NXP Semiconductors N.V. (NXPI) Earnings Call Transcript & Summary

Thanks, Seb, and good morning, everybody. Thank you for joining us on this call. I'm Ross Seymore, the semiconductor analyst here in the U.S. for Deutsche Bank. The automotive market is a very exciting one for semiconductors with content growing significantly over the years. This year is a little more challenging with the COVID impact on the unit side of the equation, but that content gain story is still one that's very powerful and surviving and thriving, even despite the current headwinds. So within that content story, one of the bigger drivers is ADAS. And within that, radar is an incredibly important driver. So that's the topic of today's call. And we're very fortunate to have Torsten Lehmann, the Executive VP and GM of NXP's RF Processing unit; as well as Jeff Palmer, the Vice President of Investor Relations. So I'm going to pass it over to Jeff to give a little introductory comment, and he's going to go to Torsten to go through some slides. After that, I'll have some questions. And then we'll open up to the queue for investors to ask questions, so please log in. This is a great venue for you to ask questions, so get in the queue starting now. And with that, let me pass it over to Jeff.

Thank you, Ross, and good day to everyone on the call. I'd like to welcome you all to the first installment of the NXP Teach-In Series. There will be other topical calls in the near term, so keep an eye out for your news flow. As Ross mentioned, today, we'll present an overview of the automotive radar market and what we view as our unique opportunity. The call is being recorded with a replay, along with the associated slides that will be available on the NXP IR site within 24 hours. Please be aware, today's call will include forward-looking statements that involve risks and uncertainties that could cause NXP's results to differ materially from management's current expectations. These risks and uncertainties include, but are not limited to, statements regarding the evolution of the automotive radar market, the sale of new and existing automotive radar products and our expectations for the long-term growth in the automotive radar market. For more information on forward-looking statements and more complete description of these risks and uncertainties, please refer to the slide titled Forward Looking Statements in the associated deck, as well as risk factors listed in our SEC filings, copies of which are available on the NXP IR website. Please be reminded NXP undertakes no obligation to revise or update publicly any forward-looking statement. I'd now like to pass the call over to Torsten. Torsten?

