Intel Corporation (INTC) Earnings Call Transcript & Summary

Good afternoon, and welcome to the Wolfe Autos Corporate Week. This is the Mobileye Intel Fireside Chat hosted by Wolfe's senior auto analyst covering autos, Dan Galves. [Operator Instructions] And now I hand the call over to Dan.

Thanks, Tatiana. Good day, everyone. Thanks for joining in as we continue Wolfe's Auto Corporate Access Week. We're going to hear today from one of the key leaders of Mobileye, a division of Intel. Even though we don't cover Intel, we think it's really important for anyone interested in the future of transportation to hear from Mobileye as they're generally well ahead of the peers in identifying trends. So we continue to pass through them to join us. In the 2000s, when the auto industry saw radar and stereo vision cameras as the sensor for future active safety systems, Mobileye stuck to single lens camera technology, and they were 100% right. Several years ago, they coined the term Level 2+, pointing out very important benefits in cost and risk versus the mainstream direction for Level 3 autonomous. They were right about this. Recently, we've sensed that they're identifying some really critical concepts on how transportation will be consumed in the future and how it will be controlled that go well beyond simply replacing human-driven vehicles with autonomous vehicles. So I'm excited to dig into that today. Before turning it over to our guests, I just want to mention that these webinars are always most interesting when we get good audience participation. [Operator Instructions] Your questions will appear on my screen. They won't be visible to anyone else. I'll ask them anonymously. Okay. We're really happy to have Erez Dagan, who is Executive Vice President of Product and Strategy for Mobileye, an Intel company. Erez has been with Mobileye nearly from the beginning, in the early 2000s. He led the ADAS business development efforts for years in terms of securing contracts with OEMs and is now also the lead on Intel's Mobility-as-a-Service initiatives. Erez, let me turn it over to you.

Thanks, Dan. I really appreciate the opportunity to speak to you all. We'll focus in a very few slides I prepared for this opening session. We'll focus on the strategy that we've been forging around Mobility-as-a-Service, and I'll be very glad to go into questions and answers afterwards more broadly as needed. So as we can see in the first slide, we've identified that there are 2 streams of evolution of transportation as we look into the future. We have the existing traditional public transit operators, PTOs; on one side and the more recent type of players, the transportation network companies or TNCs in short, both with their unique characteristics and differentiation elements. For one, we can denote that the PTOs are accustomed with high capital mission profile and high labor, while the TNCs avoid both capital and labor through a matchmaking demand and supply of existing drivers and riders. Both of them are seeking or both these camps are seeking the path forward to autonomy to proliferate autonomy through their business models, and we identified the different go-to-market models that could fit these 2 slightly different phenotypes. So if we move to the next, second slide, where the second slide depicts the go-to-market models that Mobileye and now with Moovit has decided to target. I'll start with the obvious one, the straightforward one. Mobility service supplier as an automated TNC is certainly one of our go-to-market channels, go-to-market strategies. And in that, we are -- there are other players such as the ones denoted on the left side of the right cube down there who are predefined as automated TNCs. In addition to that, there are interesting models of engagement, which allows us to effectively and in a mutually rewarding manner build an introduction of a broader introduction of autonomy solutions reaching the markets. The first one is RaaS, Ride-as-a-Service, which means at large that under another player's demand umbrella, your autonomous driving solution is operating, meaning, for example, an Uber or a Lyft or any other of these players has a request for a ride and it can get that ride from our fleet in better economies or with better ETA and service levels. And we feel and we assume that, that's going to be the case, so that's definitely an engagement model we are building and forging forward. And this, by the way, could also evolve to a symmetrical model of getting Ride-as-a-Service from other players as well to offset demand and supply spikes and uniformities. And on the other side of it, we see the same Ride-as-a-Service model as an effective one to collaborate with the public transit operators. But more interestingly, given the public transit operators DNA of capital-intensive operation and physical operation and hyper-localized services, physical services and maintenance of the fleets, we have devised a third model, which is Vehicle-as-a-Service, which means that we will either sell or lease a fleet of autonomous vehicles or shuttles to public transit operators with another layer of software of mobility intelligence to assure optimal utilization of those assets of this capital. I think that just from looking at these 3 models, one can understand the synergy between Moovit and Mobileye in addressing these objectives, catering to these objectives. Moovit has been a -- has been digitizing. In the very short way to say it is that Moovit has been digitizing transportation, urban transportation, for quite a while at a very, very large scale. And we feel that having the background of the digitizer, the one that digitized it, and the go-to-market channels, B2B channels and B2C channels of Moovit, we can be the one that takes it to the next step together and automate what's already been digitized. So if we move to the next slide, we can look a bit more carefully at the different stacks of the solution and how -- which layer of the stack is contributed by which of the players. Of course, first and foremost, the self-driving system, the HD mapping. And I would also say the teleoperability of the system, system being able to be teleoperated from remote our assets developed by Mobileye at its core for many years now and with well-established accomplishments in those field -- in this field. And the stacks above fleet control center, including an extension to teleoperation, the critical mobility intelligence layer, mobility intelligence platform and services. Mobility intelligence in short stands for the brain that oversees the matchmaking of demand and supply given the instantaneous information as well as observed patterns. The objective of that layer is to maximize the utilization of the vehicles, the fleet vehicles. You want to maximize the time that the vehicle drives while being paid, the ratio of paid mile versus unpaid miles. And mobility intelligence is actually really measured by that KPI, how to make sure that the vehicles are traveling the least amount needed without being paid for it. And when it comes to more interesting models of pool drives, it becomes even above 100% utilization if you can assume that, on average, you have more than one person on board. That's an interesting -- that's where the economies take very significant leap forward as well. And of course, on top of all, layer 5, the mobility users, the access to mobility user and partner network that I already mentioned. The assets that Moovit brings in is very critical, as I said, for the B2B and B2C channels that they have for the users and partners. But also given the very rich information and understanding of mobility -- urban mobility patterns on a global scale, it is key for devising a unmatchable mobility intelligence platform and to help us in Mobileye, Moovit as the mobility service unit inside of Intel to preplan and devise our mobility service deployments in the utmost rewarding manner. We can really identify the needs, the gaps, the geographies that require it, the business cases in advance and tailor weave more carefully the solutions, the autonomous solutions into the urban transportation fabric. We have also other areas of synergies with Moovit around public transportation that concerns a uniform proposition to public transit authorities around devices and data that allows them to monitor both the safety and the flow of these vehicles of the fleet of public transportation vehicles. Both companies have been working in front of public transportation operators separately, and we think that even before Mobility-as-a-Service or Ride-as-a-Service or Vehicle-as-a-Service, we have a very interesting and compelling proposition for improving the safety and flow of traffic today. Then I think that's -- these are the slides I've prepared. Be glad to answer any questions.

