Thematic Research and Bits & BytesCharter Equity Research publishes wireless industry thematic research and comprehensive forecasts of major trends in wireless. These pieces combine detailed studies with industry interviews and analyses to develop a working thesis on where the industry is headed and why.
To access the following research, please login. If you need login credentials, or if you are interested in receiving more information about our thematic research, please contact us at info@charterequityresearch.com
Jump to Bits & Bytes
Previews of our wireless industry thematic research:
WOLF, COHR, STM, IFX.DE, ON, NVTS: ICSCRM 2024: SiC conference headlined by Chinese competition, thin 200mm wafers. (October 4, 2024)Excerpt: Throughout the conference it was evident that the silicon carbide slowdown is not letting up, due to slower-than-expected EV adoption and inventory corrections in the industrial and energy markets. Lower demand plus stronger competition from Chinese wafer suppliers have made cost reductions a priority, particularly in materials, and data from the conference suggests the industry is making considerable progress on 200mm yields as well as thinner (less expensive) 200mm wafers.
AMD, AVGO, INTC, MRVL, NVDA: Be careful using MLPerf as a proxy for real world performance. (August 30, 2024)Excerpt: AlexNet’s stellar performance was derived from a novel model (simple by today’s standard) run on GPUs. Its success spurred much of the academic and industry work that led to the MLPerf benchmark. Since then the benchmark has expanded, but to remain comparable over time, it maintains a degree of backward compatibility in model size and metrics, runs on GPUs, and evaluates basic workloads. Those constraints have made it more a target for chip companies to chase than a measure of real world performance.
TXN, ADI, ON, STM, IFX.DE, MCHP, NXPI: Not quite wheels up for analog semis, but the plane is taxiing. (August 22, 2024)Excerpt: Quarterly analog sales have declined more than 20% since peaking in 2022 and aggregated September guidance implies the first sequential revenue increase since June 2023. More importantly, combined September outlooks mark the first improvement in y/y revenue and gross margin trends since June 2022.
IFX.DE, ON, STM, WOLF, NVTS, TXN: Webinar takeaways: The shift from EVs to hybrids and implications for semiconductors. (June 11, 2024)Excerpt: Incremental content gains when moving from gas engines to electrified powertrains are largely driven by the need for more and higher-power IGBTs and power MOSFETs. There are tailwinds to other semiconductors such as analog and digital logic but they are more often driven by the vehicle’s tier or price tag rather than its level of electrification. As a result, the vast majority of the 8%-10% shrinkage in the 2030 automotive semiconductor TAM from lower BEV adoption is due to less power content.
IFX.DE, ON, STM, WOLF, NVTS: Power needs of AI datacenters present significant opportunity for SiC and GaN. (March 28, 2024)Excerpt: As a result the power density of datacenters, or how much electricity is used for a given footprint, is going to increase substantially to accommodate power-hungry AI capabilities. Silicon carbide (SiC) and gallium nitride (GaN) power semiconductors boast significantly higher power density than silicon because faster switching and fewer losses (greater efficiency) allow datacenters to use the same amount of power in a materially smaller footprint.
SWKS, QCOM, QRVO, AVGO: iPhone 16: SWKS giving up share to QCOM; QCOM RF content likely down y/y. (March 18, 2024)Excerpt: As we discussed in previous notes and conference calls, we believe the modem extension agreement Apple signed with Qualcomm last fall included new restrictions limiting Apple’s ability to wiggle out of awarding Qualcomm RF content. As a consequence, it appears the OEM is taking RF modules from Murata and Skyworks, which it favors, and awarding them to Qualcomm, which it loathes.
WOLF, ON, STM, TSLA: APEC 2024: Tesla’s new inverter uses far less SiC but likely leads to higher SiC MOSFET demand. (March 6, 2024)Excerpt: The push for a low cost EV makes sense in light of the impressive increase in sales of the Model 3/Y after several rounds of price reductions. If the inverter in the NV9X platform uses 75% less SiC than in the Model 3/Y, it would still translate to significantly higher SiC demand given most low-cost EVs now use silicon.
WOLF, ON, STM, IFX.DE, NVTS: APEC 2024 themes: Datacenter power content, GaN in EVs, vertical GaN. (March 4, 2024)Excerpt: Larger workloads require more power (heat), creating a significant opportunity for SiC and GaN which improve efficiency and reduce the need for cooling. Google, Meta, Amazon and other hyperscalers are projecting hundreds of billions in CAPEX over the next several years to increase capacity for AI applications. If a small fraction of that spending trickles down to power devices it would have a meaningful impact on the TAM for SiC and GaN.
QRVO, SWKS, AVGO, QCOM: Huawei’s Mate 60 poses no threat to U.S. RF semi suppliers. (November 30, 2023)Excerpt: If you crack open any of the leading advanced 4G or 5G smartphones, you’ll find an RF front-end built using modules from Broadcom, Skyworks, Qorvo and, to a lesser extent, Murata and Qualcomm. This is the case even with Vivo, Oppo, Xiaomi and Honor, Chinese OEMs being compelled by their government to source from domestic suppliers but still selecting Western modules for flagship models.
QRVO, SWKS, AVGO, QCOM: Chinese RF is no threat to U.S. suppliers, on the contrary. (October 12, 2023)Excerpt: Chinese RF suppliers are, for the most part, discrete component manufacturers trying to grow into modules. Leaders such as Vanchip and Maxscend have carved a niche out in discrete switches and amplifiers but their attempts to gain share in the high-density modules for 5G and 4G Advanced phones have thus far failed. Were it otherwise, we would see them in teardowns of the leading smartphones instead of waiting for Huawei to field a 5G phone using only domestic suppliers.
WOLF, ON, STM, IFX.DE: Major themes from SiC power conference: wafer shortages, rising equipment demand, China’s growing presence and progress toward 200mm. (September 28, 2023)Excerpt: One of the clearest indications from the International Conference on Silicon Carbide and Related Materials (ICSCRM) last week in Sorrento, Italy is that despite significant capacity expansions, wafer supply remains tight, and this appears to have become more acute over the past three months.
QRVO, SWKS, AVGO, QCOM: Wish upon a Huawei star. (July 14, 2023)Excerpt: Huawei was the first in production with a fairly good 5G baseband but lost the ability to produce it once sanctions took hold. It’s trying to get back in the game using Chinese technology but those are three-to-four generations behind what Qualcomm and MediaTek use. Worse still, its application processor is at least that far behind in hardware and further in software.
QRVO, SWKS, AVGO, WOLF, STM, IFX.DE, COHR, NVTS: Chinese Ga/Ge restrictions not a threat to U.S. firms in the long-term. (July 6, 2023)Excerpt: Reductions in shipments of Gallium and Germanium from Chinese producers can be offset by suppliers in Japan, Korea, Germany, Belgium, Russia, the Congo and, eventually, the U.S. Because those materials are byproducts of zinc and aluminum (Al) production, there are few impediments to developing new sources.
SiC Wafer Spotlight - Tianyu: Chinese demand strong but competing internationally will be a challenge. (April 11, 2023)Excerpt: The majority of SiC wafers used in China are supplied by foreign companies, although trade disputes and the country’s push for technological independence should benefit Tianyu, even if its epiwafer quality lags foreign competitors.
QCOM, AAPL, QRVO, SWKS, AVGO: Apple modem delays likely push its debut to 2025. (April 5, 2023)Excerpt: When it bought Intel’s baseband group in late 2019, conventional wisdom predicted the modem’s debut in 2022. Having followed nearly every cell phone modem development since Samsung launched its effort in 1999, we didn’t believe any firm could enter the market and field a competitive baseband in less than four years.
VZ: Down the rabbit hole. Is it time for new leadership? (March 27, 2023)Excerpt: Between CAPEX for a network with scant coverage and few users, millions in subsidies to Apple to include mmWave in iPhone, and CPE radios with no buyers, Verizon spent three years digging a mmWave hole only to follow it with $45B for C-band licenses and CAPEX for C-band equipment. Its reward for such prodigious spending has been a record increase in churn and steady decline in margins.
IFX.DE, STM, ON, WOLF, NVTS, MRN: APEC 2023 takeaways: GaN-on-silicon takes the spotlight. (March 23, 2023)Excerpt: Total GaN revenue is a few hundred million now so we expect the power semiconductor companies which missed out on the silicon carbide ramp are jumping into GaN before it evolves into a multi-billion dollar market. However the difference is that SiC had two significant catalysts: a substantial decline in cost and adoption in the Tesla Model 3.