Okay. Thank you, Jeff, and good morning, good afternoon, everybody, and thank you for joining us in our automotive NXP radar teach-in session, and thank you to Ross and the Deutsche Bank team for hosting us today. So before I jump into the content, I would like to highlight that in the following presentation, there are 3 major topics that I'd like you to take away about the radar market and about NXP. First, the radar market keeps growing strongly in the foreseeable future. Secondly, NXP is very uniquely positioned within this opportunity as the #1 semiconductor company, winning through a fully scalable system portfolio and through a very close collaboration with carmakers and Tier 1 customers. And thirdly, as the market is moving towards 77-gigahertz radar systems and RFCMOS technologies, NXP will keep outgrowing the market, and we will expand our #1 position. I will obviously elaborate more on each of these items as we go through the presentation. Now let's get into the content matter of radar technology for driver assistance and for automated driving. Every year, about 1.3 million people get killed in car accidents, and moreover, an estimate additional 20 million to 50 million people suffer from nonfatal injuries out of accidents. Radar technology is a key enabler to make driving safer and to prevent accidents and saving lives. So let me explain NXP's system view of a radar node. Each radar node is a full system containing a number of key semiconductor components, as depicted here on the slide. At the heart of the radar functionality, on one side, the radar transceiver, and on the other side, the radar processor, both on the left-hand side of the slide here. Radar transceivers have evolved and integrated over time. And until recently and today in production, each of those small blocks that you see depicted in that transceiver block where individual ICs, like the receipt channel, the transmit channel, the VCO, et cetera. But today's state of the art in new designs are fully integrated transceiver chips. And also on the radar processor side, this keeps evolving, going to deeper submicron process nodes and integrating more processing power, dedicated signal processing accelerators, higher levels of security as well as functional safety up to ASIL-D level. And then moreover, radar systems additionally include auto peripheral or attached components, such as safe power management and also in-vehicle networking components, such as CAN and Ethernet, where NXP is #1. This is our definition when we talk about radar nodes or radar systems, and NXP is one of the very, very few semiconductor players fully addressing this complete application opportunity. Now let me give you an updated market view right at the beginning. Some of you might have joined our NXP Investor Day 2018. And just to compare the numbers from back then to our latest view, which, obviously, also includes the latest bar projections in view of COVID-19 and global trade tensions, clearly, the car production has evolved very differently than what was anticipated mid-2018 with actually 2019 declining some 6% and then 2020 expected down another 20% to 21% due to the COVID pandemic, and these numbers are based on IHS data. Yet what holds true is our hypothesis of NXP's growth above market. We said back in 2018 that we would grow at 1.4x market CAGR, and we have grown share in the last years, and we will continue to do so. So on the right-hand side, forward looking, we do expect a very strong 20% CAGR of the radar market, reaching some $2.2 billion, of course based on latest IHS car production forecast. Our market outgrowth hypothesis stands. We have grown share in the past and in the last years, and we will continue to grow radar share and expect to grow faster than market with 25% to 30% CAGR over the forecast horizon. While this year, radar is already contributing around about 10% of NXP's automotive revenue. Actually, the vast majority of our 2023 revenue projection is already underpinned with a lot of designed ins today. Now zooming out a little bit to the bigger picture of ADAS and automation. Most of you will be very familiar with the SAE levels of driving automation, and these are evolving from level 0 with basically no automation to driver assistance, then to partial and conditional automation and eventually all the way to full automation at level 5. Basically, 2 very different segments can be distinguished in this picture. And that's, first, the classical vehicle ownership. So this is the classical car OEM space, mainly focusing on the levels 1 to level 3. And then secondly, different segment, mobility as a service, and that is primarily the level 4, level 5 robotaxi players. This is where Waymo, Baidu and many other mobility service providers are active. And while certainly, technologically super exciting, we do not expect the mobility as a service to be prevalent for the next few years that we are discussing here today, at least from a semiconductor content perspective or semiconductor dollar perspective, let's say. So by 2023, we do expect that about 60% of all new vehicles will have some sort of driver assistance. And while the vast majority of this is still in the level 1, 2 space, we do see the car OEMs pushing it further in the direction of level 3. And we also see an important new level 2+ evolving, which essentially offers level 3 kind of features and driver experience while avoiding the liability aspect of a full-blown level 3 by requiring the driver to keep the hands on the wheel. And translating this into semiconductor content, about 99% of the semiconductor value are in these levels 1 to level 3 cars in the 2023 time horizon. As said, of course, level 4, level 5 and mobility as a service provides another great incremental opportunity in the long run, but it will not significantly move the needle in the time horizon until 2023 that we are discussing here, at least not from a semiconductor dollar perspective in that time frame. Now key driver for the broad adoption of ADAS technology, in general, and specifically also of radar technology are legislations in different countries like mandatory emergency braking, for example, as well as evolving regional NCAP guidelines requiring more and more safety features for a 5-star safety rating. As we look to 2023 and beyond, which, in automotive terms, is really just around the corner, then many car OEMs are moving towards level 2+ systems. Level 2+ enables a significant step-up of safety as well as comfort, so features like, for example, autonomous highway pilot while avoiding the liability concerns for OEMs by requiring drivers to keep their hands on the wheel. So in principle, we can say the level 2+ car is able to fully drive by itself in defined situations, like on the highway, but the driver is not allowed to let go of the steering wheel completely, and the driver has to be at all times in the position to take back over. The nice thing from a radar perspective and specifically from a semiconductor value perspective is that the content of level 2+ is very similar, if not the same, as for level 3, at least on the sensor side. And speaking of ADAS sensor, the key enabler for advanced ADAS systems and higher levels of automation, are multiple complementary sensors around the vehicle that accurately sense and detect and monitor the surrounding of the car, such that you can, for example, reliably detect a child playing on the side of the road. The most widely deployed in ADAS systems are radar and camera as they make a very reliable and complementary combination. Sometimes, also LiDAR is used in vehicles, especially for higher automation levels, though we think the economics and the system side continues to limit very broad mass market adoption of LiDAR in the midterm. And each sensor technology has certain strengths in terms of which kind of measurement it is best at. So as an example, LiDAR is absolutely great to very accurately measure speed and distance of up to 300 meters with a very few centimeter of range resolution, while a camera is obviously great in terms of objects or color detection. And on the other side, most cameras have certain environmental limitations. So as an example, a camera can be blinded by bright light or not see at night at all or with fog or snow or struggle with reflection on wet roads, et cetera. And likewise, a LiDAR sensor can also be disturbed by certain weather conditions. Radar, on the other side, is very robust and therefore most fundamental technology. It is very robust against any kind of environmental influences, and radar does not care if it is night or fog or bright light, et cetera, and operate very reliably under those circumstances. And radar keeps evolving, so radar is evolving beyond the detection of big, bulky objects, where, in the past, it would just essentially see other cars, and it's evolving beyond the measurement of just speed and distance. Radar is moving towards precise mapping of the whole environment around the vehicle, and several technological ingredients are enabling this evolution, like the transition from 24- to 77-gigahertz frequency bands, as well as moving to RFCMOS, moving to higher-channel MIMO configurations as well as auto applying advanced processing solutions with dedicated accelerators. These technological moves, all combined, allow to build much smaller and more power-efficient modules while achieving a much higher range and angular resolution and thereby achieving distinct detection and separation of smaller objects in the distance which is very important. For example, seeing individual pedestrians in the distance or having your car distinguished between a motorcycle and a truck approaching in different lanes from behind and thereby enabling your car to do a safe, automated lane changing at full speed on the highway, so very crucial functionality for higher levels of automation. Now coming back to the latest market view and NXP's leadership position within this market. After the COVID-19 setback, the