The GaAs revolution
In the early 1990s, the US led the development of an infrastructure for GaAs MMIC manufacturing. The result: A technology that lies at the heart of the mobile wireless revolution.
BY ERIC HIGHAM FROM STRATEGY ANALYTICS
In the arts and the entertainment business, it can often be said that the more things change, the more they remain the same. That's certainly the case when comparing today with 20 years ago. Back then, Toy Story was a big hit at the box office, Bjork had a top ten album by the name of Post, and The Simpsons was an incredibly popular TV show. Fast-forward to today and there is much excitement surrounding the release of Toy Story 4, Bjork is in the top ten with her album Vulnicara, and many families are still sitting back and relaxing by watching the antics of Bart, Lisa, Homer and Marge (who don't seem to have aged at all!).
Within the GaAs industry, however, there has been monumental, profound change over the last two decades, which span the publication of the very first edition Compound Semiconductor to this twentieth anniversary issue.
The Early DaysThe biggest driver for the advancement of III-V semiconductor technologies has been the US Department of Defense (DoD), which has funded many key programmes to improve the capability and manufacturability of GaAs devices.
When the semiconductor industry got its start in the late 1940s and early 1950s, III-Vs did not play a role, with efforts initially focusing on the development of germanium transistors, a promising alternative to vacuum tubes. However, it did not take long for silicon to replace germanium as the material of choice for high-volume manufacturing.
In the 1980s the DoD funded efforts to improve transistor performance through III-V semiconductor technologies. By the end of that decade, the DoD had funded a successful GaAs Pilot Line programme that sought to develop GaAs digital integrated circuits to compete with silicon. However, the advantages of silicon, such as an established infrastructure and lower cost, could not be denied.
Figure 1: Global GaAs device revenue remained relatively flat in the early 1990s, but took off with the completion of the MMIC and the MAFET programmes.
This led the US government to switch its funding to the refinement of GaAs MESFET technology and the development of high-frequency GaAs amplifiers within the Microwave/Millimeter Wave Monolithic Integrated Circuits (MIMIC) programme.
Running until 1995, the MMIC programme had an incredible level of financial support, with DARPA pumping in an estimated $400 million. Although the initial focus was on microwave and millimetre-wave integrated circuits for defence applications, it also helped develop and refine test-and-measurement, assembly and manufacturing capabilities. Test and manufacturing capabilities were then propelled to a new level with the Microwave Analog Front End Technology (MAFET) programme, concentrated on developing multi-chip module manufacturing capability, improving GaAs device performance, and funding InP technology as a means of extending the operating frequency for solid-state components. While funding for these programmes came from the defence agency, and developments targeted defence applications, there can be no doubt that it is these two programmes that gave birth to the commercial wireless market.
This view is driven home by figures produced by our team at Strategy Analytics, which detail annual global GaAs device market revenue for each year from 1990 to 2015 (see Figure 1). In 1990, at the end of the GaAs Pilot Line programme, total revenue for GaAs was just $250 million, with a significant portion coming from the sales of GaAs-based digital circuits. While the MIMIC programme underpinned refinements to all the manufacturing, test and modeling capabilities, in addition to improving process and product performance, revenue remained fairly flat. However, after establishing this beachhead, GaAs device revenue rocketed from 1995 through to the conclusion of the MAFET programme in 1999, by which time MMICs accounted for the lion's share.
Figure 2: Data by Cisco shows that the ramp in data traffic is expected to continue throughout this decade.
Growth continued after the conclusion of the MAFET programme, surging well past $3 billion in 2000. At that point in time, the entire electronics industry was caught up in the excitement of the "˜dot com' era, with many working on the theory "˜build it and they will come'. The majority of the major roads in the US were disrupted by contractors trenching multi-coloured spools of fibre into the ground, and tech companies throughout the land hoped that the bandwidth capability provided by the new fibre and network deployments would unleash a whole host of new services and benefits. Some of the GaAs products enjoying healthy sales back then were used for either conditioning signals, or deployed in modems and set-top boxes. Components sold included transimpedance amplifiers, RF amplifiers/attenuators and clock and data recovery type circuits.
