Hi everyone, my name is Colin Rusch. I am the lead for Oppenheimer's Sustainable Growth and Resource Optimization team. We've done an awful lot of work in and around physical and industrial AI, and we're thrilled to have Dan Galves, now COO of Mobileye Global Inc., here to discuss the platform. Dan, you've had a couple of stints with Mobileye and have seen significant evolution of the platform. Can you just set a baseline on where the company is from a technology perspective as it enables the autonomous future?
Yeah, so I mean, I think that, you know, where we are from a technology standpoint, kind of where the customer base is, is like a little similar to kind of where we are, where we were back in 2010, 2011, 2012, where we were trying to advocate, you know, use of camera for ADAS technology, you know, for improving safety levels, for, you know, saving lives, providing warnings. There was no, you know, there was no really evidence of consumer acceptance of this, and kind of the OEMs were understanding that, you know, they were going to need to charge, you know, extra for the car for this technology. It created a, you know, a challenging, you know, business development perspective.
Once systems started to launch with BMW and Volvo in 2012, and you started to kind of feel some of the consumer acceptance of that technology, what, you know, and some early success in monetizing it, you know, really the dam broke, and then ADAS adoption started kind of coming in with regulation, and kind of adoption rates went from nothing, you know, up to now they're like 75%. We feel like we're in kind of a similar situation, you know, where the OEMs need to make decisions, you know, based on, you know, somewhat like not a lot of information, not kind of understanding where the technological maturity is going. Where are we right now? Like we're fully focused on the execution of the Volkswagen programs and importantly, like using technological progress there to further demonstrate our maturity to other OEMs, right?
Testing and development work for SuperVision, Chauffeur, Drive, it's now happening with production intent vehicles utilizing production-level hardware, much of the next generation EyeQ6 high-based software stack. The design freeze for the SuperVision program occurs in early 2026. That's to provide time for kind of full vehicle validation and homologation work throughout 2026. In front of the launch, we've had, you know, multiple executive demos in Europe and China, including recently that, you know, continue to illustrate superior performance against other, you know, they continue to kind of evaluate us against other kind of benchmarks, competing systems. Chauffeur's on a similar track, just a few months behind. Finally, on Drive, like we continue to hit all of our targeted KPIs, including our mean time between failure targets. Software freeze, you know, in order to support the European certification process, happens at the end of 2025.
We feel like we're going to be at the point where we can start to commercialize during the first half of 2026 in the U.S. because of regulations that have to wait till the end of 2026 in Europe. Yeah, we're feeling very good about our ability to put up like multiple more proof points in the coming months and across next year.
Fantastic. I guess, as we see investors trying to keep pace with the rapid evolution of AI technology, can you talk about Mobileye's compound AI approach and how it's leveraging simulation test vehicles, inference, and pushing reliable functionality forward? I think one of the trickier things, you talked about some of the KPUs and the mean time between failure, but I think there's a lot going on from a technology perspective that's happening that doesn't necessarily end up in the public realm and can be challenging to digest for investors that are really dealing with a high volume of information and trying to kind of simplify things as best they can.
Yeah, no, I mean, I think in some ways, like, you know, Tesla has an advantage here because they're, you know, connected with all their vehicles, because they're producing the vehicles themselves, like they don't have to wait for like a production breakpoint to deploy new technology and show significant improvement, you know, and they did that a year ago, like they showed significant improvement, you know, through the use of transformer-based networks and end-to-end AI. It seems to have plateaued for a while, but they did show that improvement, and that's an important proof point, not just to consumers, but to other OEMs as well. Our kind of production proof point in the Zeekr vehicles was based on EyeQ5 kind of first generation technology. That chip was designed in 2016, 2017, 2018.