radar semiconductor market is expected to reaccelerate and grow at 20% CAGR over the next 3 years. At the same time, the transition from 24-gigahertz to 77-gigahertz radar systems is in full swing. While 77 gigahertz becomes the key growth driver, we do expect that the 24-gigahertz segment will start to decline on this time horizon. NXP is #1 in the market, which can be broken down into the radar core components like the radar transceiver and the radar processor. And on the other side, the radar attach or peripheral components, as we call them, like the in-vehicle networking and the power management, et cetera, as outlined in the beginning. And just the radar core components, so the transceivers and the processors, make up for around about 10% of NXP's automotive revenues in 2020. And NXP is the #1 in this field, ahead of a key German competitor A. There are multiple other competitors out there, but none of them with a market share anywhere close to NXP or even close to the #2 vendor in the market. We, NXP, had our first 77-gigahertz design wins already many years ago. Our designed-in pipeline is very well -- is very strong, and many of the platform wins that we announced in 2018 at our Investor Day are actually ramping right now as we speak. And keep in mind the typical nature of the design cycles in automotive, where design targets have typically a 2, 3 years lead time, and then production of this platform often runs for another 5 to 7 years. It is a dynamic that gives us a high degree of confidence. And another aspect is that the required R&D investment and the specific automotive radar know-how provide a very, very high-entry barrier to newcomers in this field. In terms of outlook, we are very confident in our continued share growth as we are the #1 in the fast-growing 77-gigahertz market, and NXP is also leading the RFCMOS transition. Now you might ask why is this radar market growing so much above car production and if it's real, and the answer is clearly yes because there are actually 3 growth vectors at play here. First, as illustrated with the mandates and NCAP ratings earlier in the presentation, the penetration of radar in new vehicles keeps growing and is increasing, and that means we see more cars with radar on the road. And secondly, because of these new use cases, we need 360-degree sensing around the vehicle, and that means more radar nodes per individual vehicle. And then thirdly, with the trend towards higher resolution and towards imaging radar, we see more silicon content per individual radar nodes driven by more transceiver channels, which translates into more transceiver silicon, as well as advanced high-performance processors capable of computing that high amount of data generated by the many MIMO channels. And that means more semiconductor content and more silicon dollars per individual node. So that means when we talk about triple acceleration, this is actually a multiplication of these 3 effects, so it's more cars with radar times more nodes per car and times more silicon dollar per sensor. And the nice thing is and the good news is that we are just at the beginning of this s-curve. But today, we have, on average, around about 1 radar sensor per every vehicle produced globally. And of course, the reality is there are vehicles with 5 radar nodes, and there are vehicles without any radar node. But the average across every vehicle produced globally is just about 1, and this will reach about 2 by 2025. And this will continue to grow much more going forward, eventually approaching some 5 or more per vehicle in the long run, while additional dollar content growth per node comes on top. So as the radar technology keeps evolving, we get closer towards the vision and the goal of 0 accidents. And this is a very exciting motivation for all of us on the business side as well as on the R&D side. As touched upon earlier, radar is going through the transition from using the 24-gigahertz frequency band and applying silicon germanium technology on the transceiver side towards 77-gigahertz transceivers and advanced processors, both in standard CMOS technology going forward. With that, much higher resolution and performance can be achieved, lower power dissipation, as well as higher integration levels become possible. And in terms of use cases, this transition enables the car to safely detect vulnerable road users, such as pedestrians and bicycles. While in the past, radar could basically just detect other cars. Now going forward, we see a broader range of market segments evolving driven by 2 directions, as indicated here. First, on one side, the need for full 360-degree coverage around the car, and that means we really want and need a complete radar cocoon around the vehicle; and secondly, the need for much higher resolution, seeing and identifying and distinguishing between smaller objects in the distance. This leads to a segmentation of the radar market, as shown on this slide. Starting at the bottom with many small sensors around the vehicle, especially corner radar sensors but partly also additional gap filler or pre-crash sensors. And on the upper end of the spectrum, we see very high-performance imaging radar evolving, able to draw a precise environmental map of the surrounding and generating LiDAR-like images, if you will. And also the classical long-range radar keeps evolving with more and more performance, for example, 300-meter range with better and better resolution and increasingly also rearward-looking mid- to long-range radars with up to 200 or even 300 meters of range. And generally, better and better angular resolution is absolutely key in order for the radar to be able to distinguish small objects in the distance but especially also in lateral direction for judging if your car can safely drive over or under an object, for example, detecting the height of small objects on the road in 50- or 70-meter distance or knowing that your vehicle can safely drive underneath a bridge or through a garage door or into a tunnel opening, et cetera. So obviously, it's very, very essential. This means also, going forward, there's a lot more innovation to come which keeps our R&D teams busy and with technical evolution, as indicated on the far right-hand side of the slide. By the way, this is also a benefit of scale, having one of the largest automotive semiconductor businesses in the industry, that our running core business can fund the new growth areas. The key innovation topics here in radar are further miniaturization and up-integration on one side of the spectrum and even higher performance levels and new modernization techniques on the other end of the spectrum. And in the right block, I mean, you can see NXP is one of the very few players covering the full range of applications as well as all relevant components of the system for complete costs as well as performance-optimized system solutions. Our product families allow customers to optimize their total cost of ownership with maximum scalability as well as reuse across different radar systems and thereby optimizing their own R&D efficiency. To make this more concrete, here are 3 examples of different radar system applications. As you can see and as indicated on the slide, customers can use the same RFCMOS radar transceiver family across different segments, by the way, including also explicitly long-range radar, which some in the industry did not believe to be technically possible with RFCMOS until recently. We can tell you it is possible. Likewise, our scalable radar process of families allow for maximum architecture and software reuse across different applications. And as you might know, software is a key factor driving the R&D build. So this scalability and the reuse of R&D investments and IP assets translate into optimal total cost of ownership and is, in the end, a key selection criterion for our customers. And speaking of customers, one of NXP's core recipes for success in automotive, in general, is the so-called innovation triangle, where we collaborate very, very closely with the carmakers as well as very closely with the Tier 1 suppliers. On one side, NXP is a trusted adviser to many car OEMs for a wide range of automotive applications, and this is obviously not limited to just radar. Especially the co-definition of future architectures and the co-definition of road maps with the car OEMs is absolutely key for us. And then on the other side, NXP is a partner supplier to our Tier 1 customers with whom we deeply cooperate for fast time to market for high R&D efficiency and joint success in the marketplace. And as a result, we are very proud to have radar designed ins with 15 out of the top 15 carmakers globally and also beyond that. So in conclusion, the radar market opportunity is very, very exciting. The market is growing very attractively with 20% CAGR over the next few years. And remember, this growth is fueled by what we call triple acceleration, so more cars with radar times more radar nodes per car times more semiconductor content and silicon dollars per node. At NXP, we are expecting to outgrow the market with a range of 25% to 30% CAGR, very much enabled by our close OEM and Tier 1 NXP partnership triangle and respective joint innovation. And as a result, NXP will further expand its #1 position in radar, which already, today, contributes around about 10% of our NXP automotive revenue. We are very confident in our growth outlook as more than 85% of our 2023 estimated sales have already been awarded to us, and we do have designed in at 15 out of the top 15 carmakers globally. So wrapping this altogether, radar is a super exciting application, enabling us to help saving lives, preventing car accidents. And moreover, radar provides a fantastic market opportunity and continues to be a top strategic growth area where NXP will outperform the market by leveraging our technology leadership. And with that said, I would like to hand it back to you, Ross, for any questions that you or others in the audience might have.