STM, ON, WOLF: Fears of lower automotive SiC content miss several key nuances. (March 10, 2023)Excerpt: Tesla’s comments on fielding an inverter that uses less silicon carbide were one part of a larger presentation on various initiatives it’s undertaking to field a lower priced EV. The discussion included copious detail about cost reductions in many other areas including manufacturing operations. Investors jumped on and amplified its SiC comments to the detriment of STM, WOLF and ON but the overarching theme of its investor day was expansion into lower price tiers, not cost and performance reductions in current models.
SWKS, QRVO: iPhone 15 RF front end content estimates. (February 14, 2023)Excerpt: Like last year, we believe the largest increase in RFFE content for iPhone 15 will be for more dual-connectivity, which are more band combinations of download carrier aggregation and EN-DC (4G/5G simultaneous operation). We showed these trends in our iPhone 14 teardown presentation (“Charter’s Die-Level Teardown of the iPhone 14, A Detailed Look at RF Components and Content”, September 19, 2022) and are confident they’ll continue with iPhone 15.
AVGO, QCOM, QRVO, SWKS: Apple is still not getting into RF module design. (January 12, 2023)Excerpt: Taking Bloomberg's speculation as fact is bad enough, but extrapolating that to Apple in-sourcing RF components is grossly incorrect. If it had been released a year ago it would have seen stiff competition from another Bloomberg article that made the same claim because its authors didn't understand the difference between basebands, transceivers and RF components.
ON, STM, WOLF, COHR, SWKS, QRVO: Takeaways from CES 2023. (January 9, 2023)Excerpt: While Sony and Samsung account for the majority of the broader CMOS image sensor market, Onsemi and STMicro are strongly positioned in the automotive portion of the market and sell a wide array of other automotive semiconductor devices so they can capture more business through bundling. Coherent’s 3D sensing business is currently geared toward the iPhone but in-cabin sensing should provide additional growth opportunities if smartphone content growth moderates.
SiC Wafer Spotlight – Showa Denko: Strong player in epitaxy making a push into substrates. (December 15, 2022)Excerpt: Customers’ insourcing efforts and new entrants cast a pall over Showa Denko’s prospects as an epitaxial pure-play which is why it’s shifting its focus to substrates. An internal source of substrates would eliminate the margin stacking from purchasing substrates externally and as a result Showa Denko could lower the price of its epiwafers to remain competitive. If successful, 200mm wafers would provide another cost advantage although given the company is less than one year into its 150mm ramp we don’t expect 200mm volume production for several years.
SiC Wafer Spotlight – TankeBlue: China’s largest wafer supplier likely struggles to compete abroad. (December 1, 2022)Excerpt: Since the Huawei ban in 2019, China has ramped investments in its own semiconductor industry to wean itself off American chips. Given TankeBlue is a state-owned enterprise we expect the company’s decision to build a new factory was influenced by the government and not necessarily due to outsize demand for wafers.
SiC Wafer Spotlight – SK Siltron: Considerable experience meets substantial opportunity. (November 29, 2022)Excerpt: The company’s long-term guidance is relatively aggressive, and while it has made considerable progress in scaling the SiC wafer business in just a few years, adding capacity is arduous for any company. Despite this challenge, SK Siltron is forecasting production to increase from 30K wafers in 2021 to 600K by 2025, roughly equivalent to the entire SiC wafer market today.
QRVO, SWKS, AVGO, QCOM: What to expect for iPhone 14 RF content gains by supplier. (September 7, 2022)Excerpt: Once Apple provides details of the iPhone 14, we’ll have considerably better visibility on exactly where RF content is increasing and who will get what. Nevertheless, we already have sufficient information to make rough predictions by vendor.
IFNNY, WOLF, ON, STM, ROHCY: Infineon’s new fab illustrates soaring demand from industrial and solar customers. (February 18, 2022)Excerpt: Although EVs have been the primary focus of management, analysts and investors alike, the industrial market has been and will remain the primary catalyst for silicon carbide device revenue growth for the next few years. We believe Infineon is building its new fab not necessarily in anticipation of EV ramps in 2024 and beyond but to fulfill stronger demand from industrial customers.
Humor dominated Super Bowl commercials, but the biggest joke were ads for EV trucks. (February 16, 2022)Excerpt: Like spots hawking cryptocurrency, the EV ads were long on the promise of happiness and short on delivery. Of particular amusement were Toyota's and GM's ads for electric pickup trucks. You could hardly ask for a better joke than an electric work truck, which is probably the supreme example of over-promising and under-delivering.
CREE: A detailed analysis of what Mohawk Valley has to do for Cree to hit FY24 targets (September 11, 2021)Excerpt: To estimate the probability of success and better understand what has to occur and when, we undertook a detailed study of Cree's material and device fabs and compared them to a rival SiC power device fab at Infineon. The results of our analysis increased our confidence Cree can achieve its FY24 targets but also highlighted the hurdles in the daunting tasks ahead.
VZ: Verizon ignores the first law of holes: If you find yourself in one, stop digging. (February 25, 2021)Excerpt: The 5G game has proven more expensive for Verizon than any other domestic carrier largely because of its own mistakes. To earn bragging rights and promote its 5G prowess, it jumped the gun on the 5G standard with 5G TF, deployed expensive, poor performing mmWave systems on a few streets while burdening handsets nationwide with higher BOM costs and itself with higher subsidies. When the performance of its DSS system fell below its 4G LTE service, it added another layer of cost to its handsets. When it was clear its mmWave mistakes and DSS band-aids weren't going to deliver a usable system, it spent lavishly on C-band spectrum.
RF Semiconductors, CREE: Semi shortage has little impact on RF and none on silicon carbide. (February 19, 2021)Excerpt: The problems in semiconductor supply are primarily focused in digital CMOS and, to a lesser extent, mixed signal (digital and low frequency analog) products, all of which are silicon-based. RF components are high-frequency analog devices fabricated on GaAs, SiGe, SOI and variety of piezoelectric acoustic materials. All of Cree's products are power devices on silicon carbide. As such there is virtually no overlap between the digital CMOS/mixed signal supply chain and the RF semi supply chain and zero overlap with silicon carbide.
AAPL: Detailed analysis of impending changes in 3D sensing in iPhone. (February 10, 2021)Excerpt: We believe changes in the 3D sensing systems employed by iPhone 13 this year and iPhone 14 next year will have significant impacts on LITE, IIVI, AMS and STM. A recent client request for a detailed evaluation of what’s currently being used and what’s likely to change provided us the opportunity to do a deep dive into the technologies of 3D sensing.
RF Semiconductors: Apple: Highly unlikely to build its own RF modules. (November 30, 2020)Excerpt: A recent press report suggesting Apple may begin building its own RF module has stoked investor concern about the long-term prospects for RF businesses at Broadcom, Qorvo and Skyworks. We believe there's little chance Apple will try to bring RF modules in-house. Apple has a lot of experience designing high-density, digital CMOS devices and zero experience in high-frequency analog. The difference in engineering between the two is the difference between a certified public accountant and an Olympic fencer.
RF Semiconductors: Surprises from our teardown of the iPhone 12. (November 2, 2020)Excerpt: From an RFFE architecture perspective the biggest surprise was antenna count, which was far above anything we thought possible. In big content years a single antenna would be added but between an ever-expanding RF component count and a limitation on the size of the case, it was generally believed six-to-eight antennas was the limit. Last year’s iPhone (11) had six antennas so we expected six to seven in iPhone 12; our teardown found 13.
RF Semiconductors: iPhone 12 RF content estimates likely too low. (October 26, 2020)Excerpt: Our teardown of iPhone 12 reveals substantial y/y content increases for Broadcom and Skyworks and higher than expected content for Qorvo. On Friday, we received the U.S. model of iPhone 12 (A2172) and performed our initial board level tear-down in preparation of a more detailed die level analysis on Monday. Of the three RFFE SKUs used for iPhone 12, the A2172 is the most complex, supporting 37 bands including mmWave. It also includes all the advanced features required for 5G including the ability to download six data streams at the same time and transmit two signals to the base station simultaneously.
