All right. Good afternoon, everybody. Thank you so much for joining us for this session with Mobileye, as part of Deutsche Bank's Global Automotive Conference. My name is Emmanuel Rosner, and I'm the lead US autos and auto technology analyst here at Deutsche Bank. I'm incredibly excited to be joined here by Dan Galves, who's the Chief Communications Officer of Mobileye. As you probably all know, Mobileye is a technology company specializing in the development of hardware and software for ADAS and autonomous vehicle driving capability. The company's product range from base ADAS on chips, for which it has 70%+ market share, to Level 2+ SuperVision, Level 3 Chauffeur, and ultimately, Drive, which is fully eyes and hands off. So very happy to have you here today to discuss, you know, some of the progress you've been making.
Thanks for being here.
Thanks, Emmanuel.
So maybe just to kick things off, can you give us an update on Mobileye's progress so far this year? First, the volume outlook. How has demand been tracking on the base ADAS side in particular? Any signs of slowing down, and maybe an update on the chip destocking that you had seen?
Yeah. So, yeah, thanks for having me here. I really appreciate it. Yeah, we started the year with, you know, a pretty disappointing announcement that we had, you know, had a kind of a buildup of inventory. And, you know, as of January 4th, we laid out a pretty kind of clear and detailed cadence of kind of how volumes would progress over the course of the year, and everything looked really on track at this point. I mean, we really needed the market to cooperate, and that's, you know, the market demand has actually been very stable, not just for our products, but overall global production. We don't have any kind of overexposure to EV or ICE or anything like that, so, you know, kind of the mix hasn't affected us.
Yeah, as we said on the Q1 call, we dealt with, you know, 70%-75% of the inventory in Q1. We were projecting that by the end of Q2, you know, inventory levels would be back to normal, and that's still what we believe. And we've seen kind of good kind of order indications for the back half, so we feel like very much on track, to what we laid out at the beginning of the year.
And if we take a step back on that whole inventory situation, is there now enough open communications on the channel to prevent the destock or stocking from happening again? I guess, what measures have you taken internally and externally to get a good feedback loop on what's going on?
Yeah. So I think as a Tier 2 , so we're always selling through Tier 1 auto suppliers, and I think probably over the years, you know, they've done a very good job of ordering to demand. And, you know, we were essentially maybe over-relying a bit on their demand outlook or their kind of order outlook. Things changed a bit during COVID, where, you know, there was a more of a desire to stockpile some chips, and we were kind of caught by surprise. We put in a lot of new processes to... You know, we're very fortunate that we can look at global production data. We know what cars have our chip, what cars don't have our chip, so we have a very good sense of what the end demand is.
As long as we're matching that to our shipments over a particular period, we should have a good sense of kind of whether we're in balance or not. All of these data points are pointing towards significant inventory reductions over Q1 and Q2. We're also, you know, yeah, much more kind of in communication with Tier 1 suppliers. Some of them are providing us detailed inventory reports. So yeah, we feel like we've got a handle on this. We definitely don't want it to happen again.
And then turning to SuperVision then. So for the year, you've been pointing to, or guiding to 175-195 thousand units of SuperVision shipments. How is this tracking midway through the year? Do you still feel comfortable, you know, with this range? And, where would you see potential risks or potential, offsets or opportunities?
Yeah. So I mean, we're – we'll report. Kind of, we'll update our guidance in a month or, you know, approximately a month, a little bit more than that, so I don't wanna, you know, get ahead of myself too much. I think that, you know, this business, we're only on five different vehicle models right now, so it can be pretty volatile. And we've seen some of that in the first half of this year, where, you know, Zeekr 001, which is the biggest volume product, was relatively low volume the first couple of months of the year, but then they refreshed the car, and the, you know, order flow and volumes went up significantly, you know, to the tune of, like, going from 6,000 a month to, like, 12,000 a month.