Thanks, Erez. Again, I wanted to introduce Trey Campbell as well, Head of Intel IR, who's with us today as well. Yes, that's super interesting, and we're going to try to unpack some of what you just said in a minute, but I wanted to start off with just a couple of quick questions on kind of what the current bread and butter of Mobileye, which is the ADAS business. You've had a lot of success sourcing Level 2+ type of programs for consumer-owned vehicles. Are the near-term launches on track? Have you seen any meaningful delay from customers in the timing of the ADAS launches?

So actually, the launches, as far as we see them, are on track. There is no delay that we can measure. There are some, of course, due to the COVID, some closures and some reduction of volume, but we don't see any latencies of major programs or slowdown due to the COVID. We continue to see a very healthy design win pipeline even throughout the COVID across a wide range of OEMs. For example, Q1 '20, we had 14, 1-4, new design wins across 12 unique OEMs. Above 10 million lifetime units already secured only within that quarter. And interestingly, I do believe that the COVID would have a positive impact in the sense that OEMs are likely to go more for an off-the-shelf proposition than developing custom technology. I think that budget-wise, we will see a bit more uniform product, seeking a uniform off-the-shelf product rather than tailoring intricate features per OEM. Of course, not all OEMS, but the long tail will probably take that approach.

We've heard that across a wide variety of parts of the vehicle. So that makes sense. And I think in some ways, like people don't really know what to expect from these kind of next-generation Level 2+ systems. So could you quickly kind of talk about like what will these systems do initially? And are there ways to upgrade the features over time over-the-air or kind of rapidly through product refreshes?

