Analysts are tipping tremendous growth for the compound semiconductor industry

In some ways, all our purchasing habits are very similar; but in others ways they are markedly different. Nearly all of us only buy and fit a new light bulb when the existing one fails. But when it comes to mobile devices, some of us upgrade very frequently, others make do with the same model for a couple of years, and there are also those that cart around handsets that they bought in the previous decade. For the compound semiconductor industry, the implications of these differences in purchasing behaviour are massive. They explain why the LED industry is tipped to grow massively over the next few years, but could go into reverse around 2020, while sales of chips for mobile devices could steadily rise for many, many years. These insights, along with revenue predictions for LEDs, GaAs ICs and a host of other devices, were provided by some of our industry’s leading analysts at the second CS Europe conference. This two-day gathering – attracting almost 250 delegates to the Hilton in Frankfurt, Germany – kicked-off with an overview of the entire compound semiconductor industry by Asif Anwar of Strategy Analytics. Greater detail on the LED industry followed in a talk by Philip Smallwood from IMS Research, while very positive predictions for the wide bandgap market came from a presentation from Yole Développement’s Philippe Roussel. Anwar delivered a very promising message, which he summed up in just one sentence: “Compound semiconductors are, and will remain, a key enabling technology across wireless, consumer, automotive and defence sectors.” At the beginning of his talk, he honed in on RF GaAs devices, which are a key component in many handsets, including smartphones. They are also used for amplification in Wi-Fi networks and will soon enable communication between one machine and another. The market for the GaAs chips used in these established and emerging applications is fairly buoyant, having grown by more than 30 percent between 2009 and 2010. Although 2012 will be a ‘challenging year’, according to Anwar, total device revenue should rise towards $6 billion by 2015. Wireless traffic is increasing at a tremendous rate. To accommodate for this, there is a build-out of faster networks involving base-stations, back-haul links, fibre networks and broadband systems. To vastly increase capacity, transmission date rates through fibre are rising from 10 Gbit/s to 40 Gbit/s and then 100 Gbit/s, generating sales for photodiodes and various forms of laser. Anwar also touched on cable TV, which can also be used by many homeowners to provide Internet access. Here, compound semiconductor devices are vying with silicon rivals for deployment in amplification systems. “What we’ve seen is a move away from silicon to the IIIVs,” said Anwar, who also spoke briefly about automotive radar systems, which have traditionally been deployed in premium vehicles, but are now starting to penetrate the mid-price car market. In defense of GaN Predictions for wide bandgap electronic components also featured in Anwar’s presentation. According to him, most GaN devices sales are for military applications. Devices built with this material – which combine high operating voltages with high switching frequencies and excellent reliability – are now being deployed in: Electronic warfare systems, which produce broadband, high-power emission to disrupt and jam RF signals used to detonate improvised explosive devices; active electronically scanned radar based on hundreds or thousands of transmit and receive modules, which can be phase shifted to form and steer a beam; and radios that can operate over a frequency range of 30-3000 MHz and an output power of up to 100 W. In Anwar’s opinion, the main opportunities for the other popular wide bandgap semiconductor, SiC, are in the power electronics market. These devices could be used in the IT and consumer sectors in uninterruptible power supplies and power factor correction power supplies, and they could also find deployment in the automotive sector, enabling DC-to-AC and DC-to-DC conversion in hybrid electric vehicles. In addition, they could find application in electronic systems operating at far higher powers, which are needed in photovoltaic systems and wind turbines. One area where silicon will continue to reign is in the photovoltaic market. However, Anwar believes that concentrating photovoltaic systems built with multijunction solar cells can be competitive in climates where the sun beats down very hard and the air is dry. This technology is already starting to have some success in south-western regions of the US, and Anwar tips installations to hit 2 GW in 2015, equating to 4 percent of the photovoltaic market. In the illumination market compound semiconductor chips are already generating billions and billions of dollars. Backlighting handset keypads and screens was the first killer application for the LED, and in the last few years similar revenues have come from backlighting larger screens, such as those found in laptops and TVs. “Ultimately, the aim is to replace the incandescents in your house with an LED technology,” said Anwar. That topic was covered in far more detail by Philip Smallwood from IMS Research. He explained that the LED lighting industry is only just starting to make an impact in a market dominated by compact fluorescents (CFLs) and incandescents. “In 2011, LEDs had a 1 percent penetration, in terms of units, but a 14 percent penetration in terms of dollars.” LED lighting is tipped to grow fast (see Table 1 for an overview of its strengths), and by 2016 it is expected to overtake backlighting, in terms of revenue, to become the biggest market for this chip. However, at that stage LED lighting will still only account for about a quarter of all LED shipments. Fast forward another two years to 2018 and sales of light bulbs based on LEDs will have grown to a staggering 6.3 billion units, overtaking incandescents, which will have a unit share of just 14 percent by the end of this decade.    Table 1. According to IMS, the strengths of LED light bulbs include: The absence of mercury and lead; 40,000 hour lifetimes; great robustness; and no cold-starting issues. Credit: IMS Research The high price of LED bulbs means that they will bring in more revenue than another other lighting technology in 2013. According to Smallwood, sales in that year could net $30 billion. This will drive a compound annual growth rate (CAGR) in the lighting market of 85 percent through to 2016. Between 2009 and 2020 shipments and revenue from LEDs employed for lighting will increase at a CAGR of 53 percent and 26 percent, respectively, but towards the end of this timeframe this market will start to stagnate. Uptake of LED lamps in the lighting sector will have a profound affect on the lighting business. IMS Research predicts that total units shipments will start declining in 2017, and revenues will go into reverse the year after, due to the very long lifetime of the LED bulb(see Figure 1).