The software was designed 2020, 2021 before you really had the benefit of a lot of these kind of new AI architectures. Just to address the value of new AI, we've seen tremendous progress in our software stack, driven by the use of end-to-end transformer-based networks, both perception and driving policy. These are replacing a lot of the kind of convolutional neural network-based software that was in the first generation of SuperVision. It's all particularly helpful with comfort. We don't necessarily see sort of the pure safety-related improvements just based on technology. That's what makes us confident that more of an engineered structure that we've been kind of planning forever, or that we've been kind of utilizing forever and continue to see major benefits to, is important to reaching the necessary precision levels, which are far away from whatever's out on the road today, right?
Like maybe you can do 500,000 mi, but what about 500,000 mi or 100,000 mi or 1 million mi? We think that that engineered structure has a lot to create benefits through redundancy. If you're dealing with edge cases or even like challenging but common occurrences, the goal is not to treat them as edge cases, but to figure out ways to bring them into the normal distribution of scenarios, right? To make them normal, how do you do that? You create a structure where you have multiple softwares, multiple sensors evaluating that scenario in parallel. You have logic that's in place to decide which system to rely on in that specific case. It's pretty complicated to do. We call it primary guardian fallback, and it's a big part of our compound AI system.
On the decision-making side, you can use pure AI, end-to-end AI for decision-making software to make suggestions about what that next action is. Our belief is that next action proposed will be right in 99% of the cases, but what about the 1%? If you have a kind of logic-based, mathematical-based guardrail on top that's verifying that this decision is not going to cause an accident, and if it is, then you go to the next decision, we see as a way to boost precision significantly. We're using simulation in a lot of cases. We don't use photorealistic simulators. We base it on REM data. We can create a realistic environment based on our crowdsourced mapping data and then have vehicles interact within that environment and go through 10 million, 20 million scenarios overnight. We're making heavy use of simulation as well.
There's a lot that I want to unpack there. Dan, you know, I know we had some questions that we want to go, but I would love to go back to the customers here, right? I mean, part of this is customer education. We've seen this cycle of customers getting involved with autonomous, wanting to do their own thing, right? Then realizing that it was too big a problem, that they could, they didn't have the capability in-house to do this. They're coming back to technology providers. I'm curious about the state of those customers in terms of their education level and understanding the complexity and their ability to deal with that complexity in an environment where you've got increased regionalization, you've got the drive train transformation that's happening at kind of faster or slower rates depending on which geography you're in.
The real question is, what's the value to the consumer and who actually is the consumer? I mean, it's a lot of crosswinds in the space. I guess, can you give us a sense of how those customers are wrapping their heads around the depth of the challenge and how they've come back to work with you? We've had this theory around a tier 0.5 in many cases, and it seems like you have some of those close working relationships with folks like Volkswagen, where you're doing iterations with them in a little bit closer way. Just curious if we start moving in towards that tier 0.5 as we get into just some really vicious technology cycles here soon.
Yeah, so I mean, I think there's a lot in that question. I think there's different themes, right? There's the powertrain theme, but there's also the architecture consolidation theme, right? For our software-defined vehicles that have less ECUs and more consolidated ECUs than previous platforms, were they going to be all EV? Maybe they were at first, now they're not, right? That delays everything, right? I think that we have seen these sort of targeted platforms that are supposed to be packed with technology. Maybe a couple of years ago they were planned for mid 2026, now they're planned for 2028. It seems to be stable, right? That creates some level of pressure to make decisions now to have advanced technology in these vehicles that are launching late 2027 into 2028. I think that's one thing that's happening. In kind of recent updates I've gotten, there is more appetite.
If you go back two, three years, the target was basically SuperVision, right? Tesla FSD is out there, it's eyes on, it's hands-free, getting a lot of press. No one knew that the penetration rate was relatively low at that point, kind of now they do, but this was the target, right? Once we get that in place, we can start to add some sensors and maybe get to eyes off. Maybe something's going to happen in Mobileye Robotaxi, but let's start with this. I think over the last two years, what we've seen is more segmentation where, okay, maybe a kind of an eyes on hands-free for premium brands, and maybe that's scalable to eyes off at some point, but what are we going to do about the mass market vehicles?
