from my limited knowledge of semiconductors, I would try to limit the 'transistions' through 'layers' as much as possible. And work with maybe a Chemist to determine if the materials used can be changed to allow better electron flow. I graduated in 01, might have been advancements in fiber and optics since then.
Well, there exist two directions to make semiconductor devices work faster
1- Improve the existing devices
2- Introduce new devices
Improving the existing devices:
If you're concerned with the mainstream of CMOS technology, the delay iseither intrinisic (channel length and interelectrode capacitances) or extrinsic from the connection lines (RC delay). Intrinsic delays may be improved by proper design of transistors (reduce channel length and reduce interelecode capacitance by proper engineering of device geometry and doping profile). The multigate MOSFET is one of these innovations. Other nano MOSFET and even ballistic transistors are invistigated. The so-called Si wire nanotransistors (SiWNT) are also investigated.
Extrinsic delays may be reduced by employing new materials for metallization (e.g. Copper-Damascene technology) and even trying to integrate optical communication lines (optical transceivers on the same chip). Actually this is hindered by the fact that the standard CMOS Si technology is not ready right-now for implementing optical devices. There exist a branch of research for Si Laser and optical devices.
If you're concerned with compund semiconductors and quantum devices, there exist an extensive research to extend their operation for higher speeds (up to TeraHertz range) at room temperature
Introducing New Devices
There exist an extensive research to hybridized Si and other semiconductor devices with carbon nanotubes. Emplying a carbon nanotube in Si Field Effect Transistors has been proven succefully .