Conference Transcript - ARBE 2026-08-11

Kobi Marenko, CEO

Good afternoon, everyone.

Speaker 1

I'm Mapp Shuli of Canaccord Genuity's sustainability team, joined here by Kobe Maranko, president and co-founder of Arbe Robotics. And Kobe has a presentation here for us, and then we'll finish it off with some Q&A. So, Kobe.

Kobi Marenko, CEO

Thank you. Thank you, everybody, for coming. So, I'm Kobe Maranko. As you said, I'm the co-founder and president of Arbe Robotics. Arbe is a radar company. We are providing an ultra-high-resolution radar for a lot of activities. So basically, when we're looking today in the market, the physical AI, the robots, the robot taxis, the drones, they are all coming into the market. They all need reliable, really world intelligence, a sensor that can see the environment in any weather, any lighting condition, and gives an output that an AI system, an AI model can really use it in order to do autonomous movement. And this is what Arbe is all about. We started the company with a focus of automotive. We were actually chasing the autonomous driving And the fact that we were focusing on automotive gives us a lot of advantages This is the real market where the quality that is needed is very high from one hand And the price that the car manufacturers are willing to pay is very low So we are coming today to the market with a product that has the best performance, the lowest power consumption, and the lowest price. And this year, this is the first year that our commercializing is expanding. so basically when there is a perception we always think of a camera camera of course is the best perception it's almost like our eyes it sees the environment but the the camera has limitation the main limitation of the camera is bad weather bad lighting condition night fog rain snow So all of those basically blocking the camera from seeing the environment. But there is more than that. There is the distance. So a camera is limited by the distance. And when we are talking about physical AI and we are trying to move fast, whether it's a drone or a robot or a robot taxi, the speed is critical, which means you need to see the obstacle in a very long range. and this is where the radar fits in. Another big advantage of the radar is the fact that the radar can actually detect the velocity of the objects, not just the depth, not just the X and Y of the object, but also the velocity. So if you take the reliability of the radar that works in a long range, low latency, see and can see the environment in any weather and any lighting condition. And you add to this a perception layer, then you get an output that is ready for the large models to fuse it with a camera and coming with a full self-driving or full self-moving robots. So, basically, our base semiconductor company, we are producing chips for radar. Our chipset basically has two parts, the RF part, the transmitter and the receiver, and our core IP is our processor, a dedicated processor that has our IP on it and allowing us to process 10 times more than any other processor in the market in the amount of channels and in the amount of points and the output of this radar. For the automotive market, we are selling this chipset to the tier ones, the companies that sell to the car manufacturers. Magna is our customer, our main tier one in the Western world. Hi-Rain is our tier one in China. For other applications which are not automotive, we are producing the entire radar system, the entire box. Like you see, it's a box in this size, more or less. And with this box, we are able to see the environment up to 500 meters, to see a very small drone, very small object on the road. in any weather and any lighting condition. On top of it, together with NVIDIA, we are providing an AI-based perception for the radar. So we are able to see the environment and to generate this point cloud. And with the AI stack, we are able to analyze the output of the radar and generate what is called free space mapping, this green area which tells the car or the robot or the drone what is drivable and what is not including of course classifying the objects and tracking them and making sure that the drive will happen smoothly and safely so basically we have three verticals first vertical is automotive we have we already have a robot taxi in that driving with our radar in the States, and a delivery van fully autonomous in China that we won that will start driving by the end of this year. We already supplied the chips to our tier one. On the second side of the story is the defense industry and the homeland security. Our radar can detect drones, even very small ones, in more than 500 meters and by that enable the defense system protect on a perimeter or on a vehicle in the battlefield or even a platoon of soldiers. We are able to detect the drones around them and make sure that they will be able to take a shelter or to operate an interceptor for that. With our radar, we are enabling also vehicles that are used in the battlefield. We have a contract with, we won a contract with the American Army for autonomous supply chain trucks. It's a big truck driving off-road where there is a lot of dust, a lot of rain sometimes, and they are able to drive themselves