We really can't put $1,700, $1,800 of cost into a $40,000 car or a $50,000 car, we don't have the confidence of monetization to do that. Can we get to a very cost-efficient highway hands-free driving that also helps us to meet future safety standards? Surround ADAS has become kind of a lot of traction to serve that area. Robotaxi, all of a sudden there's like consumer proof points that people are preferring driverless vehicles rather than Ubers. That's created sort of a rekindling of what are we going to do as an OEM. We're seeing more clear segmentation of what OEMs want to do and being able to be a one-stop shop like we are has advantages there. We feel like that's kind of a newer trend that makes sense for us.
On the negative side, I think new AI has created a lot of noise and a lot of confusion, especially at the upper executive levels, and it's also allowed for more startups to pitch themselves, right? As, hey, now that you've got new AI, this has become relatively easy and we can just use the existing hardware in the car and all we need is data and lots of compute and it'll teach the car how to drive. I think that affects, hearing that from Jensen or hearing that from some startup, it affects the executive mindset. We have to go in and explain why an engineered structure leads to better precision levels and more traceability and more predictability. We feel like we're in good shape, but I think it's created some noise.
Okay, at some point, decision makers need to harmonize product cycles and technology cycles and understand what those are. If you had to characterize this latest education cycle, where are we in that education cycle? We go back to 2016, we had this kind of autonomy was the future suddenly, right? That drove a lot of the interest in the EV powertrain because you needed to have a big battery to run the computer, right? That was part of what happened with EVs. Fast forward here, if we've got this AI disruption where folks are talking about these massive computers and not necessarily a more focused, task-driven AI learning cycle, which actually kind of accumulates. We're very much in the camp of start small and grow rather than boil the ocean with these AI compute.
Where are we at in terms of some of the decision makers understanding the scope of the problem and how to attack the problem and starting to harmonize technology investments with product cycles?
Yeah, I think it's a great question. The product cycles do feel more harmonized than they did a couple of years ago. Trying to kind of push a string to get a decision made when the vehicles that were targeted weren't clear, it wasn't going to happen. Now that the product cycles are more clear and they're a little bit more diverse by powertrain, we think it's good for us, especially for this current cycle where you're trying to get product on the road for 2028. There's not a lot of time, right? As you evaluate other solutions, it makes having someone who can actually demonstrate performance in production-level vehicles gives you a big advantage there.
That's why the Volkswagen programs are so important for us, because not only are we creating some level of competitive pressure because these systems are going to start to be in Volkswagen vehicles late next year, we're also creating the situation where we can show production-level hardware and production-level software in a vehicle that can drive anywhere. Even things like instead of having a defined route for your demo, just having the customer get in and telling them, where do you want to go, it means a lot. It doesn't mean that you're not going to continue to see investment in in-house development, maybe for that next cycle, because I think in some ways you'd probably be stupid to just ignore that there have been technology changes that may create more opportunity for you to do in-house or create more opportunity for other companies that are just starting now.
The idea of somebody, like a wave, actually getting a system into production in the next couple of years, it's not credible with the OEMs.
Right. That's super helpful. The other thing that's happening with this kind of increased interest in the space is just the capital that's gone into the ecosystem, right? One of the spaces has been Perception got, you know, a lot of capital early. There's been some consolidation. You guys have made some real strides in your own Perception suite. Can you talk a little bit about the ecosystem of, you know, Perception solutions that are out there that are really going to help support eyes off, mind off, and what Mobileye's current thought is on what's needed and how that part of the ecosystem evolves within the system level kind of functionality that you guys are talking about?
Yeah, so we're still big believers in, you know, perception systems that have different strengths and weaknesses create like a more robust structure overall, right? Because you can understand where the strengths and weaknesses are. If you have, you know, two independent perception systems, you know, each analyzing the environment, the chances of both of them failing at the same time go way down. That's like the simple explanation for why we see, you know, redundancy in sensors as really important. We have, you know, I can remember Amnon telling me like almost 10 years ago, like that there's way more runway in radar than people expect and that, you know, you could see imaging radar type of sensors emerge as, you know, a good competitor to LIDAR because they're less expensive. Even more importantly, because they're more independent from cameras in terms of, you know, what they do well.
