The RF Semis Market is about to undergo some dramatic shifts. For the past two years I have been neck deep in that market, but am no longer employed there. So I want to put my thoughts about the market into writing. These are my opinions, they do not reflect those of my previous employer, and they are based on public information only.(It has been a while since I was involved in day to day of operations.) Some of this may look familiar from the public transcripts I helped write, and some is just based on my estimate on the strategic shift in the landscape.
This site looks at a broad swathe of technology. I think the audience can be divided between those who know about RF and the vast majority of the population who do not. So first, let me explain a bit about why I think the market matters, even to those who did not know that RF stood for Radio Frequency. For those in the market, you can skip a couple paragraphs down, or hopefully be able to cut and paste those paragraphs and use them to explain to your in-laws what you do.
When we talk about things being ‘mobile’ or ‘wireless’, we mean they are connected to the Internet using a radio. Yes, I mean like the radio you listen to in your car. Half the history of communications technology over the past fifty years is the story about how we made those radios much more useful. Today, the radio in your smartphone is extremely complicated. Think of your car radio as a paper airplane, and a cell phone radio as the space shuttle, or the Starship Enterprise. There are two sides of this story. First, the industry had to find a way to cram huge amounts of information into some coding scheme. That is the story of getting very powerful computers to fit inside a phone, and is exemplified by the rise of Qualcomm. That is a pretty amazing feat, given how much data we are now cramming into the airwaves. But there is another part of the story, and that is the way we manipulate the physical radio waves, bridging the digital computer of a Qualcomm chip and the physical world of radio waves. This area has been marked by some truly incredible advances, that fifty years ago would have been thought of as defying the laws of physics (sortof). This is the story of RF chips, and today’s vendors do some pretty incredible things. We need both sides of this to work, and the major transitions of cell phone technology of the last 20 years (2G, 3G, 4G, etc.) have occasionally been held up by bottlenecks on one side or the other.
I want to make clear that both sides need to work and advance. Without the achievements of the RF semis makers we would still be wire-bound. And the coming years promise to really push the technological envelope further. The shape of the market’s participants will likely have a very real impact on the rate of progress we can expect from our wireless networks.
Market Dynamics [RF people can start reading here]
Over the last ten years this space has seen tremendous changes. When cell phones became a mass market product in the 90’s (thank you Motorola and Nokia) a large number of component suppliers appeared to support the handsets’ RF needs. There were something like 30 vendors in 2002.
Today, there are effectively three and a half left. The story of RF semiconductors is like the story of all forms of semiconductors. Big vendors slowly get bigger, their chips get smaller and they add on functions. Vendors who cannot keep up, who fail to integrate the parts that sit next to them get acquired or exit the business. This process is cyclical, with waves of growth followed by waves of consolidation, and you can track these cycles back 20 years. In the consolidation phase, the big vendors grow and become more profitable. We are at that stage now. It has become fashionable on the Street to imagine that the cycles are now over, but I think that is unlikely, for reasons I discuss below. The cycle will continue and gravity will re-assert itself. The flip side of this progress is that there is now very little room left for small vendors. Some of this is ‘good’ in that it makes the whole industry more profitable, but it is also decidedly bad (no quotes, just flat out bad) because it is now very hard for anyone to bring a disruptive technology to market.
As I said above, there are three and half vendors left. In alphabetical order, they are Murata of Japan, Qorvo (the merger of RF MicroDevices and Triquint) and Skyworks. The half is Avago, which I will explain in a moment. In the past 18 months, the last of the small companies were sold – Peregrine, Amalfi, Black Sand and Javelin. Acco also seems to have moved in a different direction. There a couple of small, private entrants pushing the technology envelope, and a handful of domestic suppliers at the low-end of the market in China. And as far as I know, that is pretty much it. Oh, yes, there is also one more entrant, coming to the market someday – Qualcomm the industry giant. I will get to them too.