Sure. So Level 2+ or, as we also coined it, pseudo autonomy, just to make sure we understand what it is, it's the case where the -- we deliver an autonomy-like experience without shifting the liability away from the driver. It allows -- the reason that the market gravitates to that direction is that the alternative of shifting the liability away from the driver is at least at this point with a challenging business case, meaning the redundancies and the overheads that you have in order to take your liability away makes the OEMs divert more and more resources in this pseudo autonomy proposition, into this pseudo autonomy proposition. And this proposition has, first of all, this comfort contribution, as I said. It's usually -- it's comprised of surround perception and HD Map as a differentiating -- as a, sorry, defining features. Very efficient. Very cost efficient. As I said, point-to-point navigation experience. And interestingly, when the vehicle is not driving for you but the driver is in control, the surround perception and the map that we have also gives immense safety benefits. It allows us to trickle down the safety concepts that we have developed in AV, in autonomous driving, to the human-driven domain. Our RSS concept can help drive the vision zero goal by offering holistic RSS-based safety envelope through preventive brake or steering actions. So the vehicle is monitoring the driver actions and keeping it within the formal safe envelope that we have identified when we devised the solution for autonomous driving. Same watchdog, if you want, and that would give AV-like safety assurance. It's a very interesting promise to say that you can tackle the safety problem well in advance to the proliferation of full autonomy. It's a very important message to say here. Over-the-air is an exciting part of this autonomous -- sorry, automotive industry evolution. There are the obvious monetization advantages it gives. You can slice and dice and sell parts and monetize more value-based pricing, et cetera, et cetera. That's straightforward. But the significance of it to the automotive OEMs is in pulling them out of the dungeons of very stringent validation and demanding validation processes because of the irreversibility of a vehicle deployment and a function deployment. If you have it over-the-air updatable, the criticality of the launch time is different, and the validation cycles can become much more agile. It saves a lot of time to the OEMs and allows them to introduce the functions in a much more attached to the state of the art and not linger behind because of the rightful need to secure the validation without the ability to update.

Thank you.

Dan, quickly, that goes without saying, our software/hardware are fully supportive of the over-the-air updates, and it's just a matter of enabling it on large scale by the OEM makers.

Got it. Got it. Last one on kind of consumer-owned vehicles. We've been hearing that OEMs are targeting a set of software and central compute unit that can scale from fairly basic ADAS to very sophisticated pseudo autonomous, as you call it. Is that correct? And what kind of strategic advantages does that provide to you, if any, and your Tier 1 partners?

So actually, it's a bit more complicated than that. There are 2 modes of operation of the automotive industry in that sense, and I'll explain. Indeed, the centralized architecture scalability is one driver. It applies usually to mid and high variants of vehicles because that's where there is not necessarily a uniform function proposition, and the OEMs still need to find a way to avoid revalidation of hardware and software components for this versatile propositions. So in those mid and high end, yes, a centralized architecture is sought by the OEMs. When it comes to the ultra low end, the basic safety-only vehicles, the OEMs are very keen on a cost-effective power-aware, power-effective solution that can be mounted on the windshield. And that does not have the intent of -- does not have the aspiration of scaling it to the higher -- all the way to the higher end. So it's a bit -- there are 2 modes of progress, especially given the recent NCAP fortifying itself as a leading safety standard and the GSR regulations, the mandates that are proliferating now for 2022 and 2024. So with this dual, we are well positioned to deal with these -- both these tracks. Our SoCs are, to begin with, well-designed to seek on a on-windshield mission profile. That's the lower -- ultra low-end vehicles that require a very cost-effective and power-aware solutions. That's where we start off, and that's our bread and butter. The -- that same capability is also in our favor when we come to talk about scalable solutions because the families of chips, SOCs that we've recently released, are scalable in a way that you can extend the computational space by adding components. EyeQ5 high, for example, is a combination of 2 EyeQ5 mids with high bandwidth with [indiscernible]. And as we move forward to EyeQ6, we will -- we have an even more careful attention to the details of scalability, to the design of scalability. And we start from the windshield and can extend it gracefully while allowing the OEMs not to validate again the hardware or the software deliverables.

Got it. Got it. All right. So let's move to kind of mobility services. So after kind of listening to your opening, it almost feels like you are looking to rely on others to drive demand to your vehicles. Is that -- do you have ambitions to drive demand yourself? Or are you really more looking to plug in autonomous vehicles to kind of existing services?

Great. So just for extra clarity -- guys. Sorry, guys. Let me close the door. I apologize. Sorry for that. Again, if we go back to enumerating the business models -- I apologize. If you go back to the business models that we just saw, we have the Mobility-as-a-Service, we have the Ride-as-a-Service and the Vehicle-as-a-Service. Mobility-as-a-Service, we're definitely going to drive our own demand. We're going to rely on Moovit as the -- which has a very, very large urban transportation user base to be that demand driver. And as said, we're going to be much more surgically introducing those services such that the demand will actually be there. We can look at the patterns of demand and supply, look for the glitches, look for the gaps and really target the business. So by all means, we are going to drive our own demand. The other models are essential because you want to leverage all of the channels available to proliferate your solutions. We're going to have the autonomous capability, the very differentiated autonomous capability, both value and cost-wise, and we are pretty certain about that. And we want to make sure that we have all of the monetization opportunities for that capability.

