Will GaN-on-Si displace Si and SiC in power electronics?
Dr. Philippe ROUSSEL
Yole Développement
45 rue Sainte Geneviève, 69006 Lyon, France. roussel@yole.fr +33 472 83 01 86


GaN is an already well implanted semiconductor technology, widely diffused in theLED optoelectronics industry. For about 10 years, GaN devices have also been developedfor RF wireless applications where they can replace Silicon transistors in some selectedsystems. That incursion in the RF field has open the door to the power switching capabilityin the lower frequency range and thus to the power electronicapplications.Compared to Silicon, GaN exhibits largely better figures for most of the keyspecifications: Electric field, energy gap, electron mobility and melting point. Intrinsically,GaN could offer better performance than Silicon in terms of: breakdown voltage, switching
frequency and Overall systems efficiency.

A $16.6b 2010 TAM and some key promising applications
GaN technology is maturing and now offers transistor, diode and even ICs compatiblewith Power Electronic expectations, at least in the 0-600V range. Looking at TotalAccessible Market, a $16.6b market size is envisioned, split in:
• Power IC s
• Power Discretes
• Power Modules

GaN power devices: a $350M market in 2015
GaN power electronics market has just started in 2010 along with the announcementsof IR and EPC Corp. about their first products introduction. To now, the maximumcommercially available Vb is of 200V that partially covers the IT and consumer segments.In their roadmap, these 2 companies and the other possible new incomers (MicroGaN,Furukawa, GaN Systems, Panasonic, Sanken, Toshiba and so on…) are announcing 600Vand even 900V devices in a very short time. Such an increase in the Vb range, will allowGaN to step by step enter into the industrial and automotive segments.Taking into account the minimum qualification period needed for new technologyimplementation, we do forecast that the inflexion point for GaN market ramp-up will occurearly 2012, leading to a $50M+ market size by 2013 and ~$350M by 2015.
In 2015, GaN device business should be equally split between IC’s, discretes andmodules.

GaN-on-Si is probably the most cost-effective solution at short term
About 67 % of Power Electronics market is looking at 0-900V voltage range, mostlymade of cost-driven consumer and IT applications”.. To address these segments require ahigh-volume manufacturing capability as well as a very aggressive market price positioning.Thus, the technical solution involving expensive bulk-GaN substrates are not compatiblewith market requirements. GaN-on-Si appears as the most cost-effective setup to reach atleast the 0-900V applications. It has been calculated that GaN-on-Si HEMT could be 50%
cheaper than the same SiC device. However, as of today state-of-the-art, it remains twiceand even 3 times more expensive than the similar silicon device. The choice to integrate GaN instead of Silicon will be made at system level, whilecalculating the overall module cost. Implementing GaN will allow reducing:
• Thermal management costs (fewer fans, smaller heat-sink…)
• RF filtering costs (higher switching frequency will need small capacitors and
• Overall housing cost (30% to 50% overall module size shrinking is expected)

So we assume GaN-on-Si is the preferred solution to enhance GaN market penetrationover the power electronics industry at least in the early times. Based on the expected priceerosion of the GaN-on-Si 6” epiwafers over the time, the GaN substrate market shouldexceed $100M in 2015.