When I segment the market, I think there are three basic product groups. And when I say there are only three vendors left, I mean there are only three left who can supply all three products. That is what happens with integration, there used to be a dozen different products involved, but to survive today at scale, companies need to supply all three. The three products are:
- Power Amplifiers
- Packaging (aka integration, modules, or just plain solid operations)
[To be clear, I am simplifying greatly, there are many other parts involved, but these are the big profit pools left.]
All three companies can do all of these, but each has their strengths and weaknesses. Let me walk through them now.
Murata – Murata is a poorly understood company, at least poorly understood by US investors. It is one of the most dynamic electronics companies in Japan, and really deserves to be compared not with its domestic peers but with the global group. They have a very strong filter business, completely dominating certain categories. The filter business is very much driven by economies of scale. Murata has a very strong manufacturing operation, that is now reinforced by its sheer size. They are weakest in Power Amplifiers. They acquired a small, legacy PA business four years ago from Renasas. I would argue that they now have a potentially very strong PA business which they acquired with Peregrine. But as that company’s public comments say, it will still be a while before that PA is shipping in meaningful volume. Overall, I think the market greatly underestimates Murata’s position and prospects.
Qorvo – First, a warning. It is almost impossible for me to be objective (or even polite) about Qorvo. I will not get into the reasons for that. You can probably work it out. So take what I say with a grain of salt, and I will do my best. Qorvo deservedly has a reputation for making some of the best products on the market. They are also very good about winning leading share at the biggest customers (Nokia and Motorola then, Apple and Samsung now). With the acquisition of Triquint, they also now have some filter capabilities. They have struggled with the operations piece. This is not really a product, but just a reflection of how well a company is run and how it gets its products to market. RFMD had a history of struggling to allocate capital in a wholly efficient manner (there, that sounded polite). And as a result, they have occasionally run into serious hurdles hurting their growth and/or profitability. Right now, they have a great tailwind in the form of merger synergies and cost-savings through the removal of redundancies between legacy RFMD and legacy Triquint. However, I would argue that at some point they will have to make some hard choices about how to integrate the two companies. If history is a guide, they may not got all the answers right on that quiz. But let me close by repeating, they make some great products.
Skyworks – In my opinion, Skyworks has one of the best-run operations in the industry. They are ruthlessly efficient in driving cost out of packaging. They have solid Power Amplifiers, arguably not as good technically as Qorvo’s but still solid. They have also been much more efficient about capital allocation which has given them the opportunity to attack much more business than anyone else. Skyworks is weakest when it comes to filters. The only filter capacity they have comes from a joint venture they run with Panasonic. This is a small-volume operation, and as I noted above, filters are very much about economies of scale. They also have a relationship with a very small company called Resonant. Despite having essentially no revenue, Resonant is public, and valued by the Street entirely on its ability to provide a disruptive filter technology to Skyworks. If Resonant can do what they claim, this would be a big boon to Skyworks. (Leading to the assumption that they would then acquire Resonant.) I am watching this development closely.
That leaves Avago. First, let me say that I think very highly of Avago. They are a very strong company, with tremendous RF semis revenue. When I say I only count them as being half in the market, I am referring to my view that they are cherry picking only a small subset of business. They are one of only two companies that manufacture a special flavor of filters know as BAW or FBAR. These filters are much in demand for certain challenging applications. The other company in the space is the Triquint side of Qorvo, but this business is much smaller than Avago’s operations. As a result of this elder brother role in a tight duopoly, Avago has been able to make a lot of money. Companies that need BAW filters (e.g. Apple, Samsung) are willing to let Avago bundle other products with those filters (a module with a PA and the filters). But Avago is very much bucking the trend towards integration demonstrated by the rest of the industry. They are not a one-stop shop, and show no sign of wanting to become one. They enjoy the best profit margins in the industry as a result. However, I really question how much longer Avago will be in the RF business. I regularly predict that some day they will sell their wireless business. Someday, someone will feel that they really need to own their own BAW filter capacity. When I make this prediction, people treat me like I am crazy, but Avago’s management are some of the smartest deal people around. They know the value of their assets, and show no sentimental attachment to them. I do not often make definitive predictions, but I think within five years Avago will get a very nice price for their wireless business. Until then, they will continue to sell in their large, profitable niche.