RF Semiconductors: Huawei sanctions benefiting Qorvo and II-VI. (September 21, 2020)Excerpt: U.S. semiconductor manufacturers have witnessed firsthand the downside to supplying Huawei and will likely avoid the OEM even if sanctions are eased, to insulate themselves from sudden shocks in demand. Mobile RF semiconductor suppliers have found suitable replacements for the lost Huawei revenue and infrastructure semiconductor manufacturers have pivoted away from Huawei.
RF Semiconductors: The 5G iPhone mmWave dilemma. (September 8, 2020)Excerpt: Normally the answer to the SKU question is "two plus". Unfortunately, in the year of COVID, nothing is normal. Verizon’s requirement that all high-end 5G smartphones support its flailing mmWave service throws a wrench in Apple’s SKU machine. Nothing about mmWave appeals to phone OEMs. They’d rather not include it, but if they don’t, Verizon won’t support the phone and given how important carrier sales are to unit volumes in the U.S., Verizon’s wishes are effectively their command.
Rapid selloff, rapid recovery: What 25 years of market corrections tells us about economic impact of COVID-19. (April 18, 2020)Excerpt: Panic regarding COVID-19 kicked off a market selloff that set records in magnitude and speed. It’s since recovered most of those losses but remains depressed on concerns the economic impact of the cure will be worse than the disease. It’s a reasonable fear given the record $10B/day decline in the S&P500 in March was followed by government shuttering of much of the economy, leading to the fastest increase in unemployment in peacetime.
RF Semiconductors: New iPhone SE uses iPhone 11 front-end, INTC baseband. (April 15, 2020)Excerpt: Apple just announced the release of the new iPhone SE. The specifications suggest it reuses the iPhone 11 RFFE which isn’t a surprise given the phone exists to address the mid-tier price segment ($400 ASP) so cost reduction is paramount. New RFFE architectures are expensive as are new basebands so, as we expected, iPhone SE reuses last year’s iPhone 11 front-end and the Intel baseband that powered it.
Semiconductors, OEMs: Coronavirus: Large and pervasive negative impact on demand. (February 18, 2020)Excerpt: We believe the coronavirus’ impact on the Chinese economy will force many of the companies in our coverage to preannounce the March period or miss consensus forecasts on the quarter and for the June quarter. The government’s extension of the Chinese New Year, its lock down of entire provinces and the quarantine of tens of millions of Chinese citizens has undercut demand and stymied production. Restrictions on movement extend well beyond Wuhan and include Beijing, where the state is now requiring 14 days of isolation for anyone entering the city.
RF Semiconductors: Huawei Mate 30 Pro teardown suggests a return to U.S. components is inevitable. (January 7, 2020)Excerpt: In December we tore down Huawei’s Mate 30 Pro 5G to determine whether it contained any U.S. RF components and if not, what type of work-around was implemented to make up for the loss of high-density RF modules that can only be sourced from U.S. suppliers. As we predicted in several notes and investor calls dating back to the Huawei ban in May 2019, the phone doesn’t use any U.S. RF components and as a consequence, Huawei had to revert to an old, discrete architecture for the RF front end (RFFE).
RF Semiconductors: How Huawei can field flagship smartphones without U.S. components. (September 26, 2019)Excerpt: Huawei can’t short-circuit 25 years of knowledge and experience in RF to fabricate its own components and it can’t buy what it needs from U.S. suppliers, which leaves it with four choices: 1) abandon flagship models, 2) release phones using what components it can obtain from non-U.S. sources, 3) pull back from difficult RF regions and concentrate on China and India, 4) develop a backchannel to U.S. suppliers.
RF Semiconductors: Mislabeling parts doesn’t help Huawei’s predicament. (August 14, 2019)Excerpt: A recent teardown of Huawei's Mate 20 X 5G phone by Tech Insights misidentifies several RF components, confusing small, inexpensive parts for large, high-ASP modules. These errors were amplified Monday by a competitor's note citing Tech Insights’ analysis as proof that Chinese component manufacturers, more specifically HiSilicon, are replacing U.S. RF modules with internally sourced devices. This is not correct.
RF Semiconductors: Why Huawei needs U.S. components to build flagship smartphones. (August 9, 2019)Excerpt: The U.S. government’s ban on most exports to Huawei has stoked speculation the Chinese OEM can replace U.S. RF components with Japanese, Taiwanese and Chinese suppliers. This is highly unlikely. The size and performance requirements for a competitive LTE Advanced, LTE Pro or 5G phone require OEMs to employ complex, high-density RF modules that are only available from Broadcom, Qorvo and Skyworks; there are no other alternatives.
RF Semiconductors: You can’t build a flagship smartphone without U.S. RF semiconductors. (July 27, 2019)Excerpt: The Huawei kerfuffle had investors questioning the possibility of the Chinese OEMs replacing U.S.-sourced semiconductors in the RF front-end (RFFE) and carrying on as normal. Huawei has made claims to that effect".
Welcome to the 5G logo wars. (June 27, 2019)Excerpt: We've been told by several folks deep across wide swathes of the wireless food chain, the killer application for 5G is the logo on the box that says "5G". It certainly seems to be playing out that way. T-Mobile has built out a 4G network on its new 600MHz band (Band 71) onto which it will dabble the smallest spec of 5G - higher density modulation on a paltry 10MHz of the 35MHz band and then crown the entire service "5G".
Over-charging for 5G modems has consequences. (June 20, 2019)Excerpt: Even before the FTC ruling, Qualcomm’s grip on the cellular baseband market was slipping. Intel emerged as a suitable replacement at Apple (at least for 4G), Huawei sourced nearly all its needs from HiSilicon, and Samsung was using its own LSI modem in everything from the Galaxy and Note to mid-tier models. Oppo and Vivo ebbed and flowed with MediaTek while Xiaomi was heavily committed to Qualcomm.
Thou shalt not pass. (June 18, 2019)Excerpt: RFFE modules will grow in size and complexity, but they won't combine low with mid-high bands. Having tracked the evolution of RF front-ends from before Triquint introduced the PAD in 2005, we've witnessed the growth in RF modules from single band to 12 bands in the 2018 modules. The relentless increase in size and content has investors wondering when they will see one module to rule them all; a single placement device combining low, mid and high bands.
Massive MiMo is French for WiFi. (June 13, 2019)Excerpt: Any conversation about 5G in cell phones inevitably involves discussion of massive MiMo (beam steering antenna arrays) because there's no practical way to establish a radio link to a mobile phone over 5G frequencies (Sub-6 or mmWAVE) without them. This makes "5G" and "massive MiMo" synonyms at least as far as connections to cell phones are concerned. The problem for carriers is that "massive MiMo" and "Mobility" are antonyms.
Impact of Huawei ban on phones, base stations and semiconductors. (May 22, 2019)Excerpt: On Wednesday, May 15, the U.S. government banned U.S. firms from shipping products to Huawei, causing an extended sell-off in semiconductor and equipment stocks. Heavily reliant upon American technology, Huawei finds itself in a similar position as ZTE when it was blacklisted in April of last year. That ban nearly drove the company out of business and even after it was lifted in July, the damage persisted as many customers shifted to other suppliers that didn’t face the possibility of future disruptions.
GaN is good, but not for handsets. (May 16, 2019)Excerpt: After a decade of gestation, GaN is finally coming into its own. The advantages over silicon and GaAs in power output and efficiency are substantial enough to get the notoriously parsimonious wireless network OEMs to pay up for GaN in 5G base station power amplifiers. The defense industry has been using it for years and every new hypersonic, stealth and ballistic missile threat increases the depth and breadth of its reach in land, sea and airborne radar.
The ripple effect of the Apple-Qualcomm settlement. (April 30, 2019)Excerpt: The litigation settlement between Apple and Qualcomm will ripple through the food chain, altering content expectations for the various component suppliers. The biggest beneficiary is obviously QCOM which experienced its largest single-day stock price increase in nearly 20 years; a testament of how much more important a settlement was to Qualcomm than to Apple, which was up only 1%.
5G and the fallacy of faster duty cycle. (April 22, 2019)Excerpt: 5G: If you're deep in the mobile cellular world and given to technical discussions sooner or later, you’ll probably find yourself in threadbare debates with 5G evangelists over use cases. The starting point is usually how the lack of spectrum in the legacy cell bands pushed 5G to higher-frequencies. High-frequencies don't propagate well so getting enough signal to the phone requires "massive MiMo".