Figure 1. Growth of the LED lamp market will go into decline in 2014 and halt in 2017, according to IMS Research. Credit: IMS Research The good news for consumers is that the prices of LED bulbs will tumble over the next few years (see Figure 2). In 2014 their average price should drop below $10, a major milestone from the perspective of many US consumers. Smallwood says that three years’ later, the average price should nudge below $5, and he predicts it will retail for just $3.59 in 2020. Figure 2. The average selling price of LED lamps will tumble during the next few years. Credit: IMS Research During the next few years Japan will account for the greatest deployment of LED lamps and luminaires. This will be fuelled by the very high cost of energy in this country, which has been driven up by the impact of the Fukushima nuclear reactor damage. But by 2018, North America will overtake Japan. The residential market will be the hardest one to penetrate for LED lighting. “Consumer’s don’t do returnon- investment calculations,” said Smallwood. “They go to the supermarket and pick out the cheapest bulb. They’ve been doing that for the last hundred years.” But households will start to buy LED bulbs when prices fall. By 2020, LEDs will be found in four-fifths of sockets in developed regions, with incandescents making up just 2 percent. In offices, fluorescents could still be widely used, thanks to their combination of high efficiency, long life time and low cost. In developing regions a slightly different picture will emerge. Although LEDs will be found in 44 percent of sockets by 2020, for the next few years the CFL will remain the most popular energy efficient lighting product, due to its lower price. Forecasts for LED lighting have been translated into LED die shipment predictions. According to Smallwood, shipments of packaged LEDs for lighting will hit about 3 billion die this year, and will rocket to more than 30 billion by 2017. The role of SiC Analyst Philippe Roussel provided a detailed analysis of the current state of the wide bandgap market, offering some predictions for the coming years. According to him, the wide bandgap market will be worth almost $100 million in 2012, rising to nearly $3 billion by the end of this decade. This rapid rise in sales will occur because of two highly valued characteristics found in all wide bandgap devices: A high junction temperature and a high electron mobility. Thanks to these twin strengths, devices have no recovery time during switching, leading to low losses and high switching frequencies. And at the system level that means lower cooling demands, fewer filters and ultimately a unit that can be lighter and smaller. Efficiency gains also result from replacing silicon devices with those made from wide bandgap materials. According to Roussel, switching from the incumbent technology to GaN or SiC can increase the efficiency in DC-to-DC conversion from 85 percent to 95 percent; boost the efficiency of AC-to-DC conversion from 85 percent to 90 percent; and propel the efficiency of DCto- AC conversion from 96 percent to 99 percent. Thanks in part to these efficiency gains, Roussel believes that we are starting to enter a new era for power electronics devices. Thyristors and MOSFETs reigned supreme from the 1970s to the 1990s, IGBTs have been the dominant technology for the past two decades, but SiC and GaN devices are an everincreasing threat to their ascendancy (see Figure 3 for details of product releases). Table 2: Many countries are using legislation to phase out the purchasing of incandescents. Credit: IMS Research The silicon-dominated power electronics market is currently valued at $17.7 billion, and is predicted to increase at a CAGR of almost 100 percent throughout this decade, to hit $35.7 billion in 2020. Today, more Figure 3. SiC diodes appeared on the market before any form of transistor,with Infineon and Cree leading the charge. Credit: Yole Développement than half of these devices are produced on 6-inch silicon, but by the end of the decade the 200 mm platform will be the most popular one for manufacturing.

Table 3: It can take many years to develop a new GaN device and ramp its production to high levels. Credit: Yole Développement Two-thirds of the power electronics market is associated with devices operating at up to 900 V, and Roussel believes that in this sector GaN is the most attractive alternative to silicon. At voltages beyond 1.2 kV, however, he believes that SiC can be more competitive than GaN. Commercial interest in wide bandgap devices has recently stepped up a gear. Funding for the likes of Transphorm, CamGaN, EpiGaN and Azzurro, plus the acquisition of Velox by Power Integration, has led to a $100 million injection of funds into GaN development through merges, acquisitions, investments and fund raising. What’s more, nearly all of the top twenty power semiconductor manufacturers are involved in developing either GaN or SiC – and the top two, Infineon and Toshiba, are pursuing both. All this effort will spur the GaN market from $13 million in 2012 to $1.9 billion in 2020, according to Roussel. Figure 3. SiC diodes appeared on the market before any form of transistor,with Infineon and Cree leading the charge. Credit: Yole Développement This analyst also considered the possible impact of overcapacity in the LED market on the wide bandgap electronic sector. He considered the situation where companies with MOCVD tools for GaN LEDs and a weak order book could consider diversifying into power electronics. This would not be easy, warned Roussel: “It’s certainly not as simple as copying and pasting”. He believes that – depending on access to intellectual property and patents - it could take up to two years to master the production of GaN-on-silicon epiwafers, and a further two-to-three year to develop good device designs. Meanwhile, SiC device revenues will grow, although not at quite the same pace as their wide bandgap rival. Roussel estimates that the SiC market for this year will be worth $70.6 million, rising to almost $950 million by 2020. This means that by the end of this decade, the compound semiconductor industry could have multiple billion dollar markets, including those for GaN power devices and SiC power devices. That doesn’t mean that there will not be painful times ahead – overcapacity in the LED market is likely to lead to some tough times over the next few years – but the long-term future for the compound semiconductor industry is a bright one. Table 4: The nascent GaN power electronics industry features some epiwafers providers, some fabless companies, and some players that produce their devices in-house.Yole Développement