Thanks for all that great color, Torsten. Seb, why don't you give detailed instructions for people to get into the queue, and then I will answer -- or ask some of the questions from my side while the queue builds.

[Operator Instructions]

Great. Thanks, Seb. And Torsten, why don't we start with a couple of market dynamics? In a couple of the slides you showed, it looked like the radar market was about the same size in 2020 it was -- as it was in 2018 and grew a little bit slower than your 2018 Analyst Day forecast. Is that simply due to the COVID impact on SAAR overall? Or has there been a change in the rate of adoption of radar over the last couple of years, whether it be because of the technology, the bill of materials, et cetera?

No. Actually, the radar adoption, as such, has picked up very well and very much as expected since 2018. So -- but what is clearly the case is that the market production has compressed, and the car production certainly hasn't developed as we all anticipated mid of 2018. And definitely, you mentioned COVID, I mean, huge impact this year, no doubt. But even pre-COVID, let's also not forget these whole trade tensions and so on has started to kick in. So looking at, for example, IHS car production numbers in comparison to last time, 2019 was already down some 6%. And then this year, it's probably another 20%, 21% down. So that is, of course, yes, also correlates then into the radar market. But if you take our old forecast and correct them for the car production, then actually you come to very similar numbers. Actually, the new forecast is even a bit better than the old one corrected for car production.