fully autonomously with our radar and, of course, with a camera. Perimeter security is another application that our radar can handle and autonomous movement or autonomous flying of drones with our radar on the front of the drone. In between them, there is the civilian market, so we have traffic management customer, delivery robots, which is something between a car and a robot, and even marine. We won a contract with a big yacht manufacturer that enabled the yacht to be fully autonomous, drive back to the port, and making sure they won't hurt swimmers on the way in. The fact that we are coming from automotive engineering is what gives us now the advantage in other industries. It's the performance. The automotive industry needs the best resolution. And with this best resolution that we can separate between a child and a car and a truck and something that is on the road, we are also able to see the small drones flying even in an urban environment, very low, and to separate them from the ground and making sure that we will be able to detect them. The architecture of a radar is a solid state. There's no moving parts. It's a box with chips inside it. It's very reliable, and it works in any weather and any lighting condition. And because of the fact that we built for scale, because automotive is a high scale we are able to supply any other industries easily looking on the competition basically today we are competing with in two basic industries the first is automotive and the automotive we have today the best price performance and power Our next competitor is Mobili. And on the drones detection, our main competition is Ecodine with their EcoShield product that has a low resolution and it's much more expensive than our solution. And also, our radar is much smaller, which enables it to be on mobile or on vehicles, as opposed to Ecodime that is mainly protecting a perimeter, a stationary place. So there is a broad shift to physically high, and the radar is a huge opportunity for that. We believe that by end of the decade, there would be around $27 billion in radar systems. Part of it is automotive, part of it is defense, part of it is the physically high. For all of those markets, we have a solution and a good solution, A solution that can fit into the existing ecosystem, connect to the AI stack, and make a difference the minute that we are in. So basically, by the end of last year, we moved from development to production. And early this year, we started generating revenues. And since then, we are increasing our revenues in more than 50% a quarter by quarter. And we believe that this is going to continue to the rest of the year. Next year, we're going to see a broader commercial deployment, especially in China. now we have some project that going to full scale of production in automotive and also in defense projects that or customers that we want this year and we started shipping them hundreds of units will become a thousand and ten thousands of units next year which will drive us to break even in and 28 and revenue growth in 29. By 2030, we believe that we will start seeing also a larger amount of revenues coming from automotive. So today, we have clients around automotive. In China, we have a full autonomous delivery van. We have a robotaxi project, robot taxi company here in the States, and we won a European truck manufacturer for autonomous trucks. In the defense and homeland security, together with Volterra, we won the contract of the American Army, and we have a defense and homeland security system integrator in three projects. One of them, the leading one, is drone detection. On the civilian and static, we have a project in China for traffic control. We have a customer for, as I mentioned, for autonomous yachts and some other smaller projects as well. Our production is already there. Our main, our fab is global fundaries, and we have with them a long-term strategic manufacturing agreement that providing us the ability to support our customers. Also in times where there is shortage in chips, we can really make sure that our customers will get it. Our guidance for this year is around $4 to $6 million in revenues, and just a EBITDA between $28 to $30 million negative. The cash burn is reducing dramatically toward the end of the year. Next quarter, Q3, we will burn less than $7 million, and by Q4, less than $6 million. And we are aiming to reduce it even more next year. We had $42 million in cash as of end of this quarter, which we believe will enable us to reach break-even in 28. So a few words about ourselves. We have traded on the Nasdaq for the last five years, 120 people, most of them in our R&D center in Israel. We have sales and support in Germany, China, and U.S. and as I mentioned, $42 million in cash. So physical AI needs reliable real-world perception and we are the sensor for that that can work in any weather, any lighting conditions. Now for the questions. I'm good with that.

Speaker 1

Maybe to start just on the technology front, Your radar has significantly more channels than kind of peers in the space. I guess how does that help support safety in a lot of these autonomous driving applications? And why is your product differentiated relative to theirs?

