The large majority of LEDs today are GaN-on-sapphire, produced on sapphire wafers 2" or 4" in diameter. A few suppliers have moved some production to 6". Sapphire is not cheap, and the larger wafers, which result in higher yields, command a premium.
A number of LED suppliers -- chief among them those whose logos appear above -- are committed to exploring a less expensive substrate, silicon. As LED production scales and becomes more efficient by moving to larger wafer sizes, silicon's price advantage will grow. LEDs are following the course set by semiconductors at an earlier stage of their evolution. In theory, LED makers could use cast-off semiconductor foundries mothballed by technology's evolution.
Silicon and GaN are not well matched in terms of their crystal-lattice spacing or their thermal coefficients of expansion. The TCE mismatch matters at fabrication time. The lattice mismatch is more serious, as it results in
strains resulting in periodic dislocations between the layers. At these sites Auger recombination occurs, reducing the LED's efficiency. (Auger recombination is widely agreed to be the mechanism underlying the LEDs' "droop" at higher current levels.) Techniques such as employing a buffer layer can compensate to some extent. Proponents claim that within 12 months the performance of GaN-on-Si LEDs will match that of sapphire-based LEDs.
It is not clear how big a difference cheaper LEDs will make to the total cost of an LED light bulb or a luminaire. Besides the LEDs, these incorporate electronics for the driver, a heat sink for thermal management, optics to direct the light, and more. Even dropping LED costs by three-quarters, as some silicon proponents have suggested is possible, might only bring down overall prices by single-digit percentages.
The analyst community is not unanimous on the prospects for silicon. Estimates of GaN-on-Si market share by 2020 range from 10% to over 20%.