Arbe Robotics Ltd. (ARBE) Earnings Call Transcript & Summary

Hi, everybody. Thank you for joining us. I'm Jesus Gonzalez Lopez, auto analyst here at JPMorgan. And I'm here with Arbe Robotics and its CEO, Kobi Marenko? Go ahead.

Thank you. I'll try to do my presentation short and leave some time for questions. So I'm Kobi Marenko, I'm the Co-Founder and CEO of Arbe. Arbe is my third venture as the Co-Founder and CEO. My previous company went public on the London market more than 10 years ago. The name is Tremor, traded also now on the NASDAQ. And my previous company was merged into an American company called Digital Turbine. After that, those 2 companies were around advertising in mobile. And then I thought that I should do something different. So we were -- my co-founder and me were looking into problems of safety and navigation in automotive. And we came to the conclusion that any car without a sensor that can see around it in any weather and any lighting condition won't be safe. And the only sensor that can do it is the radar. So we began developing chipset for ultra-high resolution radar. So 7 years past, and we have 100 people, mainly in our Tel Aviv office, but also in U.S., China and Germany, where we are supporting our customers. We have chipset on the final stages of production that consists of chips, transmitter, receiver and a processor. Our processor is our core IP. We are well funded. Just recently, we raised a follow-on on top of the first $100 million that we raised when we were listed on the NASDAQ of another '23. And we believe that we have sufficient cash in order to be profitable somewhere around early '25. We have $40 million of preliminary orders for 2024, mainly from the Chinese market and around 10 OEMs are in a final decision of adopting our technology. So the first wins that we have are in China. China is right now the most advanced market in terms of EVs, electric vehicles adoption and EVs mainly coming with a better ADAS features, better hands off driving features and so they are adopting sensors faster than anywhere else. We have a premium model that will go to production with our technology in early '26. We were selected by a major European premium brand for their China fleet. And we have 2 wins already with L4 players that has a large fleet, AutoX, the largest RoboTaxi in China and DD's trucks. So basically, when we are looking on the automotive market, what we believe is going to be the main -- and most of it, the only business around is what we call L2+, L2++. It's hands-free driving from point to point. And we designed a sensor that will enable perception for this functionality that will work in any weather, any lighting condition up to a very long range in ultra-high resolution, and we generate a full free space mapping of the environment that works in 100% of time from one hand, and from the other end, won't generate any false targets, which is the main problems of many of today's sensors. So we believe that hands off and eyes off, this is the main application in automotive and also the analysts see it like that. And Level 4 that is called mind off is definitely -- we won't be able to see major revenues from those kinds of applications in the next 10 years. Our chipset consists of 3 chips, transmitter, receiver and a processor. Our core IP is our processor that offers the ability to process 48x48-radar channels that generating around 1 tera of data per second. We are doing it with our proprietary IP that basically knows to squeeze the data, to crunch it and to take this theoretical 1 tera of data and generate from it 1 giga Ethernet point cloud that can be processed by the central computer of the car. We believe that one perception radar like we are producing is a must for L2+, L2++. L3 might be needed another 2 or 3 lower-end imaging radar, but still strong radars. And for L4, L5 customers are talking about 6 high-end radars per car. Our solution is the only solution that enable free space mapping for L2+ and L3 without LiDAR. It functions in any weather and any lighting condition, can see to a very long range in a very wide field of view, and it supports a mass production radar in a cost of $150 that basically can be adapted in almost any car and any model. This is a video showing our technology in action. On the right side, you can see the point cloud that is generated by our radar in real time. And in the center is our AI stack that takes the point cloud and generating a full free space mapping of the environment, showing to the car what is drivable and what is not where it can drive and where it cannot, where the road ends and of course, if there is pedestrians, cars, motorcycles, bicycles, we are able to detect them easily and to make sure that the car will adopt the driving speed or turn to other lane in order to make sure that it won't hurt anything. So our radar can work in real time, 20 frames per second and up and generate this full picture of the environment in real time, especially in cases that the camera face when the light is shining on the sensor or when we are getting into the tunnel and the camera hertz. Also, of course, the camera cannot work when there is problems of weather. The camera has a very limited range, but the radar solves them all of those problems. So basically, we believe that we are fit for L2+ and app, which is going to be a majority of the market in the next 5 to 10 years. We believe that RoboTaxi and full Level 4 will take long time and not sure that it will even come into a real mass production. The TAM of radar today is around $6.5 billion, and the total radars that would be sold in 2030 are going to be around 17. The high-end imaging radar will take something like 20% of this market, and we're going to be a big player out of it. Arbe is a Tier 2, which means that we are selling our chips and a reference design to the Tier 1. The Tier 1 are the big auto manufacturing companies like Magna, like Continental, like Bosch, and they are taking our radar -- our chips and building a full radar model out of it, doing all of the testing, qualification, production line that is needed for that and taking it into production. We are engaged today with 4 leading Tier 1s, Magna that bought Veoneer. Veoneer was a major player in the radars market, the third largest player in radars. Magna bought them. Veoneer selected us almost 1.5 years ago. And they are in the final stages of launching a base sample based on our radar, expected to be on the third quarter of this year and full production of radar based on our chipset in early '25, mainly for Europe, U.S. and rest of the world. In China, we are engaged with Hirain and Weifu. Hirain is the largest ADAS leader in China. Weifu is a multibillion Chinese Tier 1 partly owned by Bosch. Weifu already showed to the market a full sample and a production line based on our radar. Hirain are in the final stages, and we believe that by Q3, they will have also a full production line with our chips and will go to production. We also have a partner from nonautomotive application called Qamcom, and they are taking our technology for commercial vehicles, for off-road and so on, not for automotive applications. When we are looking today on the competition, so the radar today is a business that is controlled by NXP, TI and Infineon, none of them has a solution that can go into the high level of channels that we believe that is a must in order to generate a picture that is safe enough and can really make it for L2+, a real safe L2+, L2++ or L3. The only player today that is trying to compete with us on the same performance is Mobileye that developing a radar. They are between 2 to 3 years after us in terms of their ability to bring sensor to the market. And we also believe that their solution would be more expensive and not fit for a real mass market of L2+, L2++ cars. Worth mentioning also Ambarella that bought a company called Oculii. We haven't seen any traction for this product since they bought them 1.5 years ago, but we're still counting them as a competitor. So we are engaged today, as I mentioned, mainly with car manufacturers through the Tier 1s, but also with the car manufacturers themselves, the car manufacturers basically buying from us a full radar sample in order to start collecting data and to train their perception stack. And the full mass production is coming from the Tier 1s. So where the Tier 1s will be ready -- all of our Tier 1s will be ready with full production radar in Q3. Q4, we begin chipset production. And next year, we're going to be in a full mass production of radars based on our chipset mainly in China and in '25 in Europe and U.S. As basically, our revenues in '24 are mainly based on China. '25, it will be something like 50%-50%, 60% to 30% split and '26 is going to be something like 50% China, 50% Europe and U.S. Our guidance for this year is winning 2 significant U.S., European OEMs for active programs and 2 new Chinese OEMs. On top of it, 5 nonretail automotive partners like trucks, RoboTaxis or infrastructure. We just recently won DD4 for trucks, and we won a big project with Hirain for China infrastructure. Starting of automotive grade production by the end of Q4 and revenues are between $5 million to $7 million this year. The team is very experienced. Our Chief Business Officer is coming from automotive, spent his last position in TI, managing their sales to automotive companies. My co-founder, Noam Arkind, did his Ph.D. in Applied Mathematics on autonomous movements of robots. Last year, came to us -- joined us Gonen Barkan. He's our Chief Radar Officer, working on the strategy of our radar and our next generation is coming from GM. And worth also mentioning Shay, that is not here, that is our VP Operation coming from 25 years of experience in taking chips into production in automotive, which is a very, very, very complicated task. This is -- it as an intro, and I would love to answer your questions.

