Stick these transceivers on the CPU/Memory/GPU silicon, have them all connected via fibre and get rid of this stupid front side bus bottleneck.
This is actually the driving application for hybrid Si-InP integration research. (http://engineering.ucsb.edu/bowers/). The biggest limitation of this technology is the inefficiency of the laser. It seems logical that Intel is going after an electronically simpler application like an integrated tranceiver, before attempting to integrate with an already hot running microprocessor.
"Kill the Wabbit, Kill the Wabbit, Kill the Wabbit!" -- Looney Tunes, "What's Opera Doc?" (1957, Chuck Jones)