Unfortunately, the reality was that the dot com bubble burst; operators built networks, but no one came! In response, GaAs revenue dropped about as quickly as it rose, before floundering around at 1999 levels for several years.
The next inflection point began in 2004, a year that marked the start of what we now refer to as the wireless revolution. This kicked-off well after the beginning of mobile telephony in the 1980s − and at its start GSM accounted for about three-quarters of all mobile subscribers, but CDMA, with its higher data rates, was beginning to gain traction in North America. A turning point came in 2005, when operators introduced the first HSDPA (high-speed downlink packet access) networks into the GSM community, igniting a data arms race. Ignoring an almost flat year of 2009, when the global economy threatened to implode, GaAs device revenue has been on an upward trajectory ever since.
Market driversBetween 2004 and 2014 GaAs revenue tripled to more than $7 billion. Early success of mobile communications initially spurred the rapid growth in GaAs sales, and this led on to the evolution of smartphones and increasing data traffic, which, according to Cisco, will grow at a compound annual rate of nearly 28 percent between 2009 and 2019 (see Figure 2).
Although that growth rate is impressive "“ it equates to an eleven-fold rise over a decade - it pales in comparison to the growth rate for mobile data. Analysts at Cisco are predicting that this will shoot up at a compound annual growth rate of 75 percent over the same time frame, equating to an 266-fold increase.
Several factors are behind this tremendous ramp in wireless data traffic. The top of the list is the increasing penetration of smartphones, which, in 2014, we estimate to account for two-thirds of all phone sales. This indicates that the wireless revolution is not only about communication, but entails a whole host of services and applications that users are accessing anywhere, at any time.
This anywhere, anytime connectivity is the engine for the entire RF market. It is welcomed by GaAs chipmakers, as it allows sales of their products to rocket. However, it also brings challenges, the biggest of which is how to build a network that is capable of handling the increase in data that is expected in future.
Figure 3: Mobile data traffic is increasing at a compound annual growth rate of 75 percent, according to Cisco Systems.
Essentially, there are three options for accommodating increases in wireless data. Equipment manufacturers can make radios that are more spectrally efficient; this means more data capability for a given bandwidth. However, the price to be paid for this is the combination of additional complexity of new wireless standards and more sophisticated modulation schemes.
One alternative is for operators to acquire more bandwidth, but this is a costly and potentially time-consuming process "“ and the third option is for operators to "˜densify' networks by adding more, smaller wireless base stations. Benefits of this architecture, commonly called "˜small cells', are fewer users per cell and more frequency reuse, but there are higher capital and operating expenses. Most of the evolving wireless networks are incorporating at least one of these three techniques, as they try to help operators fend off a wireless data tsunami.
Upgrades to data capacity that will occur through the introduction of new network architectures will provide many opportunities for the manufacturers of GaAs devices. They can aim to increase their sales of products that receive cellular signals in wireless base stations, are deployed in backhaul networks that aggregate the data, and featured in wired networks, which transport data to homes or enterprises and distribute this within the premises.
However, the single largest compound semiconductor opportunity comes from consumer cellular devices. Although mobile phones are the dominant wireless device of today, there are other opportunities that come under the "˜cellular' umbrella, which have arisen due to the growing importance of applications such as tablets, e-readers and machine-to-machine communication.
Part of the reason why the cellular terminal market is so important to GaAs chipmakers is the volume that it offers "“ we estimate that 2.4 billion cellular devices were sold last year. That's not the full picture, however, as others factors are also at play in this industry.
Figure 4: Today's typical smartphone contains many amplifiers, filters and switches for several cellular bands, along with Wi-Fi capability. Shown here is a mobile phone architecture for TriQuint, which merged with RFMD at the beginning of 2015 to form Qorvo.
They are associated with efforts by operators to build more capacity into networks, by buying more bandwidth and deploying more LTE networks that offer higher data rates. Today there are more than 40 approved LTE bands that range from 700 MHz to 3.8 GHz, and operators also have to support legacy users on existing 2G/3G networks with different modulation schemes. From the perspective of a carrier, a "˜world-phone' is wanted that accommodates all the regional frequency bands and modes, because this minimizes the number of different phone variants that have to be tracked.