Let me round out the entire bestiary of RF companies. There is one other sizable filter vendor left – TDK Epcos of Japan competes with Murata but has struggled in recent years to break into the smartphone market, and thus do not have the volume to scale meaningfully. They are out there fighting away, but I imagine ultimately they will sell the business to someone who needs filters. There is also a private company known as IO Semi, that is working on some interesting parts of the component stack. They have the advantage of being small and focused, which lets them push the technology envelope. It is unclear how their future will play out. Finally, in China there are about a half dozen small PA vendors. Many of these companies spun out of RFMD, in the sense that during a restructuring a few years ago, many of the engineers who were let go set up shop on their own, supplying the China OEMs. RDA of Shanghai had a fairly sizable PA business also, but it is unclear how well this is faring inside of Spreadtrum, or whatever that bundle of companies is called now.
I should probably mention at the outset of all this, that I am writing about the RF market for consumer devices. The market for carrier products like base station RF is very different.
The whole theme underpinning this industry is the trend towards greater integration of all the components. Single Band Power Amplifiers (PAs) have become PA modules, and are well down the path to becoming monolithic multi-band PAs. The next big shift will come as the industry moves from Gallium Arsenide (GaAs) parts to chips made in standard silicon processes. (Technically, these are silicon on insulator or SOI, but that over complicates the underlying reality.)
Peregrine built itself as the pioneer on SOI technology, and is the first company to demonstrate a working silicon PA that offers the same performance as GaAs PAs. In theory, that will give Murata an advantage, depending on when they finalize the commercial product.
Eventually, the whole industry will find some way to make the same move. And whether other companies develop their own SOI processes or license Peregrine’s, the benefits of silicon are immense.
Silicon has two key, closely related advantages – one technical and one economic. The technical advantage is that silicon readily allows the addition of other functions into the RF silicon product. Historically, RF components were largely analog machines, they took a physical signal, and then handled it in some pre-determined way. Using silicon, it is much easier to embed a little bit of memory and a little bit of digital logic, giving the RF semi some ‘intelligence’. This opens up all kinds of doors, and most likely will lead to systems that allow the other components in the phone to ‘communicate’ or ‘control’ the RF semi. Someday, this will bring about tremendous gains in RF performance in the form of better reception and better battery life.
The economic advantage is that using silicon means companies can take advantage of the huge silicon foundry industry. The shorthand way to describe this is to say that by using silicon, RF component vendors can take advantage of Moore’s Law and the steady improvement in chips that has brought about. The actual details of this are a bit more complicated, but the reality is no less powerful. RF component vendors using silicon will get a big boost in performance just by participating in the global silicon ecosystem. This has played out many times over history in other markets, and there is no reason to expect RF semis to be any different. When silicon can match the performance of some more exotic material, silicon quickly drives out the other material. In RF, we have not yet gotten to the point where commercial silicon PAs are available which can match that of the incumbent Gallium Arsenide (GaAs) Pas, but it is clear that GaAs’s days are numbered.
The Big Q?
All of this us brings us to what could be the biggest change to the RF market this decade. When more of the RF semis of the phone can be produced in silicon, they will get integrated into the other silicon components of the phone. But when we talk about ‘integration’, that really means one thing – the companies that make the key digital chips in the phone (aka basebands) will find a way to own the RF stack as well. This has already happened with another half dozen parts, and RF is one of the last left standing.
Which brings us to Qualcomm. Two years ago, Qualcomm announced that they were entering the RF market. Qualcomm is by far the market share leader in advanced basebands. They have something like 95% share of the market for LTE (4G) basebands, and a very large share of 3G as well. If they can produce a working RF product, their control of other parts of the phone will give them an immense advantage in determining which RF products get built into devices. So Qualcomm’s entry, in theory, means the other RF components vendors face a bleak future.