Great to hear that's still progressing. And on another one of your slides, you talked about, about 85% of the forward growth was already underpinned with confirmed design wins. Can you talk about that design win to ramp time line for this sort of technology? There can be a wide range of time-to-revenue ramps within the automotive market. Some of this technology is evolving a bit quicker, so it might be a bit of a shorter time to market in this instance. But the design wins that -- or the business that's ramping today, when were those designed in? And how is the design win traction changed, if at all, pre versus post-COVID?

Right. Yes. The typical dynamic is really, let's say, on average, 2 to 3 years lead time from design to revenue. And then it's oftentimes a 5- to 7-year production, yes, window following that. But of course, I mean, there's also a wide range, so this can depend a bit on the region. So if you look at, for example, some of the emerging China companies, sometimes these lead times are a bit shorter. They're a bit more aggressive. On the other end of the scale, some of the Japanese customers might be a little bit more conservative, maybe having some longer lead times, and then Europe, U.S. and so in between. So that's one of the dynamics. And then the other thing is, yes, we often do big platform design wins with -- this is typical 2-, 3-year lead time. But then, often, we get also incremental awards based on existing platforms, and those can happen a little bit shorter lead time to revenue. Yes. I think the second part of your question you were asking on the -- has a design win dynamic change, right? And I mean, generally not. So -- well, let's say, if we look at platform ramps and so on that we were expecting for this year, and this is still happening as expected, so the stuff that was planned to ramp second half of this year is ramping and so on. But of course, we have the car production dip in Q2 and in the first half of the year. Looking at new platforms, also most of what we see is really still on track with the same launch time lines of vehicle. However, maybe some of the decisions that the carmaker and also the Tier 1 customers have maybe a little bit slowed during the COVID because people who were not able to travel as much and sit together face-to face. But yes, we haven't seen really fundamentally time line shifting, all of that.

Great. And continuing the market dynamics, you had a slide talking about the kind of the triangle of your relationships with the OEMs and the Tier 1s. Can you just talk a little bit about where does the design win actually take place? Do you have to fully engage with both sides of those equations? Do the decisions really come from the OEM side? And part of the ADAS story that investors hear about, admittedly more on the computing and processing side of things, is that there's reference designs based around the big compute engine that will be in these future L2+ and beyond cars. Are there reference designs that NXP and any of the radar folks get embedded within? Or is the sensing side of the equation somewhat separate from that go to market?

Right. Yes. I mean, to the first part, it's clearly totally crucial to engage with both the OEMs and the Tier 1 customers. And it's really that the -- it's very important to closely align the road maps the overall vehicle architectures and some of the innovations very early on with the car OEMs. But having said that, it also depends a little bit on the OEM. So some OEMs take total control down to the component level selection, whereas others rely way more on the Tier 1. So that's not one general answer. Therefore, simply from an NXP perspective, totally crucial that we are very deeply engaged with both the OEMs and the Tier 1s. Another important aspect, especially in ADAS, is also that we see, of course, these systems evolving quite a lot. And this has been not only an impact on the sensing side but on the whole vehicle architecture in terms of the in-vehicle networking backbone that you need and in terms of the data rate that you need to shuffle back and forward and so on. And also from that angle, every generation usually touches many, many bits and pieces of the whole vehicle. And also, there -- yes, it's good that NXP has such a broad portfolio, stretching across the sensors, the gateways, the in-vehicle networking and so on and then thereby being a nice discussion and respected trusted adviser partner for the car OEMs. Yes. With respect to reference designs, not so much, I would say, at least not covering the complete system. I mean, of course, I mean, we and also other semiconductor players provide reference designs for a certain part of the system. But as I said, since the ADAS system basically touches upon each and every element of the vehicle because it's the sensing side, it's the actuation side, it's the networking side, it's the central compute and so on, yes, it covers almost everything. And that means, yes, you cannot really do that without deep engagement with the car OEM directly.

Great. Seb, why don't we go to the audience and start taking a few of their questions?

The first question comes from the line of Karl Kroeker from Woodline Partners.

Maybe just following up on a little more detail on Ross's question. You mentioned when beyond sort of 2023, we'll be increasingly moving to mobility as a service kind of like the NVIDIA-Mercedes agreement. As part of your sales process, do you have to include as part of your engagement with the system provider, like be it an NVIDIA or a Mobileye? Or would it stay simply with the Tier 1 and the OEM?

