Okay, I apologize for the bad pun, but it does seem like the Fiber Optic market is finally trending upward. I attended the OFC/NFOEC Conference in Anaheim in March and although a bit late, I’m glad that April came and I can describe some of the developments. I was impressed with many of the technology, product and market developments on display.

At the top level, this market segment has floundered, directionless for the past several years. After good growth in the middle part of the 2000’s, the overall optical market ran into trouble in 2009 as the global economy faltered. The result was a drop in revenue of about 15%. With the exception of a short-lived spike in 2011, the market has been relatively flat. This behavior is understandable, because upgrading, expanding or initially deploying an optical transport network is a capital-intensive exercise and the uncertain direction of many regional economies only adds to the challenge. The wildcard in this scenario however, is the seemingly insatiable desire to consume data.

We tend to focus on growth in the mobile data consumption that is enabled by the vast array of wireless devices we’ve all grown so dependent on, but that’s not the whole story. As I am fond of pointing out, even with mobile data roughly doubling every year from 2009 to 2016, it will only account for roughly 10% of the total in 2016. The other portion of data consumption comes from Internet, high-speed broadband, CATV and enterprise data applications. Even with the backdrop of a still uncertain economy, it is becoming clear that transport network upgrades are essential to ensure future data increases are feasible.

It appears operators are finally committing the capital resources and this looks likely to set up the overall optical market for steady future growth. At the component level, another interesting trend I observed is the rapid conversion to higher capacity networks. It appears 10Gbps systems are the workhorse of the network, with lower capacity systems all but disappearing. Another thing that seemed clear was that while 40Gbps systems will grow, they won’t represent as big an opportunity as originally thought. There are currently four modulation schemes used for these networks: optical duo binary, DPSK, differential quadrature phase-shift keying (DQPSK) and dual-polarization QPSK or DP-QPSK. While this lack of standardization allows suppliers to differentiate their designs, it does not allow for economies of scale and the cost reduction standardization usually entails. At the same time, the cost of 10Gbps systems has been dropping quickly. These factors seem to be leading operators to consider jumping directly from 10Gbps to 100Gbps networks, skipping 40Gbps in the process. It appears from the Marketing presentations and the products I saw, that 100Gbps system will see the fastest growth, by far! For more information on growth rates, segmentation, quantities and market values for optical transceivers, please look at Capacity Requirements Driving Fiber Optic Market.

The other big topic at the show was silicon photonics. This idea seems to divide the audience into two distinct and passionate camps. One says, “we’ve been hearing this now for years, but where are the production products?” The other camp is more than ready to discuss the benefits of silicon CMOS processes for cost and power reduction. One thing that appears to be different from the past is the dramatic increase in enterprise applications and “big data”. As server farms grow, so does the need for fast, cheap, low power, short-reach connections. This need seems very well suited to the capabilities of silicon photonics. While the technology doesn’t appear to meet the needs of every application, there is certainly a lot of development effort and it will be interesting to see how quickly silicon photonic circuits penetrate short-reach enterprise applications. Stay tuned!

-Eric

With the financial reports in the books, it’s time to close out 2012. The good news is that GaAs device revenue closed 2012 up slightly to reach another record at slightly more than $5.3 billion. The small gain was driven by strong fourth quarter performance from the industry after a sub-par third quarter just about erased the gains for the entire year.

Handsets and smartphones, in particular, remain the driving force behind GaAs device revenue growth. The growth of smartphones with their increasing GaAs device content helped propel the handset segment to more than 50% of the entire market. Not surprisingly, the companies associated with handset devices remain the revenue leaders. Skyworks Solutions again saw their revenues increase faster than the market and they remain the largest GaAs device manufacturer, stretching their lead over TriQuint. On the pure-play foundry side, WIN Semiconductors continues their impressive growth trajectory and they have become the dominant company in this segment.

We expect a good uptick in cellular terminal shipments in 2013, along with smartphones continuing to capture market share. I am expecting this will propel GaAs device revenue growth in 2013 into the 8 -10% range. With some of the predictors I use to track the market, I think there are signs that this growth is taking root. For more details, clients of the GaAs service can access my GaAs Device Industry Closes up in 2012 Insight.

However, even with above average growth looking likely in 2013, all is not rosy for the GaAs device market, long-term. The first threat to growth comes from within. The dizzying number of LTE bands, coupled with a desire for the “world-phone” has given rise to the multi-mode, multi-band (MM-MB) PA. This has some serious repercussions, because this market is so price sensitive that it will not tolerate bigger and more costly parts, so these MM-MB PAs must be smaller and cheaper than the PAs they replace or it won’t make sense to use them. We’ve already seen substantial design and design-in activity, so these devices are beginning to see commercial traction.