You know, so cameras have really good resolution, but the distance is not that great. Imaging radar has, you know, much better distance perception. If you can get the resolution to as good as LIDAR, then you kind of eliminate one of the, you know, the disadvantages to LIDAR. We feel like we're there on that. Things like bad weather, I mean, you have to think about this. If like you're, you know, an eyes-off, you know, Chauffeur type of vehicle and all of a sudden you hit a bank of fog, like how long is it going to take for the car to get the driver back into the loop? In the meantime, you know, you need a sensor that's not going to be impacted by weather there.
Yeah, so I think that, you know, now that we're approaching production of the imaging radar, we're using B samples, we're getting like much more confident in the spec and the performance. We do see it as like a big advantage in scalability and cost. I think that that's, and this is kind of showing up in the data too, like when we added the imaging radar belt to the Volkswagen test vehicles, like that was another sort of MTBF boost. We're getting these like chunky KPI improvements, you know, based on specific technologies, one of which is imaging radar.
That is incredibly helpful. I want to kind of translate this back into the customer dynamics, right? Customer voice, because I think there's been a big focus on platforms, you know, choosing you guys as an exclusive partner, but you're working in a far more complex way with a lot of these customers. I want to just get a sense of, you know, as you start seeing some future proofing happen with the OEMs, the diversity of the segmentation that you're talking about, how should we be thinking about, you know, Mobileye's go-to-market strategy and tracking real customer progress? Not just with, you know, the big ones like you have with Volkswagen, but also some of the other opportunities that maybe aren't necessarily big announcements where you're doing a lot of work in the background that have long-term growth trajectories for the company.
Yeah, I think just to start with Volkswagen, the value of these programs goes way beyond the value of the revenue that we're going to get from it, because if it wasn't for Volkswagen investing in these programs, we would still be developing kind of in a vacuum. Having that customer that's holding your feet to the fire and keeping you to timelines and bringing up, like, well, what if this happens? What if that happens? How do you interact with the HMI? What about parking, DMS? All of these integration points that we probably wouldn't be able to do on our own, not to mention the customization work that Volkswagen is doing, because that becomes a proof point to other customers. Hey, this is not a black box. This is a system you can create brand identity on top of it. I think these are huge.
It also, in a way, makes you a little bit agnostic to decision making because you know the products are going to come out and at that point, they're either going to be superior to everything else or not. If they are, then the dam breaks, right? In the meantime, like I said before, this sort of knowledge that there's going to be different segments here. It's not just going to be base ADAS or navigate on pilot. Multiple segments means that running different programs with different suppliers or with multiple suppliers within a program just becomes too cumbersome for the organization. The one-stop shop, the fact that the chip is the same for each category, the camera for SuperVision, Chauffeur, and Drive, it's exactly the same perception system, camera-based perception system for each. You save a lot of money on validation. That becomes an advantage.
The scalability, our ability to, we demonstrated capability in China, right? Porsche and Audi, they don't want to have a separate, it costs them money to have a separate program in China. The fact that we can support China as well is a big benefit. Right now, I think what we've seen is we do think that there is more strategic clarity across a number of different OEMs about what they want for different vehicle segments. What are the criteria that they're looking for in a supplier? Do they want us to be tier one or tier two? Everything has become a little bit more formal. Right now, we're, and even the timelines have been holding more steady for, let's say, the last five or six months. There are a number of decisions that are targeted for the next few months.
We're very much hoping to have more proof points about more customers. I think in some ways, the advantages of Mobileye have become more valuable right now in a world where you're not putting all your eggs in one basket on an eyes-on hands-free. You also have to think about Mobileye Robotaxi. You also have to think about your path to eyes off. You also have to think about this lower cost segment that needs better performance.