Euphoria over Apple settlement may be short lived but it's not good news for Qorvo. (April 17, 2019)Excerpt: 5G: Intel thought it could, but Apple didn’t share that confidence. As recently as last week, our checks deep in the baseband supply chain showed a high degree of confidence that Intel would deliver its 5G modem to Apple on time. It also revealed Apple didn’t share this opinion and in fact, had little confidence in Intel’s ability to meet its schedule.
The 5G Fairy Tale. (March 27, 2019)Excerpt: 5G is turning into a pretty nice little bubble, especially for Qualcomm. The sell-side is teeming with reports of Qualcomm’s success in landing modem and RF content in most 2nd and 3rd tier 5G phones slated for release this year. The facts of those wins may be true, but the implications are not.
The cost of testing 5G phones could kill them at birth. (February 4, 2019)Excerpt: Unlike in LTE, 5G requires MiMo, which broadcasts multiple signals on multiple antennas, every one of which must be tested across all bands, modes and use cases. With so many more combinations to deal with, test equipment for 5G is extremely complex, much more so than for 4G.
Apple's future in phones. (January 12, 2019)Excerpt: Apple’s problems don’t have to lead to failure, but in the long run they probably will. The path it will doubtlessly follow was forged by Ericsson in 2001 and retraced by every major handset OEM failure since . Product problems at each of those companies undercut unit sales, ate into margins and put them further behind their competitors.
What to expect for Apple & the RF semi suppliers. (January 3, 2019)Excerpt: Apple’s problems are nothing new and likely to become more acute. Having covered the handset industry since 1997, we had a front row seat to a long procession of failures in phones starting with Ericsson in 2001 followed by Siemens (’05), Motorola (’07), Sony-Ericsson (’08), Nokia (’09) and finally Blackberry (’12).
Trade wars, embargoes, hurricanes - RF content is still going up. (October 23, 2018)Excerpt: Building a 4G phone without U.S. components is like trying to launch satellites without rockets. A recent note out of a competitor calls for zero content growth at U.S. RF semiconductor manufacturers on the belief that Chinese OEMs are shifting to Japanese suppliers to ameliorate the impact of trade war restrictions.
Will the U.S./China trade war impact iPhone sales? (October 16, 2018)Excerpt: The trade war could affect iPhone sales, but we believe there are other more important factors impacting unit sales which have nothing to do with China. Saturation in high-end demand. High-end smart phone unit sales have been stagnant for several years now, a trend evident in both Apple and Samsung’s results.
Qualcomm's campaign to put a man on Mars. (October 12, 2018)Excerpt: We’ve received a plethora of requests to handicap Qualcomm’s prospects in 5G, particularly in mmWAVE. These are largely the result of the marketing campaign Qualcomm pursued in January to fight off Broadcom, which was amplified more recently by a sell-side competitor’s breathless regurgitation of the company’s marketing flyers as research.
RF Semiconductors: What's ailing wireless? (September 19, 2018)Excerpt: Over the last several weeks we have fielded a large number of investor calls on the weakness in the RF semiconductor stocks, especially SWKS and QRVO. Most questions focus on three primary concerns but a poll of large holders and the companies themselves suggest there are four concerns pressuring the stocks.
Bull case on Qualcomm's 5G RF is far off the mark. (August 8, 2018)Excerpt: Lowering price targets on the threat of losing 0.8% in revenue in two years seems nonsensical? A competitor’s note on Monday claiming Qualcomm’s gains in 5G RF are a blow to Skyworks and Qorvo is far off the mark in our opinion, stretching comments from a single "expert network" consultant to illogical conclusions not backed by events and trends on the ground.
The Big Challenge for 5G (June 28, 2018)Excerpt: It’s difficult to make a singular statement about 5G other than it’s the Tower of Babel for wireless standards; a technology looking for a market. Like the wireless broadband systems of the early milieu, 5G is just the latest version of the "If you build it, they will come" school of economics.
Thoughts on GaN (June 19, 2018)Excerpt: The International Microwave Symposium (IMS) was held in Philadelphia last week. It may sound like a conference on cooking, but it’s actually one of the preeminent gatherings of engineers and scientists in RF semiconductors.
Qualcomm vs. Broadcom: Checkers or Chess? (February 13, 2018)Excerpt: With last week’s unanimous rejection of Broadcom's second offer, Qualcomm’s future will now be determined by the board election on March 6.
Points of interest on RF. (January 16, 2018)Excerpt: The 5G filter story is being muddled.
Likely scenarios for Broadcom’s acquisition of Qualcomm. (November 5, 2017)Excerpt: We don’t believe an acquisition of Qualcomm would herald a sea-change in Broadcom’s business model although it will upend the cellular modem industry.
What to expect on the quarter and in guidance for 4Q17. (October 26, 2017)Excerpt: With so many moving parts in the mobile phone business and such large exposures to the big three OEMs (Apple, Samsung, Huawei), it’s often difficult to forecast hits and misses for Skyworks and Qorvo in individual quarters.
Big changes afoot in RF Semiconductors. (September 15, 2017)Excerpt: Carrier’s quest for the 1Gbps phone is proving a boon to RF semiconductors. We just completed a series of technical discussions with leading companies and engineers in semiconductors and handsets, and have concluded the quest for a 1Gbps phone will be a boon for the leading RF semiconductor companies.
iPhone 8: What to expect for new bands. (September 11, 2017)Excerpt: We’ve been asked to comment on a recent note
postulating on the possibility of various bands showing up in Apple’s new phone. Having tracked RF
filter development for many years, we are familiar with the history, progress and likelihood of
deployment of all the bands mentioned in the note.
RF Semiconductors – What to expect for the RF stocks with Apple’s upcoming phone release. (July 7, 2017)Excerpt: We’ve fielded several questions about variations in RF content by SKU, with most investors expecting three new models in September: upgrades to iPhone 7 and iPhone 7S, and a new high-end flagship, iPhone 8. We expect RF content gains in all models.
RF Semiconductors – Trends for 2017. (January 18, 2017)Excerpt: We are highly confident that content will increase in 2017 at all the major handset groups: Apple, Samsung, China Inc. (top 10 Chinese OEMs), and China white-box (MediaTek, QRD, Spreadtrum). The largest percentage increase will almost certainly be in China white-box phones where MediaTek is accelerating its move to PADs.
RF Semiconductors – Even if iPhone’s release schedule is stretched (unlikely) it will have no bearing on RF. (June 4, 2016)Excerpt: How can Apple stretch its phone release schedule to three years when it doubled the models and halved its time-to-market for iPhone in 2016? A news report out of Asia suggests Apple will extend the revamp schedule for iPhone from two to three years because of slowing demand. While this is certainly possible, we don’t believe it’s likely…
iPhone 6S: Teardown analysis. (September 25, 2015)Excerpt: Total RF content up ~25% y/y for iPhone 6S. The teardown photos published last night confirm our forecast of a higher overall RF content from an increase in the number of LTE bands, a move to MIMO in the Wi-Fi section and larger, more complex diversity receive modules. The final word on how much all of this is worth will have to wait until the quarterly reports but discussions throughout 2015 suggest total additional spending will be in the $4.00 range, bringing iPhone 6S to about $24 in total RF content.
RF Semiconductors - The Ten Most Common Fallacies about the RF TAM. (September 22, 2014)Excerpt: #1: MMPAs- Multi band amplifiers will accelerate price erosion and destroy the market. The arrival of multi mode amplifiers (MMPA) in 2010 did drive down pricing, but only on a per-band basis, and then only in high-band count devices; they had no impact on ASPs for PAMs, PADs, FEMIDs or filters…
iPhone 6 – A Detailed Teardown Analysis. (September 19, 2014)Excerpt: Apple’s philosophy of minimum SKUs pushes total RF content up again. Contrary to what we thought two years ago, rather than segment its phones by region to reduce BOM costs, Apple has taken the opposite path, and decreased the number of RF SKUs to two with iPhone 6. This has increased the number of bands, as well as the complexity and cost per phone. Even if the unthinkable occurs and unit volumes are flat y/y, total dollars spent on RF by Apple will increase about 30% on higher content alone.
RF Semiconductors – A thorough look into the RF Content of the iPhone 6. (August 11, 2014)Excerpt: We believe that the iPhone 6 will have about $17 in RF content and that Avago will be the big winner. Our analysis suggests $16 - $18 in RF content for Apple’s new phone, lower than our previous forecast of $18 - $19, because we no longer expect envelope tracking (ET) to be used in the new phone due to the problems experienced with Qualcomm’s ET device (QFE11xx).
