With the manufacturing process of CPUs shrinking every year, it's only a matter of time before standard electrical switching is no longer an option, as it will eventually become impossible to insulate between layers. Using light to transfer data will address this problem, while also allowing incredibly fast data transfer rates.
Back in February 2005 Intel announced that it had developed the first continuous wave, silicon Raman laser. This was Intel's first tangible step towards the reality of optical switching instead of traditional electrical lines. But this was only a small step, since this original silicon laser still required an external pump.
In September 2006 Intel announced an electrically pumped, hybrid silicon laser, which meant that the first link in the chain of silicon photonics had dropped into place. But this just meant that light could be transferred in silicon, not data.
The next link in the chain was a silicon laser modulator, which would convert data into light, thus allowing it to be transferred and routed around the silicon. In July 2007 Intel announced a 40Gbps silicon laser modulator, providing lightning fast data conversion and transmission over a laser beam.
The other end of the chain is a method to extrapolate the data from the light beam and switch it back into traditional electrical signals - for this you need a photodetecor. Today Intel announced a 40Gbps photodetector which goes hand in hand with the laser modulator announced a few months ago.
With all the pieces now in place, the reality of silicon photonics is almost within reach. Mario Paniccia, Director Photonics Technology Labs, indicated that we should see real products based on silicon photonic switching by the end of the decade. In fact Intel sees photonic switching as a key component in its future Tera-scale computing platform.