The other, serious threat was unveiled at the recently concluded Mobile World Congress (MWC). Qualcomm fired the first shot across the bow with their pre-conference announcement of the “RF360”. The company calls this family of devices a complete, all-encompassing CMOS RF front-end subsystem. This subsystem consists of an antenna tuning IC, an envelope tracking (ET) IC for Qualcomm’s PA and a MM-MB CMOS PA fabricated using a silicon-on-insulator (SoI) substrate. This announcement sent stocks of the GaAs PA manufacturers plummeting to levels from which they are still trying to recover. Then at MWC, a whole host of companies announced their ET development efforts aimed at CMOS-based PAs in LTE handset applications. A detailed summary of these announcements and developments is contained in PA Market in Flux: CMOS PAs and Envelope Tracking Emerge as Major Themes at MWC 2013 from Strategy Analytics’ RFWC service.  

These events and particularly the development on the CMOS front will certainly influence the growth trajectory for GaAs devices in the next several years and merit close attention. If you plan to attend IMS2013 in Seattle, stop by the panel session I will be hosting entitled “The Death of GaAs (?)” on Thursday, June 6^th at 12:00PM. We’ve have some market overviews, short presentations from a number of GaAs and silicon-based device manufacturers and then a lively discussion. If you can’t make the IMS2013 conference, you can also catch up with me at CS MANTECH in New Orleans on May 13 – May 16. I’ll be presenting an overview of the 2012 GaAs market and I’d be happy to chat.

-Eric

I just posted theGaAs Five Year Forecast: 2011- 2016on the website and I am happy to report that the GaAs device industry continues to be very resilient. Despite an uncertain direction in the global economy, our research shows the GaAs device market closed 2011 with nearly 6% growth and record revenues of slightly more than $5.2 billion. In fact, our analysis shows the GaAs device market has not declined since 2004 when it stood a shade below $2.4 billion. I’m taking a bit of artistic license with that statement, because we have reported that the market in 2009 “declined” by less than 0.5%, but given the economic meltdown in the US at the time and the nature of the forecasting business: I’ll put that year on the good side of the ledger.

So why is the GaAs market so resilient and less sensitive to economic cycles than other semiconductor technologies (and I’m talking about you silicon)? I think the answer lies in the performance dimension of GaAs technology. We’ve all probably heard the saying “if silicon can do something, it will”. I believe that statement and we’ve certainly seen examples where if silicon-based technologies catch up to the performance of GaAs, the cost advantages make it an easy decision to eliminate GaAs. Keep in mind, among the first applications for GaAs technology were “high-speed” digital logic and where has that market gone? We are seeing SiGe devices in LNA applications and high-frequency transceivers and we are even starting to see CMOS used for handset PAs. The common thread in the application where GaAs is being displaced is a relatively stagnant technology environment. This may be the result of long design cycles, slow upgrade of standards, specifications that remain “good enough” for a long time or a number of other reasons. Where GaAs has proven resilient and risen to the challenge is where the requirements are moving “up and to the right” quickly. As handsets have become more sophisticated with the number of frequency bands increasing quickly, GaAs is still the most capable technology. As Wi-Fi standards evolve to incorporate millimeter wave frequency and multi-gigabit speeds, the displacement of GaAs in this segment does not look quite so certain.

So, why have I gone off on a bit of a tangent? It’s because we are in a period where GaAs will have to show its resiliency once again. We may be looking at a prolonged period of global economic uncertainty that will not help the business models for network capex or consumer spending. For the last couple of years, tremendous growth in smartphone sales have really helped pull the GaAs market along, but saturation is somewhere ahead and growth rates are slowing. In addition, GaAs is still seeing stiff and growing competition from GaN, SiGe, LDMOS and CMOS.

I remain optimistic that the GaAs device market will continue to grow and the report details the effect the trends I’ve mentioned will have on the GaAs bulk and epitaxial substrate market at the very front-end of the GaAs supply chain. The underlying drivers for GaAs growth; data consumption, more GaAs content in handsets, the need for higher capacity wired and wireless networks are still in place. However, I think that the economy is the wildcard. If it doesn’t improve, the growth we see in the next few years may be below historical averages. Now, I’m not a betting man, but despite the apparent gloom, I’m not so sure anyone should bet against the GaAs industry, given the track record of resiliency!