A Field Guide to Filters (September 9, 2013)Description: A meticulous 61-page primer explaining filter fundamentals (cost, insertion loss, temperature drift, distortion, and more), summarizing duplex filter products (front end modules, power amplifier duplexers, antenna filter modules), and comparing BAW vs SAW technologies.
Understanding Disparities in the Food Chain’s Guidance for iPhone 5S. (July 27, 2013)Excerpt: The stark differences in September 2013 period guidance between many of the vendors supplying parts to iPhone 5S has heightened concerns over share loss, slow demand and delays in the release of Apple’s next phone. Of particular concern is the huge difference between Triquint’s blow-out guidance and Qualcomm’s short-fall.
RF TAM Analysis: A Detailed Look at Spending on RF Components for Mobile Devices (May 21, 2012)Description: An in-depth 29-page report with detailed discussions of unit growth versus content compression, 2G to 3G conversion, High Density Amplifiers, and accelerating price erosion in commoditized devices.
Estimates for phone demand in 2011 are too low. (December 13, 2010)Excerpt: Phone growth could well exceed 20% y/y in 2011. Despite what appears to be a robust recovery in handsets, sell-in remains insufficient to satisfy the pent-up demand induced by three years of economic weakness. At 15%, the replacement rate for mobile phones hit a ten year low in 2009 after averaging 22% for the previous eight years.
Android's opportunity to jump ahead of iTunes (July 8, 2010)Excerpt: Next generation music platform is the major battle in the smartphone wars over the next 12 months. Last week's announcement that Google is activating 160K Android devices per day (a run rate that exceeds 14M per quarter), demonstrates Android’s ability to attract new users when combined with focused carrier marketing. Clearing that hurdle, the next test will be "stickiness"; the platform’s ability to retain users through several generations of replacements, as well as entice them to purchase secondary devices such as tablets and set top boxes.
Zeitgeist: A cautionary look at the world’s appetite for handsets. (June 19, 2010)Excerpt: The recession is in check, worldwide sales of handsets are up, demand for mobile semiconductors is booming and OEM inventory in the April reports remained below normal; everything seems to be going great so why are we so worried? Our growing apprehension stems from a slew of data points that individually may not mean much but in aggregate seem to spell trouble.
Zeitgeist: Arguing RF Semiconductors. (May 2, 2010) Excerpt: Nokia’s implosion last week stoked investor’s concern that the handset food chain is headed for leaner times, particularly the semiconductor manufacturers which posted record results even as Nokia’s margins reach record lows. The idea that component suppliers can’t do better than their customers is rooted in years of experience with the boom-bust cycles of the PC semiconductor industry. The rollercoaster of scarcity and surplus that drives that cycle is a direct consequence of Moore’s law and the inertia inherent in multi-billion dollar fabs.
Zeitgeist: Drifting from Reality in Phones. (January 31, 2010) Excerpt: We were asked to comment on an article from Thursday's Financial Times that endeavored to explain motives behind the big moves in mobile phone stocks over the last week. While well written, we found almost nothing in the article we could agree with but found it be an excellent summary of popular misconceptions about smartphones in general and Nokia/Motorola in particular. We present the article and our analysis below.
Zeitgeist: Barbarians at the Gate (March 7, 2009) Excerpt: My first introduction to the concept of handsets as commodities occurred in 1999 in a rapid fire debate with the head of a large technology fund who ran through the logic of why it was inevitable that handsets would follow the path of PCs and see margins plunge to low single digits. He had good reason for concern given that year saw Qualcomm-Sony, Philips and Lucent exit the business, Ericsson fall into losses and Motorola struggle to remain profitable after losing its lead to Nokia. Clearly mobile phones seemed to be following the curve of other consumer electronics where fat profits on early volumes disappear as manufacturing scale expanded and competition intensified.
Zeitgeist: Dance partners are being chosen at the baseband ball. (February 14, 2009) Excerpt: In the next several weeks, Freescale will announce the divestiture of its cellular business in what will be another major step in the ongoing consolidation of the baseband industry. The genesis of this divestiture goes back to Chicago 2000 when a small group of Motorola engineers began work on a credit card sized device that would become the RAZR. That phone’s success redefined the definition of style in handsets and set in motion events that would lead to the consolidation we are now witnessing.
The Bigger Picture: The Whiplash Effect of Excess Inventory (January 31, 2009) Excerpt: On October 12, 2008, we described a scenario in which a crescendo of bad news and preannouncements through the December period would culminate in a market bottom around February 1, 2009 ("Finding a Stable Bottom"). Our thesis has generally proven to be true even if the pace of events has been faster than we have anticipated.
Deconstructing the Decline: Separating inventory burn from declining demand in 4Q08 reports. (January 20, 2009) Excerpt: For the first time in history, the number of cell phones sold in the fourth quarter was lower than in the third period. Fourth quarter sales, which are typically up 20% sequentially, ended the year about 2% lower than in 3Q. Worse still was the timing. The freefall in late October hit manufacturers after the annual production ramp for the Christmas selling season had already begun, leading to bloated inventories even after production was cut-back in November.
Finding a Stable Bottom. (October 12, 2008) Excerpt: Now that the market has broken anchor from fundamentals and is in freefall, investors are wondering when we will see a stable bottom. Because failures in the financial sector and concerns about the economy are fueling the sell-off, it won’t be remedied by positive data points out of any of our companies in the short term. Indeed, the only thing that can stem the tide at this point is an end to the cascade of defaults and the collapse of confidence in the credit markets.
Smartphone Wars: Google’s clone enters the fray. (September 27, 2008) Excerpt: So begins the Android media juggernaut, Google’s formal entry into the Smartphone Wars, which have become the industry’s black hole consuming all media and emanating little reasoned critique. And while the Google faithful come off as deists compared to the Apple jihad, we believe expectations for the proliferation and impact of Android are optimistic. There are several major pitfalls that are routinely overlooked or discounted by new entrants in cell phones, especially those lacking experience in industrial grade communications systems and radio networks. That said, we believe the open architecture and flexibility of Google’s approach give it advantages that will help Android struggle through the rough road ahead.
Google, Inc. – Special Report: Google enabled phone. Devil is in the details but potential exists for big gains. (November 5, 2007) Excerpt: Marketing splash aside, it is too early to tell how the Google enabled phone will differentiate itself or how popular it will become. If Google concentrates on delivering seamless local based services (LBS), it could see dramatic gains. Location information is inherently valuable to mobile users, but is difficult to deliver seamlessly given the variety of technologies involved and the requirements for a visual interface.
Multimedia Phones - Revisited: Review, Update and Outlook Since Our 2004 Report (October 6, 2006) Excerpt: Two years ago we analyzed multimedia functions (pictures and music) in cell phones using their impact on handset mobility as a framework. We argued that pictures or music would be popular with subscribers only to the extent their utility in a mobile world outweighed their impact on the cost, size and power consumption of handsets. We concluded that imaging would not be popular with most subscribers because making handsets more appealing as cameras makes them less appealing as phones. Music, on the other hand, complements mobile phones because much of the utility of a standalone music player can be achieved with little impact to handsets, and music, like mobile voice service, can be used on the go.
Multimedia Phones: A Quantitative Analysis of the Impact of Imaging and Music on Mobile Phones (September 10, 2004) Excerpt: After years of anticipation, the age of multimedia cell phones seems finally upon us. Phones capable of more than just talk are populating the product lines of all major phone OEMs which now offer models that support data, music, imaging and video. Like PDA1 phones before them, multimedia handsets represent the state-of-the-art in radio and semiconductor engineering, boasting more performance and features than any other handheld device. But while technology has made such products possible, it has not always made them profitable.
Bits & Bytes
(Back to Top)
SiC furnace companies should be clear beneficiaries of China’s wafer buildout. (December 20, 2023). Excerpt: Historically there was minimal demand for merchant market crystal growth machines because the three companies that supplied the vast majority of device-quality wafers (Wolfspeed, Coherent, Rohm) designed and built equipment internally. Custom building equipment allows wafer manufacturers to protect IP and make small adjustments to the crystal growth process, although the need to procure and assemble furnace components slows the pace of capacity expansions.
