Is your glass half full or half empty? Your relative optimism is probably the most important factor in determining how you view the latest analyses of the emerging market for wide-bandgap electronics.
That's according to Tom Hausken, one of the authors behind a new report from California-based Strategies Unlimited. On the face of it, things look good – devices based on SiC and GaN semiconductors, for growing applications like WiMAX base stations, electronic warfare and hybrid electric vehicle motors, are sure to be in demand. Hausken is expecting the total market for wide-bandgap components to grow quickly, at a rate of at least 30% per year, and to treble in value between now and 2012.
So, if your glass is half full, everything looks good. On the other hand, if you were expecting GaN or SiC to be "the new GaAs", prepare to be disappointed. "They are definitely not the new GaAs," fronted up Hausken, unequivocally. Together, SiC and GaN electronics accounted for a market worth only $40 million in 2007, and Hausken predicts that by 2012 that figure will have risen to something in the region of $170 million. It's a healthy market for a couple of companies, perhaps, but hardly enough to sustain many of the estimated 150 firms that are actively working on these materials, and entirely dwarfed by the GaN LED and laser-diode business.
For a neat comparison, simply look at the GaN-laser sector. To make the blue lasers for its PlayStation3 consoles, Sony is believed to have bought some $50 million worth of GaN substrates from local vendor Sumitomo Electric Industries during 2007. That this supply deal alone was worth more than the entire market for wide-bandgap electronic devices in 2007 illustrates vividly the disparity between electronic and optoelectronic applications.
Hausken is anything but pessimistic about the wide-bandgap sector, however, and says that a successful global launch of the WiMAX connectivity protocol could spark much faster market growth. "Something is finally happening and there is a great opportunity for GaN HEMTs," remarked the analyst. "We are not going to see them in microwave ovens, but GaN is actually getting out there."
At the moment, a lot of the applications that GaN is getting into are being driven by the military, through DARPA-funded development programs and genuine deployments. For example, Eudyna Devices – acknowledged as one of the leading exponents of GaN-based RF technology, and with an estimated 31% share of the current wide-bandgap electronics market – is supplying the US military with devices that are used in Iraq. Here they operate in an "electronic warfare" capacity, where the chips can create enough microwave "noise" to disrupt and jam the cell phone signals with which insurgents attempt to set off roadside bombs. It might seem unlikely at first, but perhaps the declining number of bomb attacks in the war-torn country is, to some degree, thanks to wide-bandgap semiconductors.
Useful though they may be for the military's jamming and high-end radar systems, it is the commercial world where GaN and SiC must succeed for these semiconductors to make a real impact. In Hausken's "most likely" scenario, he has factored in a modest degree of success for WiMAX. But, says the analyst, if this technology really takes off – something that is too early to call right now, given the divisions in the wireless industry – it could change everything. In his most optimistic scenario, in which WiMAX breaks through as the 4G wireless protocol of choice, GaN HEMTs get a real boost and Hausken values the 2012 market at $300 million (figure 1).
Even that might not look so good for venture capitalists looking to recoup their investments via a high-value acquisition or stock-market flotation. GaN-on-silicon specialist Nitronex has some great technology and has racked up some $56 million in venture funding since its inception. And in a market worth less than $100 million, a profitable exit in the next five years is hard to envisage.
Hausken forecasts that microwave applications, currently responsible for some 80% of wide-bandgap electronic device sales, will continue to dominate the wide-bandgap scene, with power-management applications likely to be worth only $10 million–$20 million annually over the next couple of years.
However, that split should skew more towards power-management applications in the longer run, and the analyst predicts approximately $60 million in sales for this subsector in his 2012 "most likely" scenario. At the moment, the power-management (as opposed to high-power RF) side of the wide-bandgap industry is dominated by SiC, and largely by Schottky diodes from Cree and Infineon Technologies.
