

Given its nature and functionality, it’s hard to see how the TPM subsystem itself could be backdoored in any meaningful fashion without the rest of the CPU also being compromised. Whilst that is certainly possible, and the remote management engines (IME and such) almost certainly are, it seems more likely to me that the encryption algorythms themselves are deliberately weakened in some way. The NSA, famously, have done that sort of thing before, subtly influencing aspects of the design of cryptographic systems to make them easier for them to break. It would not surprise me, for example, to discover that the supposedly random key material is actually linked to something like your CPU’s ID.
None-the-less, using functionality help prevent someone who has your password, but isn’t a government actor, decrypting your data.

It wouldn’t be quite so bad if the previous gold rush ended first, but they seem to just be stacking up.