Got it. That makes sense. That makes sense. Speaking of capability, I think the first step is the service in Israel that you announced a couple of years ago with Volkswagen. Can you talk a little bit about an update on timing of when you expect to have vehicles operating without drivers? And from a self-driving standpoint, really, what else needs to be done?

Great. So we definitely are on track for a '22 launch with -- in Tel Aviv of a service in this partnership. We have been working for the past several years to scale our technology here in the backyard in Jerusalem. We have the large fleet of vehicles traversing a quite wide network of streets in Jerusalem, quite broad. The technology, just to step into that avenue for a second and technologically, our RSS concept proves itself in a sense that we are generalizing very easily to very versatile traffic environments, optimizing agility while assuring the safety. The Israeli backyard is exactly the place to test that. I don't know how many of you had the pleasure, but driving culture is not very different than our high-tech culture. We just go at it and maximizing interest. So it's very, very, very interesting place to test and develop the capability. The Tel Aviv deployment also, very interestingly, is going to be a generalization test. We're not planning to develop too much in Tel Aviv. We're planning to take what's been developed in Jerusalem and as another indication for ourselves for the -- what's the word I'm missing?

Scalability? Or...

Yes. For the robustness of the solution. That's the practice from machine learning. It's called train versus test. You need to make sure you don't train on your testing session. So we're going to develop in Jerusalem. We take it down to Tel Aviv and have a very strong measure of our robustness very easily.

The last time you had investors over in Israel, you talked about kind of a camera-only system and kind of a RADAR/LiDAR combined system kind of operating in different vehicles independently for validation purposes. How is that process coming? And when will those 2 be combined?

Perfect. So the true redundancy is indeed the approach, having 2 separate systems, tackling the problem so that we can -- the primary objective of that, as we mentioned, in Jerusalem is to reduce dramatically the validation burdens. Okay? We need to collect much less data if we can validate each of the systems separately on almost square root of our target meantime between failures if we want to be -- I'll repeat that, if we want to be -- achieve, prove to ourselves that we achieved 1 million hours between failures. When we have 2 separate systems, a mere 1,000 hours of each system would suffice, and then we duplicate those meantime between failures. So we are well on track on that. We have a completely stand-alone RADAR and LiDAR-based vehicles. They will also traverse in those same routes, including localization, including egomotion, no dependencies on the camera whatsoever. The last stage of patching those together to get redundant RSS decisions or safety decisions overseeing the driving policy is planned for later this year. And that's where we're going to start launching in Tel Aviv for resilience testing.

Excellent. Excellent. And then how about the cost of the system? Your CEO said in November that the system could be done for $4,000 to $6,000 by 2025. How is that possible when most others are targeting $50,000, $60,000 as kind of the at-scale cost? And what type of cost per mile does that support relative to human-based driving? If you can give us an example.

Okay. So last bit of the question varies, but let's talk about it. There are other factors that would trickle it down to cost per mile, but let's talk about the system cost. As for the system cost, definitely, we are very competitive there, starting with the simple fact that we have been devising and optimizing for cost for 2 decades our computer vision subsystem. We are monetizing on it. We're productizing it. We're battle testing it in the ADAS space. Okay? That gives us something very efficient to go to the road with -- when we hit the autonomous driving challenge. That's one step. Second element that we cannot overlook, the importance of RSS in this. Having a safety scheme, which is explicit, model-based, clear, gives us a very clear goal when we design the system for safety. We know exactly what to protect against. It's not that we're -- there's an Israeli term of shooting from within the cage, not looking at what you're shooting. When you have a safety concept such as RSS, you can prescribe your sensor set much more tightly to the needs, to the safety needs. You can prescribe your compute much more tightly to the areas that you need to protect. You know that once you protected these areas to these degrees and these areas to a different degree, you still covered the RSS objectives. So that's a key element that was used from the beginning, from the design, the early design of our system and made it very cost effective. The true redundancy is one outcome of that RSS philosophy. It's one -- another example of how it affects the design of the system at its core. So by all means, we see a path forward to the figures, the focuses Shashua mentioned. There is going to be some learning curve or some optimization phase where the Mobility-as-a-Service introduction '22, we will still not be at these costs, at these figures. But that's exactly -- once we get close to the single low-digit thousands of dollars cost per system, that's where the second phase of autonomy will launch, which is consumer AV, and that's where we really proliferate the technology through OEMs as partners.