Eric

I recently posted the results and forecasts for the GaAs epitaxial substrate market. The Excel data model is entitled "GaAs Epitaxial Substrates 2011-2016" and the accompanying Forecast and Outlook report is "Markets for Semi-Insulating GaAs Epitaxial Substrates: 2011 - 2016". Our survey results indicate GaAs epitaxial production saw a small (between 2-3%) gain in 2011. This small gain was the result of opposite trends in the two major epitaxial processing techniques, however. As I have been reporting, the demand for pHEMT devices dropped significantly, decreasing by nearly 7% in 2011. The primary reason for this appears to be several large GaAs device manufacturers converting from GaAs to silicon-on-insulator for handset switches. An increase of about 9% in MOCVD processed wafer demand was able to offset the decline in MBE wafer epi and the overall market rose slightly. MOCVD wafers are closely associated with HBT devices used for handset PAs and this underscores the important role these types of devices play in the overall GaAs device market.

Despite the small increase in epitaxial wafer demand, the market revenue grew by almost 20% to nudge jsut past $600 million. The supply chain disruption that spiked pricing in the GaAs bulk substrate market also seemed to have had the same effect on the epitaxial wafers. This price increase is likely a one-time event and epi wafer pricing will return to a more typical price reduction curve starting this year. This, coupled with the slow growth period the entire GaAs market will be in for the next several years will lead to a declining market value through 2016.

The good news is there will be slow growth in device, epi and substrate demand, fueled primarily by handset growth in general and the continued penetration of smartphones. Driven by handset growth, it is not surprising that we believe MOCVD wafer production will increase through the forecast period. What may be surprising is I expect MBE wafer demand to also increase, albeit at a much slower rate than the MOCVD production. I think most of the transition from GaAs to silicon-on-insulator has already taken place and it is unlikely that excess MBE process capacity will remain idle. I think it is far more likely that this MBE and pHEMT capacity will be re-tasked to other high performance markets and this will create the opportunity for some growth.

-Eric

I have just released our latest GaAs bulk substrate market updates. The data model, “SI GaAs Substrates Markets: 2011-2016” and the Forecast and Outlook,“Semi-insulating GaAs Substrate Markets: 2011 – 2016” show that slow growth in the GaAs device market in 2011, coupled with a shift away from GaAs technology for handset switches dropped demand for semi-insulating GaAs bulk substrates by 4%. The other major driver for the GaAs bulk substrate market in 2011 was the devastating earthquake and tsunami that hit Japan. With so much of the supply chain located in this region, the accompanying disruption seems to have raised substrate prices and the overall market value as a result.

This certainly is not how you hope to see revenue growth in a market and this set of circumstances will be short-lived as the Japanese people continue working diligently to restore the country back to “normal”. The supply chain disruptions seem to have largely settled out and I expect the market to revert to slow growth and a more traditional price erosion curve beginning this year. I am optimistic the GaAs bulk substrate market returns to growth in 2012 because GaAs device manufacturers are seeing some growth in revenue and demand and most of the conversion of handset switches to SoI has already taken place.

The survey results indicate demand for GaAs bulk substrates reached slightly more than 32100 ksi in 2011. Even though the demand dropped by 4%, the market value increased by 12% to reach nearly $230 million. This revenue increase appears to be a singular event tied to the earthquake and tsunami aftermath. My forecast indicates demand and revenue will return to slow growth with demand reaching nearly 39000 ksi and revenue growing to slightly more than $240 million by 2016.

Eric

Just back from Montreal and the IMS2012 Conference. There were a lot of interesting products and technical topics being discussed during the week, but I’d like to focus on the breakfast session entitled “Where are the Emerging RF Market Opportunities for GaN?” that we sponsored in conjunction with Microwave Journal. First off, I’d like to thank the participants from Cree, Nitronex, NXP, RFMD, TriQuint and UMS for their time, effort and input. I’d also like to thank the more than 130 attendees that took time from their schedules to hear some of our latest market forecasts and GaN product developments from the participating companies.

GaN market development has benefitted greatly from funding and research derived from the military industry. In our latest GaN market research, GaN Microelectronics Market Update 2010 – 2015, we conclude that military applications will continue to account for the lion’s share of the GaN market. Even with commercial applications starting to emerge, we are forecasting military markets will still account for slightly more than 2/3 of the estimated ~$180 million GaN market in 2015.