Yes. Speaking about the sensor side, since we're focused on radar here, it's primarily still with the car OEMs and the Tier 1s in that time horizon. And by the way, also, I mean, if I say OEMs, that would include, of course, also the mobility-as-a-service providers, where we are also deeply engaged with some of the most relevant Silicon Valley players, for example.

Got it. Okay. And then maybe just one other. You talked about a triple acceleration driving radar. For your market share, would you say it's more impactful to see more nodes per car? Or is it -- or do you think your share is more driven by more cars with radar? In other words, do you think your market share is better sort of in the level 2+ kind of radar? Or do you think the initial cars, the entry-level cars going with initial radar, is that more where your share is levered?

Right. Yes. And the answer, it's really all 3 aspects. But yes, you are right. I mean it's probably, over time, a bit -- I mean, let's say, initially, it was probably more the -- more cars with radar because in the very early stages, yes, it was primarily the front-facing, adaptive control, automatic emergency brake. And then over time, of course, the more nodes per car really kicks in strongly basically as we speak and then the more content per individual node. So that third boost is sort of just at its infancy because that's really more towards the level 3 and beyond, yes, which is just starting. But also, there, we have some very relevant designed ins.

Our next question comes from -- this comes from the line of J.P. Scandalios from Franklin Templeton.

Yes. You showed a slide where, as we move towards level 3 and level 4, 5, that there'll be also cameras and LiDAR content potentially in cars. Do you have products for those areas as well? Or will this business remain 100% radar focused over the long term?

Right. No, I mean, definitely, we see the biggest opportunity and the most attractive play for us in radar, and we see that absolutely essential and, yes, very much business relevant. Having said that, first of all, yes, we definitely see the need for complementary sensors, especially for higher levels of automation. So you really need to make sure you, under all conditions, can safely detect the environment and also have redundancy in the system. And that's where definitely we also see big play for camera and maybe, to a lesser extent, LiDAR. To your question, yes. We -- at NXP, we are #1 in the automotive processor space, and we are also selling into applications like camera in applications, and by the way, also things like in-cabin, for example, driver monitoring, et cetera. And we also have processor content in some LiDAR applications. But what we explicitly also don't have at the moment is the LiDAR upfront and so dedicated analog or, yes, laser components for LiDAR. That part we don't address.

Great. And then just real quickly, you mentioned within radar, there -- you've mentioned 3 buckets, the corner, the long range and the image recognition. Is it -- is the market such that it's a winner take all, if you will, where if you're working with an OEM or a Tier 1, you'll get designed in for all 3? Or could it be a situation where in a car model, you might have the corner and long range, but someone else has image recognition?

Right. Yes. If we look at today's situation, I mean, it can be that it's different vendors, so different semiconductor and different Tier 1s also for, for example, the corner versus the front radar. However, what does matter, and that is maybe more from a Tier 1 perspective, is that for the Tier 1, it makes a lot of sense to build various sensors across these different segments with the same silicon content. And that means Tier 1 customers that use our transceivers and micro control [Audio Gap] the development and then have a lot of reuse, stretching that across the different segments. That means, of course, having, for example, corner radar sensor with one of our Tier 1 customers also puts the likelihood very high that they would go on the higher-end systems with us. And, yes, secondly, that there might be a component as the vehicle architectures are changing over time that it also becomes more likely to have a pull-through effect across different sensor technology.

[Operator Instructions] Our next question today comes from [ James Burling ] at [ Worli Capital ].

So hoping you could elaborate on the content you expect to achieve in radar processing. To the extent that the Teslas of the world embrace sensor fusion, where it combines this -- or ingests data from radar post cameras, et cetera, do they bypass the radar processor and just ingest it in their GPU or GPU equivalent for their AI or inference? Or do they use your radar processor in conjunction, as a complement, to the solution set when they are ingesting data from a feedstock of sensors beyond just radars?

Right. That's a very good question, and you mentioned Tesla. I mean, typically, we can, of course, not comment on individual OEMs. In that particular case, though, I can hint you to an interesting publicly available teardown report from EE Times on the Tesla Model 3, where you will see in the teardown that our NXP radar components are included, actually including the processor, which probably answers a part of the question how they do that today. But I guess, underlying to your question is probably more the general trend, as obviously there will be more central compute. And of course, you really want your driving system to do more sensor fusion and get, let's say, raw enough data from the sensors, such that you can do meaningful fusion. The countering trend, though, is that as we go to this more advanced sensors, the sheer amount of data generated by, for example, an imaging radar sensor is easily 10 of gigabits per second, and you need a very low latency loop between your transmit and receive and the algorithms that you apply, and you need to adapt that very dynamically, depending on the concrete situation, such that also the processing requirement in the sensor node keeps increasing while definitely also the central compute keeps increasing. So it's not an either/or, but it's an end function. So it's more processing, actually, on both sides going forward.

