Take microcode for example. At one time (as I understand it) microcode was not a signed blob. However companies wishing to hide details of their microarchitecture chose to encrypt it.

My guess is that these encrypted blobs grew first out of corporate closed source culture, which is strong in HW companies. If they are subverted with actively malicious code it was probably by secretive efforts, not the NSA simply propositioning the HW manufacturer.

Finally I'd like to point out that unless you design your CPU chip yourself and oversee the layout of it on the die, it is also possible that the semiconductor manufacturer you hire could embed their own nefarious processor within your design.

In practicality, I think running RISC V on an FPGA would have a very low risk of subversion. Though the FPGA design tools might add nefarious logic too.