Got it. Got it. And just this -- understanding that cost per mile has a lot of variables outside of the cost of the system. But maybe put -- I'll put it in terms of a question that I'm getting from the audience is, do you see suburban activity for rideshare as viable? Because, I mean, in my opinion, that's a cost issue. Personal cost of ownership is very low in the suburbs. So do you see that as an attractive market, the suburban areas? And can you get there on a cost to give a better value proposition than owning a car?

So we will definitely get there earlier than more expensive systems. But it's important to say that when you look at the cost per mile, the 2 dominant factors are the cost of the self-driving system and the utilization. Utilization is a macro element. It's something that has to do with the demand density, the demand patterns. If you cannot utilize the vehicle, for example, shared or pool drives in suburban area is not a very easy thing to imagine because of the sparcity of the demand and the misalignment of the trip requests. So yes, in later times, there will be a model that would allow, for example -- sorry, tackling the suburban, the more common model to assume is what's referred to as first/last mile. In the suburban area, driving to a well-polished public transit funnel, say, the train station, would be catered by A to B robo taxi services. And from -- dispersing away from those stations back to home is the more viable economical story to tail there. As we move forward, of course, and proliferating autonomy to consumers, interesting models of car sharing could pop as well, like car sharing between the users as we move forward.

Got it. Got it. Let me ask you about how Mobileye and Intel's thinking on fleet-based autonomous has evolved. There was a time when your efforts were kind of all about partnerships and contracts with OEMS. Then my perception was that you decided to focus on the self-driving system yourself. If you create a good enough system, you're going to have plenty of customers for it. And then it looked like you were kind of pursuing your own fleet of autonomous vehicles. And now it looks actually much broader where it's a partnership model and a variety of different -- I don't see really any inconsistencies there, but talk a little bit about kind of how the -- what drove this evolution in your thinking.

Very good. So I think that the traditional industry of supply management, the traditional OEM industry, which expertise is supply management, supply chain management and is usually effective in face of a known problem, known objective. When you go out to exploratory spaces where you need to learn how to build an autonomous system and where you need to learn how to drive a service, that's not necessarily the right DNA, the kind of -- that's in -- in machine learning, it's a difference between exploring and exploiting. Exploiting not in the battles, about using and leveraging existing assets and capabilities. So I think that when it comes to exploring or in our terms, in nascent markets where things -- the value chain is still not clear. The requirements from the different players in the value chain are even not defined. You cannot drive it by suppliers' management. So what we said, we wanted to reach independence in the areas which are more exploratory. Self-driving system were more exploratory back then. Self-driving system, the Mobility-as-a-Service, both these drive -- these steps up on the vertical value chain were driven by the fact that we know that we have a differentiated technology and capability, a very strong asset, and we don't want to depend on the market learning curve when we want to monetize on it. So we need to create the monetization opportunities for ourselves. Dan, I lost you. [Technical Difficulty]

Guys, can you hear me?

Yes.

Am I back? Okay. The old reboot method worked. Okay. Sorry about that. So my next question was basically, as this thinking has evolved, like how does Intel Corporation fit in? Could you have had these ambitions without becoming part of a bigger entity?

No. I think that's right on the spot, Dan. The privilege to think bigger and to be masters of our destiny in terms of building our monetization channels and building the technology with less coupling on existing links on the value chain is definitely by the backing and the capital access we have by Intel. To say clearly, we are currently a self-funded unit in the sense that our revenues drive our development, including autonomous driving. But for example, inorganic growth, as we did with Moovit, is not a step we would have considered that easily without the strength and backing of Intel. Also important to say that in the day-to-day project that we have in ADAS, we get a lot of horsepower from Intel from day 1 around optimizing compilers to our hardware or working on customer platforms on software and hardware designs. So very, very, very substantial contributions since we started being part of Intel.

Got it. I've been intrigued by this idea of personal transportation unbundling. So maybe people still own one car for long trips. But other use cases, different modes could be chosen depending on kind of what was the specific use case, the urgency, the traffic situation. And that could all be controlled through like a super app, which could consolidate lots of different transportation modes from different services, could price compare, could dynamically route, could process payments. Do you buy into that view of the world long term? And is Moovit a platform that can really help you guys be a -- to take a large role in that process?