So I would say, from that perspective, you know, maybe Zeekr 001 is, is maybe a little bit better than we expect, but there's still six months left in the year. I think on, kind of on the balancing side, Polestar 4 is a vehicle that was, intended to drive a decent amount of volume in the second half. You know, their launch looks to be on track, so we're still encouraged by that vehicle. But, you know, it's produced in China and intended for sale in Europe and the U.S. Like, the U.S. might be a little tough with, the 100% tariff. So not a, you know, a super high amount of volume, but, you know, that could be kind of a balance, to, to other, to other good news.
I think it also sort of it speaks to the fact that this is an extremely important, kind of high-visibility business for us, SuperVision in general. But right now, when it's on 4-6 different models, you know, and a lot of that is in China, there's gonna be a lot of volatility. I think in general, you know, investors are, you know, paying very, very close attention to even weekly sales models, sales volumes. You know, the numbers are really not even that material. You know, 10-20,000 units is, you know, $12 million-$25 million of revenue on an annual basis, which isn't really material to us, but it definitely gets a lot of focus.
So I think, you know, that business will remain a bit volatile just because of the small number of-
This year, you also have potentially FAW launching in the fourth quarter, which was not in the range that you gave?
Yeah, that's correct. Our next launch customer is FAW, and that's the kind of the driver of, you know, should be, you know, maybe some volume this year. We'll see what happens, but more volume next year.
Okay. And for the benefit of everyone who is here in person, those Zeekr vehicles are actually downstairs, available for at least checking them out. I'm not sure about test drive, but the Zeekr 001 and 009 are all downstairs. So your main customers right now are basically, you know, Geely out of China. Can you talk about the potential impact of these recent tariffs imposed, first by the U.S. administration on China EVs, and potential ones in Europe, sort of like in any day? Is this a topic that has come up, you know, with Mobileye, because it could eventually impact growth of supervision outside of China? I'm thinking also for models like Polestar 4. How should we think about that?
Yeah, I mean, I think it's back to supervision is kind of a long-term growth driver for us. I think that Zeekr, you know, having the opportunity to launch that system with Zeekr was huge for us, right? Because we fully believe that if we didn't have that proof point of the system on the road, that we would not have won the VW business that we announced earlier this year. So China's an amazing kind of proving ground for this technology, 'cause consumers are gravitating towards these types of systems, and OEMs are moving very quickly to innovate and bring systems onto the road. So from that perspective, it's been a huge positive to our business. But also it, like I said before, it leads to volatility of kind of geopolitical issues back and forth.
Like, we don't like the kind of tit for tat, you know, back and forth between China and the U.S. and Europe. I think tariffs in general, like, it's kind of a good and bad thing for us. Like, the good side is, you know, there has been a lot of market share shifts, you know, inside China from kind of foreign automakers to domestic automakers. There's a lot of potential for China to export vehicles into Europe and, you know, eventually the U.S. And I think in general, like, our main customers are non-Chinese automakers. So from a perspective of, you know, market share solidification for our own customers, maybe it's a little bit helpful.
But we also like this dynamic of China exporting systems outside of their home country, because we do think it leads to more sense of urgency from foreign automakers if they're gonna be competing with Chinese automakers that have, you know, kind of very fast innovation times and a desire to kind of push the envelope of intelligent driving. So kind of a, you know, positive and negative type of thing.
And then more generally, how exposed is Mobileye to the push out of new EV model launches, or the recent demand softness? I think that is always a bit of concern among investors that, yeah, you're basically powertrain agnostic, but in the end, a lot of these launches were gonna be on EV. Is that true?
So I think that the base business, the EyeQ, kind of single chip, single camera business, is still 90% of our revenue, and if anything, we're underexposed to EVs there, 'cause we don't have business with Tesla. SuperVision, the first set of customers was all within this Geely group, and it was all, you know, essentially pure EV companies and EVs, so definitely a 100% EV. I think our future launches are a mix. You know, as we talked about with the VW deal, primarily Porsche and Audi, it's about 50/50 EV versus ICE. I think the bigger picture is like we've...
