IIRC copper conducts electrical signals at around 2/3rds the speed of light in a vacuum - so that's 2 x 10^8 meters per second as opposed to 3 x 10^8 m/s - still pretty fast in my book 😛.
From what I've read, the big problem in optical processing is switching or steering the light signals into the correct logic (directional) path without having to transduce it to an electrical signal, gating it to the appropriate output, and then transducing it back to an optical signal. If that can be done cheaply and efficiently, then we're talking processing speeds in the terahertz range as opposed to a few GHz as with electrical CPUs or GPUs. Basically optical switching uses input optical signals to gate or switch optical output signals, requiring much less power and having switching times orders of magnitude less than their CMOS logic gate counterparts.
However this seems to be yet another 'promising' technology always stuck in the lab, sorta like those infamous carbon nanotube enhanced batteries that DailyTech and THG used to go on and on about for years. Still waiting on that 48-hour battery that seemed to be "right around the corner" back in 2008, BTW... 😛