Yes. No, thanks so much for the presentation. I think I have a question to start off on the broader industry. Given your closer proximity to semiconductor manufacturers, the most auto part suppliers, can you provide any color on the supply and pricing environment and how that's tracking you to date?

Which, our supply?

Yes, your supply of semiconductors.

Yes. So basically, we're engaged with GlobalFoundries is our fab. We are one of their top 3 strategic clients and the first company to go into production with their automotive process. GlobalFoundries is also doing for us all of the testing and management of inventory from their facility in Vermont. The fab itself is in Germany in Dresden. And we have a capacity that they saving for us for '24 and '25. This is part of the lessons from the problems of supply chain. And basically for each car company that we are doing -- that we are submitting an RFP, GlobalFoundries is submitting a comfort letter that they will supply the chips on the volume and the time line that the OEM wants.

Got you. And just in terms of pricing, what are you guys seeing for semiconductors and the overall inflationary environment there?

So first of all, we see a pressure. So the automotive industry is under real pressure coming from the EV. The EV is expensive. The margins are low. Tesla pushing the prices down. And the only way for car manufacturers to generate more profit is from, I think, ADAS features. So they are trying to maximize it by pushing down their prices as much as they can and also using the sensors that are cost and effective. I think that we have a very good solution for that. So our chipset, as I mentioned, basically enable Magna and our supplier and our Tier 1 partners to sell a full radar model in $150, which is a very attractive price point, definitely as opposed to LiDAR that cost $1,000 and up.

Yes. No, I think you kind of hit my next question right on the head. But given OEM's focus on reducing costs across both EVs and ICE vehicles, do you expect to see increased quote activity for your radar solution as the industry kind of realizes the cost benefit you can bring?