You know, kind of the global OEM have gone through a significant period of shock is probably the wrong word, but I think we did see, over the last four, five, six months, a significant kind of replanning of the portfolio of the global OEMs long term. And in some ways, that probably delayed a bit kind of their decision-making on SuperVision. I'm going back a few months now, so there's no change that I'm trying to signal here. We're still on track for, you know, a number of design wins in the second half of this year.
But I think that the positive that came out of that was a more kind of balanced portfolio approach, where I think that there was never any connection between EVs and supervision from a technology perspective, but, you know, when an OEM was thinking about, like, "Where do I want my leading-edge technology to go? I want it to go on a flagship vehicle." And, you know, as of a year ago, those flagship vehicles were basically gonna be EVs. I think that that's changed now, so it actually creates a more balanced business in the future for us. And, you know, maybe we even got a little bit lucky that we hadn't booked, you know, a number of deals before this change happened. So I think that, you know, the OEMs are looking at this-...
kind of been a, you know, powertrain in a more, balanced way going forward. And, you know, maybe there was some, you know, "Hey, I wanna focus my technology advancement towards EVs because this is kind of where the market's going, this is what people are gonna wanna buy." But now, they understand that this transition is gonna take longer. Again, this is my perspective, I don't wanna speak for the OEMs, but they understand that this transition's gonna take longer and, you know, you need to compete on ICE vehicles and hybrids as well. And the way you compete is to have good technology in the car. So, you know, we're seeing a lot of kind of positive traction in terms of, you know, the supervision product set, you know, new design wins, and they're definitely gonna be balanced between EV and ICE.
Let's focus a little bit on your pipeline of future business. How are you thinking about your customer pipeline for supervision? I think, over time, you've, you know, you reported some design wins. You have previously stated you had 14 OEMs in advanced discussion or agreements in place. Any details you can provide on the current appetite from OEM for these advanced autonomous systems? And it sounds like, based on your previous answer, that you still expect additional supervision wins in the second half this year. Is that correct?
We do. We do. Yeah, so no kind of change in kind of the progress. Many of these engagements have kind of evolved from technical evaluations to more of a formal RFQ process, which, you know, we think you know tends to solidify the timeline quite a bit, so that's been a positive.
I think the other thing that we've noticed over the last three or four months is that there's been more of a kind of crystallization of kind of how global OEMs want to deploy this intelligent driving technology across their portfolio, where I think, you know, 6-7 months ago, it was kind of like, okay, we have SuperVision on a path to Chauffeur for premium brands and, you know, fairly high priced vehicles, not, you know, not $100,000 above, but, you know, $50,000-$60,000 and above, and then we have base ADAS, right?
What we've seen over the last four or five months is that more OEMs are looking for a leap in functionality for mainstream brands, and there's a way to do this, and kind of specifically the deployment of hands-free systems on highways, you know, with automatic lane changes, but not really targeting off highway, not necessarily needing to be scalable to chauffeur. But, you know, the idea is you kind of keep your hand in the game with leading-edge technology at the high end for the premium segments, but you wanna be able to differentiate in the mid-range as well. And so we've seen a lot more appetite for highway hands-free, and we can support this with one EyeQ6 High.
We can take advantage of a lot of the content that's already on the car, such as the front camera and the four kind of near vision, wide angle parking cameras around the side of the car, and support, you know, a pretty big leap in functionality. Think of it as kind of an, maybe an improvement to current gen Level 2+ systems for a relatively minimal incremental cost to the OEMs, because it's only one EyeQ 6, and because you're actually taking advantage of quite a few of the sensors that are already on the car. So this has been, I think, probably the most positive development that we've seen in terms of, like, a solidification of, you know, we don't wanna just see kind of innovation at the high end, but we want kind of a mid-range in the meantime.
And then you kind of see how SuperVision and Chauffeur go, and if it's successful and if consumers love it, then you can kind of proliferate that down the portfolio, but at least, at least you have something more high-end for the mid-range and for the time being. So that, that's been a kind of a pretty big bit of an evolution of the business development that we've seen in the last few months.
