speed them up a
lot.  They've been out for a few years.  The design is patented; I've
read the patent, and there are plenty of other ways to do the
calculations.

The Mytronx chip, the MYK-80, has a full modular exponentiator on it,
as well as SkipJack.  The other name of the chip is Capstone.  It's
not yet shipping.  I take it, though, that this is unsuitable.

There are also at least four commercial announcements of European
exponentiator chips that I have seen, as well as some academic work
which is going to silicon in Britain.

There's no shortage of the chips, just the will to deploy them and the
market awareness for the need for them.

Eric