Let's take it down to results, right? You guys just posted a great 2Q, right? You had, really, we thought, constructive 3Q guidance. You judiciously kept 4Q guidance pretty modest, right? Like, you took the wait-and-see approach. I guess, what are you seeing in the market right now that's giving you that pause around 4Q and just wanting to keep numbers from getting ahead of what's actually happening in the industry?
I think we're not seeing anything specific, but I think we'd all agree that there's probably more on, you know, there's more kind of geopolitical uncertainty than, you know, typical, right? I think that, you know, yeah, I think the market has kind of looked at it as like, okay, either they know something bad is coming or they're just sandbagging. I don't really think it's either. We have kind of, you know, non-cancelable purchase orders for Q3, so we, you know, essentially know what the volume is going to be in Q3. We don't know how many vehicles are going to be produced in Q3. We've got, you know, a sense, but you know, what if something kind of like lowers and you end up with like extra inventory on Q3 and it impacts Q4? That could happen.
You've got high car pricing, you've got high interest rates, you've got tariffs. I think if you look at it, you know, mathematically, we did 18.1 million chips in the first half. The high end of our guidance for the second half is 17.4 million. The midpoint is 16.4 million. We definitely don't want to be below the midpoint and see that as, you know, pretty high probability. If you look at the IHS production forecast for the back half, our top 10 customers, you know, they're expecting our top 10 customers to produce about a million less cars. Going from 18.1 million to 17.1 million wouldn't be crazy at all, right? Now, is IHS being overly conservative because they expect, you know, some weakening in the consumer, some weakening in the environment? Maybe.
I think like the bottom line is like sort of the mid to the high end is kind of where IHS numbers would sort of imply. If we're looking at what our customers are asking for for the rest of the year, we'd probably be above the high end of guidance. We're not really willing to, you know, take that risk. That's the sense is we're not seeing, you know, we're not hearing about anything specific. We're not like seeing any kind of lowering of orders. We just wanted to be kind of, you know, rational and kind of, you know, continue to rebuild credibility after we had a tough year last year with near-term results.
That makes all the sense in the world. I guess, as you think about not only the back half of this year, but moving into 2026 and thinking about the value proposition and value sales, or value-focused sales, can you talk a little bit about the pricing dynamics? With some of these emerging markets, we always see a lot of crosswinds in terms of sometimes wanting to lower price to enable new markets. This seems like a different end market in terms of creating extra value to enable new functionality, which should be outsized extra value to monetize. How do you guys think about that as you get through the balance of this year and think about potential revenue growth through not only units, but also through incremental value to your customers?
Yeah, yeah. Pricing has been challenging in China, given the mismatch between how OEMs value price versus performance, let's say. You don't have the same testing criteria in China that you do in other countries, and maybe the consumers aren't as kind of focused on it. We did, yeah, but it's been stable for the last year, right? Middle of last year, when sort of volumes had become concerning to us, we reopened discussions with a number of OEMs in China. We agreed on a lower price, but also agreed on kind of export dynamic, right? What's very clear with the Chinese OEMs is they are not confident in their ability to hit kind of ADAS performance targets in Europe and some other markets. They need allies like Mobileye. It takes resource to do that. We're not going to do it for just very small volumes.
There was some level of commitment to volumes in China. So far, we feel like we've kind of found this equilibrium point where our prices are higher than some of the competitors, but there's still value there for the customers. We don't see a lot of pricing pressure in Europe and the U.S. because we're continuing to get higher safety standards that are requiring more functionality and more features. This dynamic of each generation, you give more for the same price, is still holding in well. I think surround ADAS creates a different environment where all of a sudden you're going to have a chunk of volume that is at much higher content per vehicle. Does that mean that the vehicles that are left behind that are okay with lower ADAS performance, does that create some pricing pressure? Maybe.
I think that kind of the triple, triple the price for surround ADAS as regular ADAS, more than overwhelms anything like that. We feel like we're in good position. It's competitive like it always is, but we're not seeing anything to be concerned about on pricing.
