Forget gearboxes, a belt drive would be superior until you start cranking out huge power, and in that case you could try chains instead. Also, don't forget that a 60hz motor does not have to spin at 60rpm to generate 60hz.
The centripetal acceleration of the rim of a rotor of radius r is rω². Rotating at 600THz (i.e. 600 trillion rotations per second), ω = 600T2π/s ≈ 3.8 × 10¹⁵/s. If your radius is 1mm, then your acceleration is about 1.4 × 10²⁷ G. The smallest rotor you can make out of atoms is probably around 0.1nm, which reduces the acceleration to only about 1.4 × 10²⁰ G. If your rotor was, say, an orthohydrogen H₂ molecule, with a distance between the nuclei of about 62pm, and thus a radius of about 31pm, the acceleration is about 4.5 × 10¹⁹ G, which would be a weight of about 74 micrograms pulling on that single covalent bond. The nuclei would be whirling around the covalent electron cloud that bound them together at about 11.7 kilometers per second, and each of them would have about 1.13 × 10⁻¹⁷ J of kinetic energy. Unfortunately, hydrogen's ionization energy is about 2.2 × 10⁻¹⁸ J, so that's about five times as much energy as you'd need to rip the molecule apart. I think. It could work out in the infrared, maybe. You'd just need a way to get the molecule started spinning.
So I guess you'd have to make your rotor a lot smaller than a diatomic molecule, or a lot stronger than a mere covalent bond.
Generally a 60Hz generator must spin at 360rpm or slower to generate 60Hz.
Gearing that down to 60Hz at the nanoscale — without losing most of the energy to friction — could still be a significant challenge. I don't know of any hundred-billion-to-one gearboxes.
The basic difficulty with the nanomotor approach, I think, is that electrons are lighter than nuclei, so it's easier to get them to oscillate over useful distances in any particular frequency range, and this is especially tricky in the terahertz to petahertz frequency range. A nucleus, under the influence of the same electrical field as an electron, will accelerate about three or four orders of magnitude more slowly.
Ultimately this should be a scale advantage for mechanical computation, since it means you can localize an atom to a much smaller region, given a certain momentum uncertainty, than an electron. The atom can't tunnel as far, so it can store a bit reliably in a much smaller region. I don't think we're there yet.