So definitely -- and not just preferences, but circumstances call for different transportation modes. In some cases, you won't have a choice but to order a taxi or you won't have a choice but to go down to the subway given your time pressure or context that you're operating. And we definitely see the value of aggregation, not just due to the business -- the immediate business interest that you can see there, but also because at the end of the day, the city is an organism that needs to be sustainably served. And being able to see all of the modes interplaying with one another is key to responsibly introduce mobility solutions into that city. If you don't see that, if you operate out of context, you can easily burden the city in a way that would burden you economically, but would burden society in a way that is really not a very good business to go after.

And clearly, from the -- from your opening slides, it looks like you have -- you're very open to working with TNCs today like Uber and Lyft. Are you also open to working with companies that people see as your competitors like Waymo and Cruise and Argo? And kind of how would -- if so, how would the services work together?

Premature to answer this one, I have to say. We're open, as I said, for any monetization channel of our differentiated asset as cases proved in the past. The one with the differentiated assets, usually there is an incentive for the other side to engage and build a collaboration with. Think about if whatever other player of AV, the cost per mile would be higher -- by definition is incentivized to at least, to some extent, refer rides to a cheaper ride supply -- supplier. A classic example. But it's really very way up there to -- sorry, ahead of us to talk about these specific collaborations.

I see. What's your view on kind of -- it sounds like you -- well, what's your view on mixed fleets? Like is there -- is it possible to have kind of a fully autonomous rideshare service? Or do you need to supplement that with human-based drivers for -- to go outside of the domain or for kind of demand matching?

So actually, that's part -- that links back to your question about aggregation. When you are operating in a context, you can be a bit more selective in what you're offering and when you're offering it. And by that, not burden yourself with a mixed fleet. If you are only catering to demand that fits the geography that you have targeted your AV to operate in while other means of transportation can serve out of that geography, and as I said earlier, if you tailor the geography to where the demand lies, you are minimizing or avoiding altogether the need of human-driven fleets.

Got it. Got it. Let's see. I've got a couple of more. I think we're nearly out of time. So back to the cost of the self-driving system and your comment that there is a path to consumer-owned AV in the future. Are you back to kind of the current ADAS model in that world where you are kind of contracted with OEMs with a Tier 1 integrator in the middle?

Definitely, definitely. We want to -- as I said, we want to forge the autonomous driving proposition in the Mobility-as-a-Service space. And once it's forged and we have the blueprint, let's say, we definitely need the traditional supply chain and we definitely can benefit from the expertise of manufacturing at scale, testing at scale, integrating at scale and all of the other qualities that the -- our partners up the value chain bring.

Got it. Two more. One is, can you kind of give us an example of how you would work with a public transit operator? And do you have -- I think you have projects in that vein. Can you just give us a little bit of detail about how you expect to interact?

So the models are -- as the ones that I showed in the initial slides, one key model is a Vehicle-as-a-Service, meaning that we will lease or sell them pre-installed autonomous vehicles or shuttles and a software layer of mobility intelligence that allows them to operate these vehicles in a way that maximizes their utilization. And by that, making the economies of the public transit operator optimal. Another model of interaction or engagement that we're exploring is similar to the one that we're heading with the TNCs, which is Ride-as-a-Service. Let's assume that a public transit operator has a large train station out of which or into which he wants to get people, and it does not necessarily want to equip itself with owned autonomous vehicles to do that. He can summon rides of our vehicles traversing that area and really buy packages of traffic into and out of that train station.

Got it. Got it. Okay. And last one, I have to ask one question about Tesla since there's so much debate about what role they're going to play in this world. What's your opinion of kind of what role they will play? And do you agree that there's a large data advantage that Tesla has by having a cell link and a WiFi link to hundreds of thousands of vehicles operating?

So I can't relate to any specific competitor, of course, or any specific company, but our access to data and access to our design -- by design, access to data and access to knowledge and understanding of how to solve the autonomous driving problem itself and the mobility service problem. On the other hand, with the Moovit acquisition, places us in a very, very differentiated position compared to many other players.

Got it. I really appreciate the time. I thought that was super interesting. Thanks, Trey, for supporting this. And thanks, Erez, for getting on the call with us. Hope to talk to you soon, and good luck.

Thanks, Dan.

Thanks, Dan.

Yes. Thanks, guys.