The intrinsic limit comes from the frequency range for which the fiber is transparent (or rather, the attenuation is low "enough"). And this in turn depends on the composition of the glass the fiber is made from.
The wavelength range currently standardized for telecommunication is 1260nm-1675nm [1], which corresponds to 59 THz as far as I can tell. I don't know which wavelength range Vint Cerf had in mind.
I'd be interested to hear how close this record gets to the theoretical maximum, and where the intrinsic limit comes from.