Labor stoppages could lead to more conservative December guidance for automotive semiconductor companies. (September 19, 2023). Excerpt: Given the growing prevalence of long-term supply agreements, most revenue would likely only be delayed, not lost, but automotive is a primary growth driver for companies such as Infineon, NXP and Onsemi so a temporary standstill would likely have a material impact on results.
Aggressive SiC buildouts increasing the risk of oversupply after 2025. (July 13, 2023). Excerpt: Indeed, SiC power device sales are expected to increase from more than $2B this year to more than $10B by 2030. However given the number of new fabs announced, it now appears the supply side of the equation is increasing faster than demand.
Expect a wave of M&A in GaN. (April 21, 2023). Excerpt: Although Infineon acquired GaN Systems in March 2023 to solidify its strong position in GaN, the transaction may very well ignite similar acquisitions among competitors. There are few roadblocks to consolidation because GaN companies are small (GaN Systems’ revenue was in the “low/mid double digit millions”) and because potential acquirers in the diversified analog space have posted robust financial results for the past two years and have ample resources for small acquisitions.
Foundries are the key to reducing GaN-Si costs. (April 18, 2023). Excerpt: A large portion of legacy power semiconductors are high-volume commodities and as such the leading suppliers are integrated device manufacturers where margins are driven by manufacturing scale rather than ASP growth. This necessitates the multi-year construction of new fabs and can amplify the negative effects of a downturn as utilization declines. Many GaN-Si companies, in contrast, are fabless and outsource production to a handful of foundries, including TSMC, UMC, Vanguard, and, more recently, Samsung.
Electric vehicles shifting from 400V to 800V is negative for GaN. (April 13, 2023). Excerpt: If automotive OEMs can get the same performance from SiC at a lower cost, there’s little question as to which technology will win in next-generation EV inverters. GaN does boast a higher operating frequency than SiC which allows for faster charging, but quicker charging can also be achieved by moving to an 800V architecture, which favors SiC.
Fab partnerships will be an attractive solution for sopping up 300mm power capacity in a downturn. (February 22, 2023). Excerpt: Long-term, 300mm production should protect the margins of leading suppliers and drive share gains against smaller players, but a supply glut is increasingly likely as shortages ease and demand wanes. New capacity won’t be entirely incremental as a portion is designated for replacing 200mm production and some companies are splitting 300mm production between power and analog. However each new fab can generate $1B-$2B and with seven new fabs hitting the market between 2021-2024, new capacity represents a material portion of the ~$25B power semiconductor market.
Apple WiFi effort is no threat to Broadcom near-term. (February 21, 2023). Excerpt: More recently the OEMs’ continued struggle to develop its own smartphone modem led to the transfer of most of the engineers working on WiFi to the modem group after it signed the three year supply agreement with Broadcom. With that agreement expiring this year, it now appears its on-again, off-again WiFi modem effort is on-again.
Centralized automotive software poses substantial risk to legacy OEMs. (February 16, 2023). Excerpt: As systems matured, many of the engineers that wrote the code retired or moved on, leaving the Tier 1 with legacy software that no one dared mess with lest it fail or require requalification. With the arrival of electric vehicles (EVs) things began to change. Newcomers like Tesla, with no legacy systems, took a more centralized approach to software.
Is Murata gaining share at Qorvo’s expense? (October 20, 2022). Excerpt: In the rush to be the first country with a 5G network, China rolled out equipment before the full standard was ratified. Known as 5G non-standalone (NSA), it transmitted data over 5G links but had to rely on the 4G network to control the phone. As a consequence, and for the first time in the history of mobile phones, the handset had to transmit on two bands at the same time.
SiC modules unlikely to become commoditized like silicon counterparts. (October 13, 2022). Excerpt: High wafer costs will always make SiC devices more expensive than silicon. However the efficiency of SiC delivers significant cost savings at the system level, and these savings increase at higher voltages where the performance gap between SiC and silicon widens. At high power levels, multiple SiC devices are used in modules to maximize performance.
Apple’s modem isn’t looking too good. (October 12, 2022). Excerpt: iPhone design cycles are about three years long and much of it is built on the modem. If modem design is in flux late in year two (our present position for iPhone 15) then there is not much hope for a phone in year three.
Recession beginning to impact long-lifecycle markets. (September 22, 2022). Excerpt: As demand falls, price elasticity increases, limiting semiconductor manufacturers’ ability to raise prices to offset higher costs. That was evident in consumer markets in the 2Q22 reports of many semiconductor manufacturers and we believe it will be manifest in industrial and automotive markets in 3Q22 reports, if not in results certainly in guidance.
II-VI/Infineon LTA another sign that wafer supply will remain tight. (August 23 2022) Excerpt: With demand for wafers outstripping supply, we expect device manufacturers will pursue more and bigger supply agreements over the next two years until they have a better idea of their internal materials capabilities. Securing external wafer supply also provides industrial, energy and automotive customers the confidence that device companies will have sufficient capacity to fulfill near-term design wins.
Wolfspeed’s options for funding ambitious CAPEX plans. (July 26 2022) Excerpt: While leverage is not excessive, we expect management wants to avoid piling on more debt until the company is approaching breakeven, which isn't expected to occur until CY23. However there are a few creative options to attain “bridge funding” until financials improve enough to allow for more conventional funding.
The rationale and roadmap for Wolfspeed’s accelerated CAPEX schedule. (July 21, 2022) Excerpt: Wolfspeed is pulling forward its CAPEX forecast to meet higher-than-expected demand for silicon carbide devices. Our roadshow with CTO John Palmour in June and follow-up conversations with management shed additional light on the company’s CAPEX plans for a new materials (wafer) factory and a second device fab once the Mohawk Valley facility is up and running.
Breaking down Wolfspeed’s updated CAPEX plans for Mohawk Valley. (July 19, 2022) Excerpt: Following Wolfspeed’s May earnings call and recent sell side conferences, there has been confusion regarding the company’s CAPEX plans as recent communications suggest the company will have to spend more than previously expected to reach FY24 and FY26 guidance.
GaN-Si vs. SiC: A stubborn rivalry or a runaway victory? (March 30, 2022) Excerpt: Peak intensity probably occurred in late 2018 when MACOM claimed it was “already in production,” “shipping material revenue” and that it was “quite shocking to see the unit volumes.” By contrast the SiC camp said little about GaN at all. This led many to believe the massive cost advantage of silicon substrates over silicon carbide (~20 times cheaper) was more than enough to make up for the lower performance of silicon-based GaN devices, at least for 5G base station OEMs which were on the front lines of adoption.
Skyworks tops the list of likely suitors for Wolfspeed’s GaN-SiC RF operation. (July 14, 2022) Excerpt: Given the federal government's extensive funding of Wolfspeed's R&D efforts and the fact that GaN-SiC provides massive benefits to defense radars and electronic countermeasures, we don't believe the company would ever be allowed to sell its GaN-SiC operation to a non-U.S. buyer. The government's intervention to block the sale of the device and materials business to Infineon in 2017 is overt evidence of that reality. As such we believe the most likely targets for this asset are Skyworks, MACOM and Qorvo.
Sale of Wolfspeed’s GaN-SiC RF operation likely. (July 7, 2022) Excerpt: The company has already shed its non-strategic assets (lighting and LEDs) and is now operating as a pure SiC product company (materials and devices). The materials business is arguably the most valuable and strategic asset, so we see zero chance of an M&A event there. However unlike materials or power devices, the GaN-SiC RF device business is increasingly mismatched to Wolfspeed's strategic targets and CAPEX investments.
Apple's modem is not a failure and the release schedule isn't a surprise. (June 29, 2022) Excerpt: We have always believed Apple's aspirations in modems were overly ambitious and would result in it missing its original target of 2022. At the same time, our checks have consistently indicated the company is hell-bent on ridding itself of Qualcomm and has made its modem effort the highest investment priority in the company for several years.
C-band build out in North America won't be nearly as lucrative as it was in China. (June 28, 2022) Excerpt: U.S. carriers may be slaves to the administrative state (FCC, FAA) but they still must answer to investors, so money is an object when gauging how much and what kind of 5G infrastructure to install. A year into Verizon's and AT&T's C-band deployments and it’s imminently clear the days of large scale mMiMo networks are over.
Considerations for a spinout of Wolfspeed’s RF GaN device business. (June 21, 2022) Excerpt: Because Wolfspeed supplies GaN devices to radar applications which are critical for national security, non-U.S. firms are all but precluded from acquiring the GaN RF business. This was made clear in 2017 when Infineon attempted to buy Wolfspeed but was nixed by the CFIUS on national security concerns.