The wide range of frequency bands and modes has prevented the semiconductor industry from introducing a single power amplifier that handles all the requirements. Instead, handsets tend to be equipped with a few amplifiers covering multiple bands and multiple modes. This results in a mobile architecture that contains many amplifiers, filters and switches for several cellular bands, along with Wi-Fi capability. This is good news for the chipmakers, because there is significant RF content in each smartphone (see Figure 4 for a representative block diagram from a leading maker of power amplifiers).
Thanks to the high degree of RF content in the handset, and sales of billions of them every year, revenue from the cellular sector accounts for more than half of all GaAs sales (see Figure 5). And if other wireless applications, such as Wi-Fi, wireless infrastructure, backhaul and VSAT are included, the wireless portion of GaAs device sales can account for 80 percent of total revenue. Meanwhile the military market, dominating GaAs revenue in 1995, now accounts for a tiny proportion of sales.
The handset has been the killer-application for the GaAs power amplifier
Given the heavy dependence of GaAs device revenue on wireless applications, and in particular cellular, it is of no surprise that the largest GaAs device companies are those that are most closely associated with cellular devices. The four biggest device companies in 2014 were Skyworks, RFMD, TriQuint and Avago Technologies; and along with the largest pure-play GaAs foundry, WIN Semiconductors, they accounted for nearly 68 percent of GaAs device revenue. The top four companies for GaAs device revenue have not changed for several years, although there has been a tussle between RFMD and TriQuint for second-place.
The stasis at the top of the GaAs device-manufacturing pyramid belies what appears to be an accelerating trend of consolidation. This is seen in the "˜merger of equals' that took place earlier this year: The marriage of RFMD and TriQuint to form Qorvo. A very top-heavy industry has resulted, and if 2014 revenue from TriQuint and RFMD is attributed to Qorvo, then between this firm and Skyworks, they account for 55 percent of last year's GaAs device sales.
Additional consolidation in the GaAs device industry includes: the acquisition of long-time industry stalwart, Hittite Microwave, by Analog Devices; and Avago Technologies expanding its presence by buying LSI Logic, and aiming to add Broadcom "“ the later is waiting for approval. Although the Avago activity will not make a big impact on its GaAs activity, these moves offer an interesting glimpse into the direction of that company. While it may appear that acquisition and consolidation is bigger than ever, it has been a staple of the GaAs industry since its inception. This point is illustrated in Figure 6, which tracks the history of what were the seven largest publically traded GaAs device manufacturers. This chart reinforces the idea that GaAs development activities existed prior to the 1990s, and it also shows how the main players in the industry consolidated to acquire capability and reach scale as commercial applications have grown.
Figure 5: The majority of GaAs device revenue comes from the cellular sector.
Those within this industry are now working feverishly to make the promise of 5G a reality. This communication standard is not well defined, but it promises to allow us to enter a mobile era where "˜any place' and "˜any time' are joined by a third attribute − "˜any amount'.
History attests, in the form of the GaAs market of 1995, that it is important to continue developing manufacturing, test, process and design capabilities to realize the vision of 5G. What is encouraging for GaAs is that the network requirements that are under discussion seem to line up well with the performance advantages of this technology, so the emergence of 5G could propel chip sales to a new level.
During the lifetime of Compound Semiconductor magazine, the GaAs device evolved into a mature, dominant technology for most RF applications. However, this industry must not rest on its laurels. There will always be challengers to GaAs device revenue, and currently they are coming from various directions, due to developments associated with GaN, SiGe, InP and silicon CMOS technologies.
The introduction of new system architectures and performance requirements will provide challenges and opportunities for the industry, but the future looks very bright. For that reason, I look forward to future developments in the GaAs industry, along with continuing excellent, insightful coverage provided by Compound Semiconductor.
Figure 6: Since the turn of the millennium, there has been significant consolidation in the GaAs Industry.
