In search of distractions from the real world, we have lately been doing a lot of work around RF semiconductors. This is a perennial topic here at Digits to Dollars, 3 of our top 10 ever-viewed posts have touched on the subject. And a lot has been happening in RF lately.
RF, or Radio Frequency, Semis are on one of the key subsystems inside mobile phones. They bridge the divide between analog radio waves that reach the phone’s antenna and the digital systems inside the phone that decode the signals into data. Chips that deal with real world, analog, signals are a special domain of semis. They have to contend with real world conditions, not merely moving around 1’s and 0’s. Moore’s Law does not work the same way in RF.
This makes them harder to engineer. RF engineers are a distinct, proud subset of chip designers, practicing a degree of Black Magic. This has sheltered the industry. Modern phones simply would not work without all the clever engineering, creating serious competitive moats.
Key concepts for the industry include:
- RF systems contain dozens of components grouped into a half dozen or so packages. Packaging is thus an important element of the industry, different from other semis.
- The scarcest components today are high grade filters. These are not chips, they are highly precisioned pieces of ceramics. These are very hard to manufacture at scale but crucial in all modern (4G and beyond) phones.
- As a result of the above, there is a lot of interchange among the suppliers with vendors selling components to be packaged in modules sold by their ostensible competitors.
State of the Industry
Today there are four major RF incumbents, one massive new entrant, some new technologies over the horizon, and something going on in China. The players today:
Broadcom has one of the best franchises in electronics. They are the leading provider of advanced filters (Fbar or BAW). Every high-end mobile phone, and many not-so-high-end phone, needs some form of these. There are essentially two vendors for BAW today (Qorvo below is the other), and Broadcom does it better, and more profitably, than anyone. This has given them monopoly-like profits. They have chosen to focus on this (multi-billion dollar) niche rather than expand their footprint into other lower margin RF segments.
Murata of Japan is a leading supplier of ‘passive’ components. Their strength rests in their incredible manufacturing capabilities, especially for a wide class of ‘commodity’ filters (SAW) for which they enjoy significant economies of scale. They also produce some of the most advanced ‘active’ components through their acquisition of Peregrine Semis six years ago. Murata logos may not appear in all the phone teardowns we see, but their components likely appear in some package in every phone.
Qorvo and Skyworks of the US are the traditional leading incumbents in RF. These companies survived twenty+ years of industry consolidation and command a large market share especially for the active components in a phone (i.e. everything other than filters). Qorvo has some advanced filter (BAW) capabilities they acquired ten years ago (via Triqunit), but have been underperforming on multiple fronts in recent years. Skyworks is a well-run company with excellent operations (especially around packaging), but suffer from a lack of filter capacity. They have been building this out, but it has taken a long time with little to show so far.
Those are the four incumbents. The massive giant is Qualcomm. Qualcomm makes all the other important chips in a phone (except screens and memory). Over their history they have steadily expanded their footprint inside phones expanding their product line to integrate all these other subsystems. RF has been the one hold out.
Their position in other parts of the phone theoretically gives them a big advantage – by tying the RF chips to the digital parts of the phone a complete system could offer tremendous performance gains. So almost ten years ago Qualcomm announced they were entering the RF market. At the time, Qualcomm announced an ambitious system called RF 360 which tied together all of their components. Qualcomm’s size was enough to scare the market, but the vision behind RF 360 was seen an existential threat to the other vendors.
Then RF 360 remained just that – a vision. The Holy Grail of mobile semiconductors, never quite achievable. Qualcomm promoted their RF efforts publicly and made several acquisitions including Epcos, the RF business of Japan’s TDK . The Epcos acquisition was also scary for the industry as Epcos has a sizable SAW filter business and some prospects for BAW.
Despite this, Qualcomm struggled. They had technical problems, RF 360 when announced, was not far beyond the PowerPoint stage. They had to master a whole new set of operational and manufacturing skills that they lacked. And mostly they just struggled with integration and organizational problems.
Every year, at the Mobile World Congress trade show (MWC) Qualcomm would announce design wins for their RF products. But when pressed, the handset vendors awarding those designs would all say privately that they were strongly “encouraged” to grant those wins (and that word is doing a lot of work in this context). Put simply, Qualcomm’s RF products could not stand on their own.
That being said, over the past year, the situation has changed dramatically. Qualcomm now has a real RF business. There were signs of this in MWC 2019, and we suspect one of our big takeaways from this year’s show, had it not been cancelled, would be this emergence.
There are two important things happening within Qualcomm’s RF business now. First, their basic RF products are (finally) standing on their own, generating meaningful revenue. They have put their house in order, with a unified management structure and products that work. Their products may not be the best RF components out there, but they are well beyond the threshold of ‘good enough’ and can compete in many categories.
The second big development is the advent of 5G. As we keep repeating, 5G really comes in two phases. The first phase is what we have today, and it looks a lot like 4G. For phones, the RF components are largely similar, and so the industry bears the same shape for 5G as it does for 4G. However, the next phase – mm Wave – looks very different from an RF perspective.