In the U.S., there's no great data on consumer interest in supervision-like features. Investors worry that end demand may or may not be there. Take rate of S-Class FSD has been reportedly low, although we haven't seen any reliable data of conversion rates since they cut the subscription price. Based on Mobileye conversation with the OEMs, is there sufficient demand for this kind of technology? Or is there a risk that OEMs skip Level 2+ altogether and just wait until a robust eyes-off solution is available?
Yeah, I mean, it's a good question. I think number one, there's no perfect information out there. That is true. You know, what we know is that, you know, we saw a survey of Zeekr users, about 600 Zeekr users, and 85% said that they were either very satisfied or satisfied with the system, and 60% said they'd be willing to pay $3,000 to kind of continue to have the functionality. So we thought that those were, you know, fairly solid, solid statistics. Now, the reality is nobody's really charging for it in China because it's become part of the price war, and it's become kind of part of the cost of selling a car in that segment. So I mean, our information is basically that if you-...
are trying to sell cars in China in the $45,000 and above price segment. It's very difficult to sell cars if you don't have some sort of hands-free, or I shouldn't keep on saying hands-free. It's really called Navigate on Pilot in China, so point-to-point, you know, plug in their address, the car drives for you. You have to keep your eyes on the road, but the car drives for you. So there's five or six of these types of systems on the road in China, and it's become very difficult to sell cars in the $45,000 and above segment without having that.
So I think from that perspective, it's actually a big positive to the global OEMs because, you know, obviously, you can't extrapolate completely from China to other markets, but it's a good sign that if you put these systems on the road, they will differentiate your cars. I think, you know, we'll see what the actual, you know—maybe we'll never see what the actual data is on penetration rates for Tesla. Still pretty expensive at $8,000, $100 a month.
You know, I think the expectations were probably pretty high of the consumers that did download it, and like we said, we don't really have definitive data, so we haven't seen any impact at all on our customers in terms of them, you know, that putting any sort of caution in their viewpoint.
And if I'm looking at, you know, 2025, 2026 and beyond, what's a good framework to think about these sort of like midterm supervision and volume growth? At CES, you provided for a first time, a very good customer launch table, which outlines vehicle model, the model numbers and the timeline and a timeframe for launch, which technology goes into it. But can you help us tie that table back to the kind of volume expectation we should consider for supervision? I feel like in speaking to investors, sometimes 2025 is sort of like the most worry, because it feels like, unless I'm wrong, the only one you're really launching between now and then is maybe FAW-
Mm-hmm
... and I'm not sure to what extent is large volume versus not.
Yeah. Yeah, I can't really get into volumes too much in the next couple of years. I think again, it's kind of back to like when you only have five, six, seven, eight different cars, like, it's gonna be tough to predict, you know, when you're still 6-12 months away. I think that, you know, what we kind of look at most importantly, internally is that, you know, you have an industry where right now there's 90 million cars produced globally, and probably about 60 million of those have some sort of driving assist system. We see that 60 growing to around 80 over the next five or six years, as markets like India and Brazil and kind of lower end segments in China and South Asia, kind of catch up on ADAS adoption rates.
There's actually, you know, significant, kind of important growth drivers that, you know, will likely, you know, drive that volume higher. The big question is gonna be: What's that pool gonna look like in terms of segments? And I think you know, there's lessons in the past, right? I mean, I think AEB in 2012 was, you know, a high-priced option on cars. And, you know, I think we did 700,000 units in 2012. But you could see the writing on the wall that, you know, regulators were getting involved and, you know, cust-- Well, I mean, maybe you didn't see the writing on the wall. I think at that point, you didn't know kind of where the business was going.
You know, you fast forward 10 years, and we were above 30 million units from 700,000 in 2012. So I think that's kind of where we are in this kind of evolution of kind of more sophisticated driving systems. It, it's hard to tell kind of how quickly the adoption curve is gonna run.
Let's speak a little bit about the technology side. There's been quite a bit of focus on robotaxi and autonomous vehicle with the Tesla- upcoming reveal of their vehicle in August. You shared some pretty good information on Mobileye's approach to full autonomy, as it-
Mm-hmm
... is different from, you know, Tesla's end-to-end approach. And we hosted ourselves a fireside chat with Amnon Shashua just a couple of weeks or so ago, to go deep dive on this. If you haven't seen it, and you're up for an intense technology discussion, there's a replay available. But let's say we're not up for it right now. So what are the practical pros and cons?