RF content growth looks good, unit demand, not so much. (June 16, 2022) Excerpt: The more we dig into the details of 2022 flagship phones, the more confident we are in RFFE content growth, particularly for iPhone 14 and Samsung’s A series models. That’s good news for largest module suppliers (Broadcom, Qorvo, Skyworks). Unfortunately, the more we dig into the economic picture, especially in the U.S., the less confident we are in unit demand,
especially at Apple which is the world’s leading supplier of luxury smart phones.
Prevalence of trade secrets in SiC materials a barrier to new entrants (June 7, 2022) Excerpt: To prevent the unwanted spread of information, some companies compartmentalize their operations so that crystal growth employees don’t interact with those from the epitaxy department, et cetera. While trade secrecy is
commonplace in high tech manufacturing, we expect the greater-than-usual degree of confidentiality among SiC wafer companies in the U.S. is partly due to early influence from the government and military, where national security took priority over economics.
Recent SiC power device TAM estimates in conflict with revenue targets. (May 17, 2022) Excerpt: Virtually all major SiC device manufacturers are expanding production but a large portion of this incremental capacity won’t come on until 2023 or later, so conflicting market and revenue estimates will remain a point of contention until the TAM targets being published today are 12-18 months out.
The difference between Qorvo and Skyworks: Part 1. (May 12, 2022) Excerpt: It's always tempting to think of Skyworks and Qorvo as interchangeable, especially after
reading their 10-Ks, which sound like they were drafted for the same company. The stocks often reflect investor's penchant for this grouping but it isn't always the best approach. They share a lot of the same technologies, products and customers, but their cultures and histories
are quite dissimilar and those have led to two very different firms on the ground.
GaN-Si vs. SiC: A stubborn rivalry or a runaway victory? (March 30, 2022) Excerpt: Peak intensity probably occurred in late 2018 when MACOM claimed it was “already in production,” “shipping material revenue” and that it was “quite shocking to see the unit volumes.” By contrast the SiC camp said little about GaN at all. This led many to believe the massive cost advantage of silicon substrates over silicon carbide (~20 times cheaper) was more than enough to make up for the lower performance of silicon-based GaN devices, at least for 5G base station OEMs which were on the front lines of adoption.
All the right stuff in all the right places: SE3 uses iPhone 12’s RFFE. (March 10, 2022) Excerpt: More detailed (and revealing) technical documents have yet to be filed with the FCC but from we’ve already seen it seems clear the SE3 is using the same RF front-end (RFFE) employed in iPhone 12 but without mmWave. This is typical. Rather than invest millions in developing yet another custom RFFE, Apple usually reuses an older RFFE which speeds time to market.
As conflict intensifies, so does demand for GaN-based radar. (March 8, 2022) Excerpt: The vast majority of GaN-based radars are manufactured in the U.S. and for obvious reasons are considered highly sensitive to national security, so historically the U.S. was hesitant to export the technology, even to allies. However over the past two years, the government has shown it is more open to exports, evidenced by Raytheon’s $2.3B contract to supply seven GaN-based THAAD missile defense systems to Saudi Arabia.
China's move to 5G SA likely to pressure content growth in domestic phones. (March 3, 2022) Excerpt: When the impossibility of the endeavor finally dawned on the government, it quietly allowed carriers to deploy non-standalone (NSA) equipment. That was great news for RF semiconductor suppliers because phones in NSA systems carry substantially higher content than in SA. NSA requires two transmitters operating at the same time (called "EN-DC") while a SA system uses only one.
RESN, MRAAY: Murata's purchase of Resonant does nothing to stop its decline in RF modules. (February 15, 2022) Excerpt: We never believed the filter EDA tool model was viable but the little hope it did have was extinguished with the Murata investment given none of the largest filter producers (Qorvo, Skyworks, Broadcom) were likely to work with a competitor’s vassal and the Chinese had so little share of the RF component market their volumes were immaterial.
NVDA: Collapse of Nvidia/Arm deal could mark the end of a highly-acquisitive era for semis. (February 11, 2022) Excerpt: With the recent correction in semiconductor stocks and fears of peak semi cycle growing with each new $10B fab announcement, a repeat performance of the past few years may prove difficult. This pressures equity-based transactions in two ways, as acquirers will have to dole out more shares while target companies may be more hesitant to accept stock-heavy offers if markets remain volatile.
SHWDY: Momentum likely to continue for Showa Denko’s SiC materials business. (January 20, 2022) Excerpt: Historically, Showa Denko applied epitaxial layers to silicon carbide crystal wafers purchased externally, and five years ago it acquired the SiC operations of Nippon Steel-Sumitomo Metal (NSSMC) to bolster development of SiC boules and crystal wafers. Prior to the acquisition, NSSMC struggled to develop cost-effective, high-quality wafers, and five years after the acquisition it seems Showa Denko has made progress in solving the problems which plagued NSSMC.
VZ: Audience for Verizon's FWA isn't worth the cost of production (January 4, 2022) Excerpt: In 2019, Randall Stephenson, AT&T’s CEO, panned Verizon’s plan as uneconomical. Having tested the technology and reviewed its abysmal history (“wireless cable" Local Point Distribution Service (LMDS)), AT&T had seen enough of the story to know how it ends. It decided discretion was the better part of 5G and limited its deployment of mmWave to just enough to say it had it, but no more. That didn’t stop Verizon however.
TSM: Hard times spawn goofy ideas. (December 9, 2021) Excerpt: The design and implementation of complex devices like MCUs is difficult and the time required to get a chip fabricated after a design is complete is substantial. This increases the cost of mistakes and their impact on production schedules. Without extensive experience in custom chip design, OEMs will likely spend vastly more time and money on an internally developed part than if they hired an experienced firm to do the design or purchased the component on the merchant market.
SOI: More SiC partnerships for Soitec not necessarily a positive sign. (December 3, 2021) Excerpt: The recently announced Mersen partnership is focused on polycrystalline SiC (poly-SiC) wafers, which have been generally unsuccessful in commercial production. Poly-SiC is not new; pundits have been claiming it is an effective method for reducing the cost of SiC wafers as far back as 1999, yet it has no material share of the market for wafers or devices.
VZ, T, TMUS: Forget higher speeds, how about less "zero speed"? (November 30, 2021) Excerpt: Hundreds of billions have already been spent on 5G and as of yet, nobody has figured out how to pay for it. 4G pushed data rates above those required for live streaming, video and music, so without some brilliant new application that has subscribers throwing dollar bills at their carriers, adding more speed doesn't add more value.
ON, CREE: GTAT's sale to Onsemi betrays its bullish claims of prowess in SiC. (August 26, 2021) Excerpt: GTAT began research on furnaces to grow SiC boules in the 2000s but didn't begin growing material with the intent of selling it on the merchant market until 2016. By contrast, the three companies that account for over 90% of the SiC wafers have been growing materially for nearly 30 years and the most recent arrival, II-VI, for more than 20. Silicon carbide material growth (boules) is exceedingly difficult and the market for crystal wafers made from boules is dominated by three companies because all others failed or have perpetually struggled with low yields and high cost.
QRVO, SWKS, QCOM: Who benefits from surging demand for Honor phones? (August 25, 2021) Excerpt: Under Zhixin, Honor regained access to Google, TSMC (basebands) and U.S. RF components. As a result, the brand is now thriving. Unlike under Huawei, Honor now produces high end flagship devices like the Magic 3 which is nearly indistinguishable in form and function from earlier the Mate and P series devices. However, unlike its predecessors, Magic 3 is powered by Qualcomm's Snapdragon 888 instead of HiSilicon’s Kirin platform.
5G Telecom: As goes 5G MiMo, so goes GaN. (August 24, 2021) Excerpt: MIMO base stations can deliver high bandwidth in densely populated areas but lack the range of macro base stations and are substantially more expensive, precluding widespread deployment. China’s buildout in 2020 targeted dense areas, which spurred the deployment of GaN-based MIMO base stations. With the more populous areas now adequately covered, China is shifting to macro base stations for suburban and rural areas.
Google determined to dump Qualcomm but not by developing its own modem. (August 19, 2021) Excerpt: Like Apple, Google believes differentiation in phones is increasingly tied to the application processor and less so to the baseband. It’s also of the opinion that to maintain its lead in advertising, it needs more direct control of Android phones. To that end, it shifted its strategy for Pixel from a low-volume, aspirational phone to a high-volume iPhone contender.
