My approach to the solution would be to have a completely open hardware solution for TPM modules with design, manufacturing, and distribution overseen by a coalition including the likes of the EFF.  These would take the place of government issued passports, and would fit the bill of being relatively cheap, but not so cheap as to allow trivial mass accumulation.

To further gain confidence in the node distribution, employ methods like Kaminisky's 'nOOter' and Eli Ben-Sasson's 'PCPs' as presented at the SJ 2013 conference.