I'm up for it.
It's only me. What are the practical pros and cons of an end-to-end AI approach to solving autonomy, such as Tesla, if that's indeed what they're doing, versus why do you think that yours is, you know, essentially more, more likely to succeed?
Yeah. So I think you'll notice we were very careful kind of not to say, like, this is kind of a right or wrong thing. We have to kind of look at kind of the actual end market, and I think also, you know, in kind of the intricacies of kind of the automotive end market, as opposed to maybe a large language model, and also kind of what's happening in AI in general. I think, first of all, I would say that Mobileye has a history of kind of rapidly integrating new evolutions in AI, which, you know, this latest one is not the first. It won't be the last.
You know, convolutional neural networks were kind of a 2012 academic research, and we had CNN in our launch with Tesla in 2014. Right, so rapid integration there, deep neural networks started to kind of become a thing in 2015. This was the first time that NVIDIA started saying that kind of data plus compute equals self-driving. And we had deep neural network technology in our EyeQ4, which launched in 2018. And we have, particularly what's called BEVFormer. So it's a transformer-based architecture that creates a bird's-eye view of everything around the vehicle through kind of, you know, consolidation of all the 11 camera data. This is something that's kind of very prevalent in China right now.
This is a transformer-based architecture, and this is in EyeQ6 High, which launches in the first half of next year. So we've had a very good history of kind of rapidly integrating new technologies, but we don't believe the right thing to do is to kinda completely pivot to one thing or the other. And I think we're in good company there because even, you know, ChatGPT, you know, even OpenAI and Google, in kind of their latest versions of their large language models, have moved away from a monolithic end-to-end AI system and kind of inserted more, you know, other techniques to create efficiencies in the system, create scalability in the system, and to look at specific kind of use cases of the system and say: Okay, I can get better accuracy through this technique rather than end-to-end. Okay?
This is called Compound AI systems. It's kind of the latest in kind of the academic research around these areas. You know, this is even more important than automotive because accuracy is key, right? Like, I mean, ChatGPT can't kill you. A car can. So getting to this kind of 99.999% accuracy is really. That's kind of the main goal, particularly for eyes-off systems. And we think the best approach is through Compound AI to get to those accuracy levels. And then finally, I'd say, like, you know, OEMs will not go for a Monolithic AI system for two reasons. One is that when the system does fail, and it will, you're gonna have to be able to explain why it did, to find the root cause and then identify a fix.
If you just have a kind of a black box, monolithic AI, you don't really know what's going on in the system, and so you can't kind of dig in and find out, you know, what was the root cause and the fix. You just have to try to find more data, retrain the system, and hope it doesn't create other problems. So that explainability is really important for OEMs, for regulators. And then the controllability. I mean, the one thing that we know that's a complete deal breaker for OEMs is for them not to be able to control the driving experience and fine-tune kind of how this particular, like, you know, I'm approaching a stop sign. Do I want to gradually slow down? Do I want it to be a little bit more harsh?
That braking profile, it can, there's infinite different ways you can approach it, and they wanna be able to control it. They wanna be able to set it the way they think that their consumers will like it. And then if they're wrong and they get complaints, they wanna be able to tweak it in real time, and you can't do that with a monolithic AI system either. So I think there's, like, a variety of reasons why we feel very confident in the approach we're taking. But I guess the last thing I would say is, you know, as always, things are more complicated. There's been this kind of dichotomy of, like, AI versus rules-based, and we've kind of been put in the rules-based camp. It's much more complicated than that. We have plenty of AI in our system, but it's not 100%.
So that, that's kind of our viewpoint.
I wanted to ask you also a little bit about the competitive landscape for your solutions. You have often framed an OEM's decision about autonomy to as having to either write the algorithm themselves or they can source a solution from Mobileye. Is this still the case, or are some other players now offering competitive, personalizable driving algorithms to automakers? Maybe some of the chip makers like Qualcomm or others, where automakers can just put a few inputs and then, you know, still have a solution. Or do you still feel like the only turnkey solution is Mobileye?