Nothing is quick with SiC. (August 17, 2021) Excerpt: Fear of new suppliers upsetting pricing in wafers is justified in silicon, GaAs and SOI but not in SiC because unlike all other merchant market semiconductor materials, doing anything in SiC takes a long time and doing it well takes even longer. Cree and II-VI, the world’s leading wafer manufacturers, both announced 8-inch wafers in 2015 yet won’t be entering production until 2022 and 2024 respectively.
GaN mix likely to shift from telecom to radar in 2021. (May 18, 2021) Excerpt: Much of China’s 5G deployments over the last two years were massive MiMo base stations in densely populated areas that required copious data capacity. Now that urban areas have largely been built out, China’s focus is shifting to rural and suburban which is to say from capacity to coverage which is more economically achieved with macro base stations.
VZ: Verizon offering $500 for clients to take its mmWave fixed wireless access (FWA) service isn’t a sign of strength. (April 22, 2021) Excerpt: Ultimately the physics of data transmission over radio waves prevents any FWA service from competing with data service over wires. This is why after forty years of cellular service over a network of hundreds of thousands of cell sites with high ample coverage in high-density urban environment, fixed wireless access has no material share of home data service.
Huawei: If it walks like a duck and quacks like a duck, it’s still Huawei. (March 23, 2021) Excerpt: Unisoc recently hired HiSilicon/Futurewei’s entire 50-engineer modem and transceiver design team in North Carolina, which recreated Huawei’s HiSilicon entity in everything but name. Using HiSilicon and Unisoc IP, Unisoc is now ramping its second-generation 5G platform, the Tiger T7520, which compares favorably to Qualcomm’s Snapdragon 855 and the former HiSilicon Kirin 990 and could challenge Qualcomm’s best in a year or two.
AVGO, QRVO, SWKS: Will there be an all-in-one, low/mid/high-band PAD? (March 12, 2021) Excerpt: Between the unwillingness of mid/high-band PAD suppliers to integrate SAW filters and SAW filter suppliers’ inability to field competitive BAW we find it highly unlikely any OEM would ever move to a single module that combines them all.
Huawei: From Premium Phones to Pig Farming. (February 23, 2021) Excerpt: Recent reports that Huawei has begun expanding its technology into alien industries such as pig farming and coal mining illustrate the pain brought on by U.S. sanctions, but also highlights the absurdity of claims in 2019 that the OEM could rapidly develop an RF front end solution to displace Qorvo and Skyworks. Huawei proclaimed it could achieve self-sufficiency in RF semiconductors, which was parroted by the media and sell side alike, but the company’s statements ignored the inherent difficulty of developing advanced RF modules. Bringing a functional product to market has become increasingly difficult for RF semiconductor suppliers over the past decade, and the supplier base has diminished from more than 10 companies during the 3G era to four in 2021.
Samsung, AAPL: Galaxy S20’s short life suggests impending trade-offs between revenue and margins. (February 17, 2021) Excerpt: Strong results for the iPhone 12 implies continued life for big ticket models but sales may have been skewed by lower replacement rates over the last several years in anticipation of the first 5G iPhone. Between upside demand for iPhone SE2 last year and Samsung's experience with the S20 and lower pricing for the S21, we expect ASP pressure in handsets and the ripple effect on BOM cost is here to stay.
MTSI: MACOM (MTSI) has substantial ground to cover in GaN-on-SiC. (January 26, 2021) Excerpt: This is MACOM’s latest move in its shift to GaN-on-SiC, which began in January 2020 when it effectively euthanized its GaN-on-silicon aspirations. The announcement is a positive development for MACOM’s GaN business but does little to narrow the gap between it and Qorvo, Cree, Raytheon, and Northrop Grumman. All four U.S.-based incumbents began developing GaN MMICs 20 years ago, and in 2013, the AFRL turned this specific GaN process (0.14μm MMICs) over to these companies to accelerate research and production.
QRVO, SWKS, AVGO: 5G shaking up market shares in RFFE. (January 14, 2021) Excerpt: As 4G evolved, RF component suppliers concentrated on and have come to dominate different sections of the RFFE. Broadcom is heavily weighted to the main path at Apple and Samsung, while Skyworks is the king of the diversity section. Qorvo leads in the main path in Chinese 5G handsets but hasn’t had much of a role there or in the diversity sections at Apple or Samsung. That appears to be changing with the arrival of EN-DC, more specifically, the use of EN-DC on Band 41.
VZ: More WiFi spectrum dims the prospects for Sub-6 GHz 5G, and extinguishes them for mmWave. (December 22, 2020) Excerpt: Given most people consume most of their data indoors, any wireless technology hoping to service that demand needs to penetrate walls, or paper, or cloth, all of which can obstruct user terminals indoors. WiFi 6E can do all that, mmWave can do none of it.
Resonant: Fighting a losing battle in a declining market. (November 23, 2020) Excerpt: Between a flurry of press releases and a recent shoutout from Murata, Resonant Inc. (RESN) is attracting interest (and generating questions) about its potential to displace incumbent filter suppliers in handsets. We believe the chances of that occurring are effectively zero and a more likely and imminent outcome is the exhaustion of its balance sheet and/or a fire sale to Murata.
VZ: Verizon's new spectrum is good, not great, but still better than mmWave. (October 8, 2020) Excerpt: At $1.9B, Verizon outspent rivals in the recent Citizen Band Broadcast Service (CBRS) auction, capturing 557 PAL (Priority Access Licenses) versus Dish, which took second-place with ~5.5K licenses for $900M. The difference in cost per license reflects the quality of the geographies, but it also speaks to the difficult spectrum position Verizon finds itself in for 5G.
VZ: The beginning of the end of mmWave at VZ. (September 17, 2020) Excerpt: Verizon was a big winner in the recently concluded CBRS auction, grabbing 40MHz of 3.5GHz spectrum that will go a long way to fielding a 5G service consumers can actually find. It’s struggled along with mmWave because that’s what it had, not what it wanted.
Verizon: When all you have is a hammer… (September 1, 2020) Excerpt: Verizon is on a quest to be the king of 5G and will chase that goal with mmWave no matter how many red flags it needs to push aside to get there. In January, the company announced it will double the number of cities with mmWave cells from 30 to 60 and increase small cell deployments by a factor of five.
BBB tells Verizon to knock off the 5G hype and get real with its ads. (July 16, 2020) Excerpt: On Tuesday the National Advertising Division (NAD) of the BBB asked Verizon to stop claiming its 5G service is widely available in cities across the country, and is broadly and readily accessible in cities where it has been launched. As the referee for the self-regulating body, NAD's requests don’t carry the weight of law, but they're a high profile reprimand that Verizon will comply with.
Saving face with the Mate. (July 9, 2020) Excerpt: It appears Huawei’s Mate 40 will be more about saving face than selling phones. Having become a political football in a U.S. / China grudge match, Huawei is cut off from the chips and software it needs to field flagship phones, but has to uphold national pride by releasing high-end models. It’s an impossible predicament that’s leading to some odd products, chief among which is the Mate 40.
mmWave costs vastly more than it’s worth. (June 9, 2020) Excerpt: The cost of supporting 5G millimeter wave (mmW) bands is about 10X that of sub-6 GHz, so the increase in ARPA (average revenue per account) for mmW service must be commensurately higher otherwise carriers will abandon it for sub-6 GHz when it becomes available. The problem of course is that it isn’t.
What happened to Verizon’s wireless fixed broadband to the home story? (May 27, 2020) Excerpt: In 2018, Verizon spoke at length about a brave new era of fixed wireless broadband (FWB) to the home. Since then, we have not heard a peep about progress, much less success. We expect this is because trying to use RF to compete in the high-speed data access to the home market is a fool’s errand.
Most 5G bands don't need BAW, but there are exceptions and they're expensive. (May 19, 2020) Excerpt: China’s requirement that N77-N79 bands be included in 5G phones will oblige OEMs to add one to two more modules to 5G phones this year. The impact on content led Qorvo to double its 2020 5G TAM forecast last October and has many investors asking about the ripple effect on BAW filters given the new bands (N77, N79) are well beyond the 3.6GHz N48 band already planned for 5G and the 2.4GHz threshold for SAW.
|
|