The panelists all did a great job of describing actual GaN products developed at their respective companies. They also confirmed that the properties of GaN make it well suited for products addressing EW, radar and communications applications, by highlighting products with some combination of high power, high efficiency and wide bandwidth performance. Of note were S-band 240W transistors from Cree with 60% PAE, 50W GaN-on-Si MMIC amplifiers from Nitronex operating from 0.8 – 2.2 GHz with 55% PAE, a 500- 2500 MHz amplifier from NXP with between 50 and 75W CW output, 50 – 1000 MHz 15W amplifiers from RFMD with more than 60% PAE, 10W PAs from TriQuint that operate from 2 – 18 GHz and 15W X-band PAs from UMS. The examples just mentioned are only the tip of the iceberg, so please visit the websites of these companies to see the depth and breadth of their product offerings and the extent of their GaN development efforts for military applications.

The presentation of commercial products also proved to be quite interesting. Initially, it seemed power conversion products in automotive applications and power amplifiers in infrastructure applications would lead GaN penetration into commercial markets. When we surveyed the industry for the GaN forecast referenced above, it seemed clear these applications were not seeing any significant GaN adoption. However, it appears that the infrastructure situation may be changing quickly as a couple of the panelists mentioned GaN capturing market share from LDMOS in macro and small cell applications. The efficiency and bandwidth performance of GaN seem to be offering enough of an advantage to achieve design wins. It should be noted that not everyone agreed that GaN, even with the performance advantages (that the LDMOS developers are working hard to minimize) was close enough to LDMOS in price to capture significant market share. So, this is certainly a dynamic market segment that we will be watching closely.

In our research, we found GaN amplifiers for CATV applications are seeing significant adoption. Our panelists agree, with most of them mentioning products for these applications. For CATV infrastructure applications, the higher efficiency of GaN-based amplifiers reduces the power consumption and OPEX, by extension. The other dimension for GaN into this market is in green-field applications (primarily), the networks can use fewer amplifiers to maintain the necessary power levels to the users. We anticipate this will be one of the fastest and largest commercial RF applications for GaN in the future.

Thanks, again to everyone who participated and attended this event. While military applications will continue to grow and drive fundamental development, I think we are on the verge of rapidly increasing commercial adoption of GaN. While CATV, infrastructure, Satcom and conversion applications are the likely initial candidates, many other applications are currently under evaluation. Keep an eye on the Strategy Analytics website as we continue to update our forecasts and thoughts on the GaN market!

Eric

In the course of the last month or so, I have had the opportunity to attend CS MANTECH, CTIA Wireless and The Cable Show. Even though these conferences address different industries and different points on the supply chain, it was very interesting to see similar threads running through all three. The conclusion is inescapable: data consumption is the engine that is driving consumer and enterprise devices and networks. These networks, whether wired or wireless, are also becoming increasingly intertwined.

The onslaught of data consumption is not new, Strategy Analytics has been following the dramatic increase for years and most top-level market presentations include some reference to this trend. I think the most succinct explanation of this trend came during a panel session at The Cable Show. Actor, director and writer Edward Burns characterized the current landscape (and I am paraphrasing here) as one of access not ownership. This was fascinating to me because Mr. Burns is not in the wireless or wired broadband industry, but he realizes the importance of the network and how consumers access his craft. It also goes a long way toward explaining the demise of “big-box” electronics and audio stores and the growth of audio and video streaming services. It also bodes well for the continued growth of the broadband industry since the high-speed broadband networks enable access.

On the convergence front, one of the big announcements at The Cable Show was that US MSOs Comcast, Time Warner Cable, Cablevision Systems, Bright House Networks and Cox Communications will allow their broadband subscribers to connect to the Wi-Fi networks of any of the companies in this agreement. This combined network will total more than 50,000 hot spots and is again interesting from several fronts. First, it shows the commitment cable companies have made toward having a wireless component to their networks. This is being done with an eye toward maintaining the “broadband experience” for customers who are nomadic outside of their homes. It also shows the concern about wireless broadband capturing share of the total broadband market.

There were presentations from CEOs of major wireless, cable and device companies that were upbeat about the trends in their respective industries. Universally, the drivers for this optimism were increasing data consumption and the advances in the networks and devices to support this consumption. Most, however, also sounded a cautionary note about spectrum availability for the wireless industry and the most efficient way to increase spectrum for the wired industry. Addressing these issues will provide both opportunities and challenges for device, equipment and network manufacturers and will likely determine the trajectory of future growth.

It has been a very lively past few weeks. It is clear the growth engine in the compound semiconductor industry is still firing on all cylinders and convergence is occurring in all segments of the electronics industry. Please keep an eye out for more detailed summaries of the individual conferences in the coming weeks.