Yes, one last one before we get to a couple of final financial questions, just the learning cycles between China and some of these other end markets. Obviously, with a lower bar around type certification and some of the cycling that you're able to do in China, and then this dynamic of around, can that technology that's been developed in China really get exported into other markets? We've seen early stages in other hardware markets there where the view was it wouldn't travel well, and then it ultimately did, right? Curious what you guys are seeing around that kind of cross-geography pollination and how strict the rules are and what you can take from China into other geographies to help move those markets forward in a new way and how those competitive landscapes might start shifting over time.
Yeah, I mean, we're definitely seeing a bit of like a two markets developing where there's not a lot of cross-pollination between systems that are developed in China and systems that are developed in Europe or the U.S. using data from Europe and the U.S., using hardware from Europe and the U.S. I mean, I think some of it's geopolitics. Some of it is inability to really access a lot of data in these markets, and it makes it much more difficult to develop kind of hands-free systems if you don't have enough data from the market you're trying to move into. We haven't seen any competition on the base ADAS side from Chinese suppliers for North America or European programs. It still seems like the performance of those chips is not good enough to get anything.
We've seen some bad examples, like BYD tried to deploy a system outside of us in Europe, and it got the worst safety rating ever given in Europe. That was a Western supplier that provided that. We feel like we've got really good competitive advantage in the base ADAS space. Higher level systems, I mean, it's interesting. Probably most people have seen these videos of this benchmark test that happened in China, and these were all high-end vehicles, Chinese vehicles that went through 20, 25 different ADAS, more like ADAS type of tests, and generally failed, right? These were not systems with Mobileye technology. Even the high-end systems that are, you get in them and they're driving you, right? It's very impressive, but don't have really much ADAS performance as well. That's something that the OEMs have to think about as well.
You can't just go with something like SuperVision or Chauffeur functionality and forget about the safety criteria. I'm kind of rambling at this point, but yeah, we're seeing sort of two markets develop. There's still definitely openness for Western OEMs to use Western systems in China. We're seeing that. We'll see what happens.
Okay. I guess the final question, because we're running into time here, is really around the Intel sales shares, right? You know, that happened in an orderly way. You guys participated in buying some of that stock back, and given some of the cash issues that that platform has, obviously, you're not representing Intel, but just want to get the basic sensibility from the Mobileye perspective around that relationship, which seems like it's been a pretty productive one today at this point.
I think the relationship's still very good. They're a great owner. They kind of leave us alone, but they provide support or kind of advice when necessary, like the board members. We really appreciate the Intel employees that are on the board. I think it has to be a balance, right? My interpretation is they're trying to balance their cash needs. They're trying to balance their sort of public statements to investors about divesting non-core assets with trying to balance that with other public statements that they make, like we continue to have strong conviction in the long-term growth opportunity of Mobileye. I think that they didn't sell for two years, stayed well above 80%. I think they probably needed to show some action on divesting non-core assets, and they did that.
I think the interpretation of the market is that maybe they're not so price sensitive and that this is going to be a continued headwind. To me, I think that that's the bigger kind of source of pressure on the stock over the last month or two, or month really, than anything about the guidance or the near term. I think we see a lot of positive catalysts in the near term that could change sentiment. I think Intel is aware of those as well. We don't think that they're in a rush. As far as the buyback, this was a one-off. It shouldn't be seen as a change in our kind of general capital allocation strategies, but I think, like you said, showing support for Intel's process, helping to maybe accelerate the sell down at least in a small way. We did about 10% of this offering.
We think our end kind of expressing that we see the current valuation is very compelling. We're all reasons why we thought it was a good idea to do the buyback. I think we'd probably do it again if there were other offerings, but it shouldn't be seen as the start of a different capital allocation strategy for Mobileye.
Perfect. Listen, we are running into time. Dan, it's always a pleasure. I love the capacity to go in a wider interview conversation with you. Just for everyone in the audience, we're happy to help you get a little bit deeper into the Mobileye story as needed and put you in touch with the company as it makes sense. Dan, thank you again. I hope everyone has a great balance of the day here at the conference. We'll talk to you all soon.
Thanks a lot, Colin. Great job.