I think, it's kind of different inside China and outside of China. Outside of China, we still feel the same way. It's, we haven't seen any kind of entrant or, you know, existing competitor that can provide, you know, an actual solution to the automaker, including software and hardware. NVIDIA and Qualcomm have some position in the business, but it's really as a processor. It's like a chipset provider, where somebody else has to provide the software, and that, kind of the only option there we see is the automakers.
You know, there's been a lot of development in that area over the last five or six years, and it's, you know, we feel like the pendulum has much more swung towards a desire to have software delivered, as long as the OEM has the ability to control how the consumer customer-facing aspects, like I was talking about before. Inside China, there's been more success with in-house systems, and most of them use an NVIDIA processor. So Li Auto, XPeng, Huawei, Xiaomi, NIO. So this has been fairly successful, like with good, kind of respectable performance on the road. Our advantage there is these systems are using, you know, at least one, you know, sometimes two LiDARs. They're using compute power that's, you know, 10x-20x higher than us.
So our benchmarking would say that our system is 50%-70% of the cost of these kind of in-house systems, and that's not including the in-house R&D, which can be at least 1,000-3,000 people. So I think over time, it will be difficult to kind of maintain that type of cost structure and kind of investment structure. We don't see this, we don't see this kind of approach being adopted by the bigger, kind of more established domestic China OEMs. We see them wanting a more of a supplied solution, but it's not gonna necessarily gonna be only us, right? I think there's also a big push to have kind of local content in China. So our, our view is ultimately there will be three or four, you know, suppliers of hands-free system solutions in China, and we'll be one of those.
We've never said that we would have, you know, 50%, 60% market share in China. So we feel like we've got a lot of advantages in terms of cost, ability to support performance in U.S. and Europe, path to eyes off, you know, lots of advantages. But I think that we'll eventually have three or four suppliers, and maybe Horizon's one of them, maybe DJI is one of them, but that's kind of how we see things going. But it's a pretty unpredictable market right now, so we'll have to see how things play out over the next few years.
Maybe just finally, when I look at higher levels of autonomy, you have the Chauffeur products and then Drive eventually. What is the status of a discussion with and interest from customers right now?
Yeah. So I think, I think Chauffeur is kind of the North Star of a lot of OEMs. Like, the kind of development of SuperVision. You know, they see benefits and, you know, potential, you know, profit pools from that product, but they really see a clear value proposition with a system where, you know, maybe 90% of the time you can do other things while you're driving. So Chauffeur is a big driver of development of SuperVision and Chauffeur in general. And I think we feel pretty confident that we're the only supplier that has, like, a true eyes-off development program, series production program, and that's the one with Audi. So, you know, we're encouraged by that business.
Drive is more challenging from kind of regulatory, and there's lots of other aspects that you have to deal with to have no driver at all. But the potential payout is huge because you're removing a human driver. So, you know, we're selling these kind of Drive systems for $50,000-$60,000 apiece. So if things work out, you know, it's a big opportunity. I think what we're trying to do there is be agnostic to use case. Like, we've never really thought the right approach is to, you know, develop in a few different cities and then basically do your own robotaxi service and kind of assume that customers will come.
We wanna be more agnostic to use case, and the use case that we're seeing probably the most interest in right now is the conversion of large fixed route buses with drivers to autonomous, on-demand, smaller shuttles, where people could sign up for rides, and basically, the car would pick people up and bring them into the city center. So we're seeing a lot of demand for that from public transit operators. So that may... But I think, you know, that could change, right? What's the best use case? What region is gonna have the most kind of open, openness to this technology? Like, we wanna be agnostic, and that's why our approach, that's kind of based on the consumer level driving system, we think is really important, or we think is really the right approach.
Great. Looks like we're exactly out of time. So Dan, thank you so much for your time and insights.
Thanks, Emmanuel. Yeah, thanks a lot.