Hausken expects SiC to continue that domination for the foreseeable future, as early adopters begin using SiC diodes and transistors in electric vehicles, and in DC-AC inverters that connect continuous energy sources, like photovoltaics and wind, to regional power grids. The market opportunity here is quite different from that for GaN HEMTs, explains Hausken. While GaN components wait for the opportunity to emerge in high-speed wireless, applications in power management are already plentiful and would potentially support far higher unit volumes than wireless base stations. The key factor holding up SiC here is its current cost, and the big challenge is how to make the wide-bandgap technology competitive.
Though more mature than their GaN equivalents, SiC substrates and processes still need to improve to meet that goal. This can be done by increasing wafer sizes and reducing defect densities to improve manufacturing yields, especially for the larger chips used in power-management applications.
At the moment the power-management market is dominated by applications in computer servers. Although servers only represent a tiny proportion of the wider computer market, they do consume huge amounts of energy upholding internet infrastructure. Hausken estimates that these servers are responsible for something like 3% of all electricity consumption in the developed world. And with companies such as Google running huge server farms approaching a remarkable 1000 MW in power consumption, it is clear that even a tiny improvement in power management to increase the server efficiency would have a very significant impact.
With the general move towards higher-efficiency products from consumers and through government legislation, it is feasible that SiC components might be able to tap into the considerably wider PC market.
"It is much easier to design a new product into a power supply than into a wireless infrastructure network," Hausken said, illustrating that SiC component makers might be able to make an impact much more quickly than GaN specialists. Although GaN-on-sapphire diodes could also be used in power management, as Velox Semiconductor described in the April issue of Compound Semiconductor, Hausken believes that the market is much more likely to stick with SiC, unless GaN offers a much cheaper alternative. Velox does have the considerable backing of European electronics giant ST Microelectronics, with which it has a distribution agreement, although ST also has its own SiC-based technology for power applications.
But the single biggest problem with all current wide-bandgap device technologies is their very high cost and, when compared with more mature semiconductors, the low quality of available substrate material on which to produce epiwafers. And, as has been the case with LDMOS technology in the wireless sector, silicon is not standing still. One start-up that has just come onto the radar is Qspeed, based in Silicon Valley. Though at an early stage, it already claims to have a silicon-based 600 V power factor correction technology that can "easily replace SiC Schottky diodes at a fraction of the cost".
Wide-bandgap materials still typically sell for a few thousand dollars per 2 inch wafer. While even this expense can be acceptable in the context of a high-value RF system for military or telecommunications use, a serious drop in cost will be necessary for these devices to penetrate the more commoditized power semiconductor applications, and to hold back the challenge presented by companies like Qspeed.
Part of the problem is that, unlike LEDs, where active-layer defect densities as high as 107 cm–2 are tolerable and still allow reasonable wafer yields, electronic devices are not nearly so forgiving.
In this respect, wide-bandgap electronic devices are similar to GaN-based blue laser diodes, which are at least now being fabricated in production volumes, thanks largely to Sony. Sony has shown that it is possible to scale up production, albeit at a high cost and not without problems. Sony's lasers are, of course, used internally, and it may be that this is the best way for wide-bandgap electronic devices to make an impact. Some very big Japanese hitters, such as Panasonic, Toshiba and Toyota, are all investigating the technology and could implement it if they wanted to. It is possible that Eudyna, formed by parts of Fujitsu and Sumitomo Electric, will end up as the major beneficiary if that does happen.
So is Hausken's glass half full? The analyst admits that his estimates are based on a currently very small set of data and should not be regarded as a precise guarantee of the future market growth. "It is a ballpark estimate of the opportunity," is how he put it. "It includes many potential opportunities that may not come to pass." But he adds that the total available market for wide-bandgap electronics is itself a growing one, and is set to reach $1 billion by 2012. A breakthrough in any one application could be the spark that is needed to ignite it.
• Wide-Bandgap Electronics: Technology Trends and Market Forecasts – 2008 is available now from Strategies Unlimited. Visit strategies-u.com for more details.