Qualcomm has made a big bet on this transition. In particular, they are now promoting a unified RF system for mmWave that ties the RF chain to the digital components of the phone – which is exactly what six paragraphs up we described as the Holy Grail of RF. Qualcomm’s long-term strategy has always revolved around mastering the transition from one G to the next. So it is no surprise that they are focusing their RF efforts on mmWave. And while we are skeptical about the timing on mmWave, Qualcomm’s system can probably be re-purposed to cover whatever stage of 5G the market glides through over the next few years.
This will eventually be a big problem for the incumbents, especially Qorvo and Skyworks. They will not be driven out of business entirely any time soon, but there is definitely a clock ticking around the size of their market.
Beyond Qualcomm’s innovation around the digital integration of the RF chain, there are a few other new technologies emerging that could shake up the industry further.
Foremost among these are the exploration of new filter technologies. Several companies are looking at ways to improve the performance of less expensive filters to capture sockets that today require more expensive (BAW/Fbar) filters. Murata has been working on this for several years, but even their original estimate did not have this in commercial production for another year or so.
Two smaller companies also merit some attention – Akoustis and Resonant. (Note: We own a small position in RESN stock.) Both companies are following this path of filter substitution, albeit using very different methods. Akoustis has built their own production facility, and after a shaky start, this appears to be making progress. By contrast, Resonant has a software and IP suite that allows them to design very complex filter modules. They are working with 3rd parties, including Murata, to manufacture these products. This is a new path for filters, analogous to the fables model used in traditional semiconductor manufacture. Both companies have shown promising early results, but in RF, manufacturing yields are hard to predict, and ultimately mean more than pure technology in determining viability.
And then there is China
As with all things in technology today, we also have to consider what is happening with RF in China. There is obviously a lot going on in semis in China today. China has set out to reduce its reliance in foreign semiconductors, and this includes RF.
As we wrote about last week, there are over 1,000 chip design companies in China today. Many of them are small firms doing custom RF work. Several years ago, Qorvo, closed their plant in Beijing, and that talent seeded a network of RF shops, in addition to pockets around Shenzhen and Shanghai, traditional handset centers. As far as we know, none of these companies are large scale, but the potential exists for one or two to break out and reach greater success.
As with everything else in this post, the big question is filters. We know of two companies in China that have some filter capacity, and they are both tied to China’s electronics academic complex, so not hot, nimble start-ups. They have been toiling away quietly for several years, and their lack of prominence implies limited success so to date.
Finally, there is Huawei. No company is more in the hot seat for its reliance on foreign components. A big focus in our conversations with investors about Huawei is its ability to acquire RF products for its phones. Huawei designs most of its phones’ silicon itself, but relies heavily on foreign companies for RF. There are signs that last year they filled warehouses last with Skyworks and Qorvo parts in advance of tighter US restrictions. And Murata, a Japanese company, also seems to be doing very healthy business with Huawei. Moreover, there are signs that Huawei has built its own RF capacity through its chip subsidiary, HiSilicon. Details on this are hard to come by, but HiSilicon has recently proved itself highly capable, so we would not surprised if Huawei will soon be able to design many parts of an RF chain internally. But as above, their ability to source filters will remain challenging.
We are living in the Golden Age of RF. The advance of the 4G standard, with all its very heavy, complicated RF requirements led to a surge in demand for highly engineered solutions. After 15+ years of consolidation, there are only a handful of suppliers left. Combining these two has meant strong demand and healthy prices.
It is too soon to say this Age is over, but we are definitely close to, maybe just beyond, the peak. The transition to 5G is shifting the technological challenges for RF products. The need for old solutions will not go away, but the premium paid for past innovation will diminish. When we eventually get to mmWave systems, the RF componentry will look very different. This is exactly what happened with every past ‘G’, and each successive generation of the wireless standards has seen a wave of consolidation among the supplier base. The incumbents will need to invest heavily into mmWave RF, they cannot merely extend their existing capabilities. Not every company can weather that transition.
At the same time, there are problems from above and below. The growth of RF providers in China will put pressure on the bottom half of the market. And then the incumbents face a serious threat from Qualcomm.
The pressure looks likely to fall most heavily on Skyworks and Qorvo. Both companies are aware of the problem. Note their repeated public statements and actions about non-mobile business. Both companies have been trying to diversify away from handsets for a long time. And the non-mobile RF business looks very attractive, offering higher margins for highly engineered solutions. The problem is that those markets move at a much slower pace. For both companies, the non-mobile business contributes about the same share of revenue today as a decade ago. Neither company is particularly adept at M&A. Qorvo tends to overpay for sub-scale companies that then need years of investment. Skyworks, by contrast, seems to be afraid of overpaying, to the point of paralysis, thus doing very few deals.
We think ultimately, one of those companies gets acquired, possibly by the other one, or by some other larger vendor.
And as for all those new technologies, if any of them can prove their technology in the market, it is unlikely that they hit hyper growth and become dominant players. Instead, they get acquired into one of the incumbents. (Note: Take this with a grain of salt. We own a position in RESN). This is a mature industry with highly a concentrated customer base, so there are a limited number of outcomes. After many years of healthy, fun growth, it looks likely that the industry will be changing a lot, soon.