Eric

I had a chance to provide some thoughts on changes in the CATV industry in April's cover story in Microwave Journal (Architecture and Amplifier Device Developments in CATV Networks). For an industry that had a reputation as "static" not that long ago, there are now many evolutionary (and revolutionary) changes underway. Driving these changes is the ever-increasing consumption of data. The CATV network and industry was born out of the need to impove over-the-air television reception, but increasing consumer appetite for more channels, higher definition, video-on-demand and faster internet speed has changed this network into a primary source for all communications needs.

 The CATV network plays a central role in the convergence of voice, video and data into the “triple-play” that cable and telecom network operators are bundling so aggressively to consumers. With video and internet data consumption increasing so dramatically, the traditional coax CATV network has become the HFC (Hybrid Fiber Coax) network with fiber pushing deeper into the network because of the bandwidth advantages it provides. In response to the bandwidth advantages of fiber, cable operators have responded by increasing the bandwidth of their networks, along with channel bonding schemes developed in conjunction with the DOCSIS 3.0 specification. As telecom operators like Verizon and AT&T in the US have paused to consider their fiber to the home strategies, cable operators like Comcast have been only too happy to fill the void with competitive internet and television offerings.

 So what does this mean for compound semiconductors? As is the case with wireless communications, bandwidth is still a precious resource and the need to increase the information contained in a relatively fixed bandwidth allocation means more spectral efficiency and sophistication in devices. These trends at the system level usually are enabled by the performance advantages offered by compound semiconductors. Traditionally, this was an industry that relied on silicon BJTs as the building block for the system amplifiers that boost the signal as it travels over long distances and many splits from a headend where content is added to a consumer premises. As GaAs MMIC technology has matured and the performance has improved, designers have been converting these amplifier building blocks to GaAs MMIC and hybrid technology devices.  

The latest development in the CATV network is adoption of GaN-based amplifier building blocks. The performance characteristics of GaN have long made this a favorite “replacement” technology for power devices, but for a variety of reasons, commercial adoption has been very slow. In our latest GaN market update (GaN Microelectronics Market Update: 2010 - 2015), we’ve found that GaN finally appears to be getting commercial traction in CATV networks. Initially, the thought was that the power performance of GaN-based amplifiers would allow operators to eliminate some of the system amplifiers in a typical network architecture. Operators were reluctant to adopt this idea because they didn’t want to disrupt a working architecture (“if it isn’t broken….don’t fix it”) However, these same operators have been very receptive to the idea that if they keep the network architecture the same and operate the GaN-based amplifiers at the same output conditions as the incumbent GaAs or silicon amplifiers, they realize an energy savings. The higher efficiency of the GaN devices has translated into operating cost savings (electricity), which is also ties in to the "green" initiatives that have become so important for comanies and the environment. This energy saving feature has been the single biggest reason we are beginning to see commercial adoption of GaN. 

Leading the charge are companies like RFMD, Nitronex, TriQuint and ANADIGICS. In public announcements, Nitronex claims they have already shipped more than 200,000 GaN devices for CATV applications and RFMD believes GaN for all applications will account for $15 million of revenue in 2012, with this figure doubling in 2013. They anticipate 25% of their GaN revenue will come from CATV applications in the future. 

There are still challenges to GaN adoption in the CATV market. The reliability concern is diminishing as companies build a history of operation with the technology. The biggest remaining challenge seems to be cost. Our research showed that GaN devices are commanding a price premium of 15-30% over GaAs devices. While this seems counterintuitive in the extremely cost sensitive commercial market, there is not as much price pressure on CATV infrastructure parts as there would be on a mobile handset and the operating expense savings is overriding the acquisition cost increase. Nonetheless, there is still pricing pressure. Even allowing for the premium, the cost of GaN devices appears to be in the $0.50 - $0.60/W range and some GaN foundry companies wonder whether this price reflects the true cost of the devices and if it is sustainable long-term.  

So, the technology that has long been rumored as “about to take off” seems to finally be gaining a foothold in the commercial RF market. However, along with the volume benefits of a commercial market comes the disadvantage of steeper price erosion curves. This is the driver behind the efforts to reduce die sizes for GaN devices and the efforts to develop cheaper GaN-on-silicon (versus SiC) alternatives. 

The dynamics of this segment of the compound semiconductor market are changing all the time, so stay tuned as we at Strategy Analytics stay on top of developments!

 Eric