Then something strange happened. To date, Qualcomm has not yet shipped a working PA. They have a couple other RF parts, but they are missing the key elements. As I write this, I expect Qualcomm to make some big pre-MWC announcement about RF, but at this stage, they are far behind. Their RF product, which they call RF 360, is so far behind expectations that the Street no longer thinks Qualcomm can ever deliver. I think that view is misplaced. Qualcomm has the brains and the resources to eventually get RF 360 shipping, but its slow start has made the industry aware of the company’s vulnerabilities.
Even when they can build a silicon PA, they are going to need filters. Filters are not semiconductors, they are essentially ceramic parts, that require the kind of manufacturing expertise that Qualcomm cannot build internally. So they are going to have to partner with someone to get filter capacity.
Also, the small pieces of RF 360 which have shipped have gotten very poor reviews from the industry. This has made handset OEMs very nervous, as they are worried about having no choice but to use poorly-performing Qualcomm RF parts. Or worse, being forced to use some Qualcomm partner’s GaAs parts, delaying for years the potential benefits of silicon RF.
Finally, when Qualcomm originally launched RF 360, investors hoped that this would open up a new market for the company, extending and protecting their core baseband market. Instead, RF 360 has had to rely on the baseband business to protect it and build its toehold. It is unclear how long it will take for this situation to be reversed.
So if the whole history of the RF market is one of steady, cyclical moves towards consolidation. And we are now poised for a major technology transition to silicon which tilts the board heavily towards further integration. Does this mean the end is near for makers of RF components?
First, I think there are more mergers in store. I am speaking solely from my view of the market, I certainly have no inside information that anything is pending. My guess is that we need a few more catalysts to really get things going. In practical terms, the next move is really Qualcomm’s. If we can leave Barcelona with tangible signs that RF 360 has made technical advances, and can expect a commercial product in the next year, then that will set the course for the rest of the industry as everyone scrambles to respond. On the other hand, if they announce that they are partnering with an incumbent vendor, then the industry can breathe a sigh of temporary relief. The last vendors will have time to develop their products and integrate their acquisitions fully.
Regardless of where they stand with their silicon products, Qualcomm will still need to source filters and packaging from someone, and there are few choices left. Whoever that supplier is, it will get a big boost from an attachment to the leading baseband vendor. But it will the kind of partnership where both sides are always wary of the other. The filter vendor will have to face an eventual retreat from their PA business, as Qualcomm slowly gets that part working. And Qualcomm itself will have to watch as its filter partner supplies the other baseband vendors like Mediatek and Spreadtrum, who likely have their own RF ambitions.
Set behind all of this are ongoing changes to the smartphone component market. The past few years have seen tremendous growth in RF content in phones. The transition to 4G has required much more complex, expensive RF components. For the next few years, the big growth in 4G devices will come at the bottom end of the price spectrum, as operators deploy 4G networks in markets that emphasize low-cost, simpler devices. This will mean a steady erosion in average prices for RF parts and a return to less heady times for the RF vendors. We saw this with 2G and 3G as well, the cycle will re-assert itself.
The one exception to this may be Apple. The iPhone has been a huge driver of RF dollars in recent years. The latest devices have very complex RF circuitry, but this will likely peak at some point. Apple, and all the handset OEMs, want to see more integrated RF components, as it will save them money and space inside the phone. I have no idea if this will happen in this year’s iPhone or next year’s, but it will happen at some point.
I recently wrote a note about the positives and minuses of being an Apple supplier. That note may never see the light of day, but suffice it to say, that Apple pushes all of its suppliers hard. If the past year or two have seemed like glory days for the RF industry… well, glory fades. There are still a lot of people competing for Apple sockets, and eventually that will take its toll on industry dynamics.
In the past decade, we have gone from roughly 30 RF component vendors down to roughly a half dozen. All indications are that five years from now, there will be even fewer.