Why don't I jump in with the next question? Torsten, I want to go back into the 77- versus 24-gigahertz side of things. Talk a little bit about the advantage of going to the 77? And then I have a few follow-up questions on that. And really, what I'm getting at is not the market share advantage, but the technology difference between the 24 and the 77, the step-up in content that it could offer and even some of the technological challenges of doing it in RFCMOS versus some of the competition doing it in more specialty materials like silicon germanium?

Right. Yes, so quite a few questions. So yes, I mean, first of all, going from 24 to 77 is from the sheer physics, I mean, going to higher frequency range. You can go to smaller antennas, smaller form factors, you get better resolution, et cetera. But that is not all. It's actually -- if you look at the frequency band that is allocated in 24 gigahertz, then it's just a very narrow frequency band. While in 77 gigahertz, the reality is, in many countries, actually up to 5-gigahertz wide frequency band from 76 to 81 gigahertz. And that means you not only go to higher frequencies, but you can also use more bandwidth. And with that, plus applying advanced processing, you can reach up to a factor of 20 better-range resolution. So that is very, very significant and very important then also for these advanced use cases that we're trying to achieve. Yes. Along with that, of course, the module sizes, and along with that also going to RFCMOS, going to smaller or lower-power dissipation to higher integration levels and so on. So those are all factors that drive this switch. And maybe want to add also is that there's also a legislation impact. So in Europe, for example, there have been multiple extensions, but there is a clear desire to also free up that 24-gigahertz frequency band from radar applications. And yes, from top of my head, I think the EU requests that from 2021 onwards, new cars shouldn't be produced with 24-gigahertz systems anymore, unless they already had a -- we're already in production or had already licensed previously. But in essence, I mean, this will be freed up over time and will also, from a regulatory angle, move to 77 gigahertz.

And on the RFCMOS side of things, Torsten, obviously, just from a holistic level, people familiar with the semiconductor industry know that the performance benefits of specialty materials is always a trade-off. When you go to CMOS, you can have lower power, lower cost, et cetera, the ability to integrate, but sometimes there's a trade-off between the performance side and those benefits. Do you believe there is any performance detriment by going to RFCMOS? Or have you guys overcome any issues that may have been there at one point?

Yes, the latter. So we have overcome that, and we master that. But you are fully right. I mean, first of all, it's really an art, and it's super hard to do, and not many people can do it. So doing really high-performance RF in CMOs and at 77 gigahertz is super challenging, and that's also why it provides a very high-entry barrier. And like with many other technologies, indeed, in the early days, oftentimes specialty processes provided an advantage because it is so hard to do. But in the meantime, we've overcome that. And with our latest-generation RFCMOS, we achieved better performance levels than what was previously achieved with silicon germanium. And that means once you get to that point, yes, there is really 0 good reason to do silicon germanium anymore. Because RFCMOS -- I mean standard CMOS have the -- all the benefits of 12-inch standard processing and all the economies of scale but while providing the advantages of much lower-power dissipation, much higher digital content integration and, yes, better cost efficiency and everything else.

And with those benefits that you have and the performance now being on par or even superior to some of the specialty materials, as you move from the corner radar to the long range, as you had in your slide, and then eventually to the imaging radar, is the ability of what you guys can deliver via radar actually replace the need and remove the need to have other sensors such as LiDAR? How do you think about the complementary aspect of those other sensors versus the substitution of them with your technology?

Right. No. I mean we definitely believe in the need for complementary sensors. There's no doubt. And also, as I said, I mean, we definitely, at least, usually see a combination of radar and camera, and there could be also systems with LiDAR on top as we go to very high levels of automation. Having said that, though, yes, I mean, we are approaching really LiDAR-like resolution with imaging radar sensors and similar angular resolution. And that means depending on which automation level and which amount of redundancy and complementarity you want or need in the system, yes, you might be fine with radar and camera. So yes -- and radar provides certain advantages over LiDAR with respect to being less vulnerable to environmental aspects. And also typically, they are more cost-efficient than smaller, et cetera. But yes, I mean, having said that, there is a play for the various sensors, so it's not our ambition to sort of kill any other sensor, but we see very positive prospects for the radar technology on a very broad adoption basis.