Yes. So definitely, this is right now the case. So the car manufacturers are looking for new ways to generate revenues, ADAS and Level 2+, Level 2++ or point-to-point driving, hands-free driving, I think, is the only way for them to generate more revenues from services. And so this is, I think, going to be their focus for the next 5 to 10 years.

Got you. No, that's helpful color. And then lastly, on the ADAS side. A lot of OEMs are talking about pushing out their time line for full autonomous driving. Ford announced that they're shutting down Argo. So given that L2 is probably going to have a longer ramp -- a longer life cycle, do you expect to see increased revenue from that market? Or is the big opportunity for your company still L4?

I don't -- from the beginning, my co-founder, when we started the company, wanted to do a full autonomous driving. I told them that I'm supporting his hobby, but it won't be a business in the next 10 or 15 years. And we designed the chipset that by the end of the day can support L4, but the main benefit and it fits from a cost perspective also to L2+, L3. I think that L4 as a mass market is something that will take 15 years maybe until it will mature. So it's not -- to put a RoboTaxi in one city within hardware that costs $200,000 per car, I don't think -- I don't see how it scale up to being a business, definitely not for us.

No. Got it. Makes sense. And then I think we had some audience questions in the back.

I wanted to find out of your early customers, are they -- what kind of sensors are they using? Are they planning to use your sensor in lieu of LIDAR or in addition? And then I'm assuming cameras are part of it? Or am I wrong there?

Yes. Camera is mandatory. So there's no way to build any kind of service without, you can say, 10 to -- 8 to 11 cameras. Everyone has their own configuration and a majority of those cameras are short range. Only part of them are long range that are more expensive, but camera is given. So vision only can take you far away. I would bet around 90% to 95% of the problem. The main issue is that on the last 5%, you might kill people. So the last 5% is the most problematic. And for that, I think radar and especially imaging radar is the solution. As much as redundancy as you want, you should add also LiDAR, LiDAR has a huge advantage in L4. If you want to do a full autonomous driving, and you need to do a real-time mapping, I think LiDAR is the way to go. But since L4 is far away, I'm not sure how much business there is in L4. For L2+ and L2++, I think that the problem can be solved as long as the driver is still in a way ready to take action. I'm not sure that LIDAR is needed. For L3, I don't see a car company taking the responsibility without putting a LiDAR just in any case that they will be sued, they can say that they have everything. But seriously, I think for the beginning, collecting data with the LiDAR and the radar, having the LiDAR as a ground proof for the radar and training, the algorithms is the way to go. And maybe over time for L3, not for L2++, there is -- you can consider whether a LIDAR is needed or not. I think most of the customers that are working with us, combining radar and LiDAR right now.

For that extra cost, what are they achieving with?

Sorry?

With the additional cost of radar and LiDAR, what are they achieving?

They're achieving -- with extra cost of radar and LIDAR, they can achieve a real point-to-point driving with the hands-free. And even, I would say, a bit of eyes off. So you cannot get to sleep, but you can relax and maybe in a situation that something went wrong, the car will tell you, please take control.

The LiDAR vendors are indicating they can do that without radar. So that's what I'm asking. With the addition of your radar, what are they achieving?

I haven't seen their service doing it without -- I haven't seen anyone doing it without radar. I haven't seen such a system working. And to be honest, I don't think that it will work because of the range, because of problems of the LiDAR by the end of the day is vulnerable to the same problems as the camera. If the sun is shining on the sensor, the range of the LiDAR get reduced dramatically. If there is rain, the LIDAR hasn't issues. So I don't think that the LiDAR can compensate on some of the problems of the camera, but not on all of them.

Do we have any other questions from the audience? Yes. So I mean, just kind of since your production is supposed to ramp at the end of Q4 or going to 2024, what's your view on light vehicle production across the geographies you have business in '24 or '25? And can you give any color on that?

So as I mentioned, '24, we will start -- we have preliminary orders of around $40 million from the Chinese market. And we believe that this is the first market that we will start shipping chips and see revenues. '25, the mix would start change to support our European and U.S. customers and in '26, it's going to be something like 50-50.

Got it. And then in terms of like the structural cost benefits and OEM can realize from using your system versus your competitors or pure LIDAR system, can you kind of like quantify that? And in terms of like what that would mean as they're ramping production?

So if you're trying to compare imaging radar, our imaging -- an imaging radar, this is based on our chips to the next imaging radar in the market that is based on TI chips and XP chips. Basically, our radar has 10x more radar channels and in radar channels is like pixels. So it has 10x better pixels and more or less on the same price as the next. If we're trying to compare it with the low-end radar. So if, for example, our radar is around -- radar based on our chips is around $150, today's front radar -- regular front radar that is doing pure adaptive cruise control is a radar of sub-$100, something like that. So the extra cost for the car company to adopt imaging radar is just $50 because the $100 radar is already there.

Got it. That's all I had in terms of questions. I don't know if the audience, is there anything else you want to ask? Well, thank you so much for your presentation.

Thank you.