And sticking with the competitive intensity side, in one of your slides, you talked about, I think, competitor A or company A has a significant amount of market share, and then it's a big drop to a whole bunch of others. This is clearly a very attractive market, not an easy market to get into, but an attractive one. And the semiconductor space tends to attract a number of aspiring new entrants. Are you seeing the breadth of new competitors widening? Or is it really the strongest companies, the largest companies that are doubling down in this business and winning? How are you seeing the competitive intensity change as we start to move into ADAS being adopted?

Right. Yes, indeed. I mean the entry barrier to this market is very high. you, though are fully right. I mean everybody sees the attractiveness of the market. And of course, also some people try to enter. But we've also seen through the COVID crisis that some competitors, yes, there's a bit of a shakeout. And we, at least, hear rumors that some competitors might consider to scale down their efforts on radar. And the reason why the entry barrier is very high is that, I mean, you need automotive quality. You need functional safety. You need security, and all of that is deeply at the DNA of NXP. And then specific to radar, as mentioned, I mean, doing 77 gigahertz in RFCMOS is super hard, and only very few can do it. And by the way, also our biggest radar competitor seems, from our perspective, to lag behind in this space. They are not yet being ready with RFCMOS, for example. And then on top, yes, you need all the processing, know-how, and yes, all the algorithmic and the system know-how, and you need deep pockets and a very long breath. So it's quite a number of factors coming together, and there's really -- not many or probably not any competitor I can think of who has all the ingredients that NXP brings to the table to the same extent. But of course, I mean, we are paranoid about our competitors, and we take them very seriously. And they are all very respectable great companies, no doubt, but we feel very confident on our position.

Final question on the competitive landscape. There's the processing side and then the transceiver side, which just tends to be a little bit more analog mixed signal-ish. How do you view the competitive landscape from those 2 sides? Do you have to be an expert on both sides of the equation? Or does this fold in with the RFCMOS and then the ability to integrate? If you don't have RFCMOS, you don't integrate, and it doesn't need -- the company that's competing with you doesn't need to have both sides of the equation. Does that limit, in your view, the competitive landscape and the number of players in it, more so going forward than it has in the past, before the integration capability was there?

Yes, indeed, absolutely. That is spot-on because, indeed, in the past, I mean, you had players that were maybe only focused on the RF side and others that were only focused or primarily on the processor side and so on, which was maybe okay at the time with the less integrated system and the market just evolving. But indeed, I mean, the key advantage and the key value proposition of NXP is that we have the complete system expertise. We have the full system offering, and we are a one-stop shop. And indeed, I mean, going forward, as things integrate more, you need to have both sides of the equation, at least the transceiver, the RF side and the processing side, and you need to have the capability to bring that together. Plus, there's also a lot of really system linkage between those in terms of algorithms, et cetera. And yes, that means also for our customers is a big R&D leverage and a big advantage to scale with our platform, then it really limits some of our competitors who might be a bit narrow in that sense.

Great. And I think we're approaching the top of the hour, so why don't I ask one final question, and this will go back to -- I think it's Slide #4, where you talk about the 20% CAGR for the SAM of this market but the 25% to 30% growth estimate for NXP as you take share in the market. That share gain, it's driven by a lot of things you've already talked about, the integration of the 77 gigahertz, et cetera, et cetera. But if we look on Slide 6, you have the number of radar units or nodes per vehicle rising. You also mentioned that as you go from corner to long range to imaging, the content goes up. So if you're talking about the incremental share gain or the incremental CAGR that you have, how would you summarize that between those metrics? Is it more the units or nodes per car? Is it the content per unit going up? Or is it just you guys taking a greater share of the market itself, the number of units?

Right. Yes, it's a combination, to be honest. I mean, first of all, we are leading in the 77-gigahertz market, and we have less exposure to the 24-gigahertz market than some of our competitors. So that's already the better segment to be strong in, certainly. And then it's really the combination of playing across the whole system and having the transceivers as well as the processors as well as the attach, and, yes, having more radar and more nodes per vehicle and then incrementally also the more content per individual sensor. So let's say in the nearer term, it is more the higher penetration of radar and the more nodes, so the corner radar, and let's say this whole imaging radar and more content is something that is slowly starting and will further accelerate also beyond the time frame here we're discussing.

Great. Well, that takes us exactly to the top of the hour. So Torsten, thank you so much for your time on this, and thank you, everybody that dialed in and those that asked questions. I believe, Jeff, the replay of this event will be available on the NXP website shortly, and so you should be able to see the slides and see the replay on that. And if you have any follow-up questions on my side or I assume on Jeff's side, please feel free to reach out to us individually. So thanks, everybody, and have a great rest of your day.

Great. Thank you, everyone. Have a great day.

Thank you very much, Ross, and thank you, everyone. Yes. Stay safe. Thank you. Bye-bye.