Kobi Marenko, CEO

Yeah, so in radar, channels means pixels, and more pixels means a better picture, and better resolution, and better ability to fuse it with the camera. So the current radars that exist, especially in the automotive space, basically they are focusing mainly on detecting the velocity or one or two points on the object. And if you have two points on pedestrian, you cannot use it with the camera because there's no AI that will be able to say, yes, those two points is a pedestrian. And the fact that we have so many points in our dense point cloud allowing us actually to show the shape of the human being and by that enable the AI stack that companies like Wave or NVIDIA are generating to fuse it with a camera and to take decisions in real time. Our radar today, we have an application that can actually see even a gun that is hidden in the bag or something like that. So it's a radar that can actually see the details, and this is because of our channels and our ability to process these channels. So this amount of channels is not coming without a cost. So the cost of it is a lot of data that our radar is generating, and the way to process it is with our processor.

Speaker 1

Great. And I guess maybe just next, I think what we've seen a lot recently is OEM pullbacks just with respect to pursuit of L3, L4 autonomous driving applications, and I think just kind of broad disarray within the market. How has that kind of shift impacted your timing with a lot of these OEM developments?

Kobi Marenko, CEO

So, I think that autonomous driving in the end of last decade, everybody thought that it's behind the corner. We always thought it will take longer than expected. But with the help of the COVID, this behind the corner was pushed even more. I never thought that the OEMs will be able to solve this problem there's no way that this problem that is so complicated will be solved by people that are working for GM so if someone finished university and is a good AI researcher and is getting an offer to go to work for the Silicon Valley or to go to Detroit to work for GM, there's no question what offer it will take. So I always counted on companies like Wave or NVIDIA to solve this problem. But actually, I think before the LLMs and the shift, the paradigm shift in AI, there was no way to really solve this problem so four or five years ago the car manufacturers as well as even way more and and those companies were were building something that was a bit of a rule based system and because of the fact that the road is so unpredictable and there can be hundreds if not thousands of corner cases every minute this is not a way to solve the problem and just when the processors became strong enough and the LLM and the VLM whatever the implementation of the LLM to the virtual to the visual world happened this is what basically allowing the problem to But I think we are still two or three years from solving the problem, real solving the problem, without remote driver, without safety driver, in a way that it can be scalable and it won't take months to map the city before you can start driving. So, I believe that the robot taxis will start scaling up around 28, 29. For passengers' vehicle, it probably will take longer, if at all.

Speaker 1

Yeah, and I guess that kind of brings me to my follow-up here. You mentioned your partnership with NVIDIA, and they've kind of developed a more just baseline application that OEMs can then develop their own software on. And with that technology, have you seen a re-acceleration in OEM demand for these applications?

Kobi Marenko, CEO

I think that the OEMs will be the last to adopt it. At first, it will be robot taxis and the companies that will provide car as a service to the customers. OEMs will be the last to adopt this technology. because they won't take their responsibility on those accidents. They are very conservative, and I don't see them taking this kind of responsibility.

Speaker 1

And maybe just on the robo-taxi front, on your most recent earnings call, you mentioned that you have some robo-taxis currently testing on the roads now. If you could just provide a little more color around what the current volumes look like, perhaps who your partners are with that, what the geographies are of where those deployments are going, and maybe just what you ultimately kind of see that program leading into.

Kobi Marenko, CEO

So we cannot say the name of the customer, but the driving are here in the States, and there is around 100 cars driving with our radars, collecting data, and helping them improve their algorithms and their software stack.

Speaker 1

Great, and then maybe just internationally, you have your partnership with HiRain in China, and you cited in your presentation that you have a program with a state-owned OEM kind of on the ramp right now, and that's kind of slated to go into production 27 to 28. Could you perhaps just talk about what that will look like as it nears and what are the key kind of milestones that you need to achieve before you reach that scale?

Kobi Marenko, CEO

So, first of all, we already start shipping chips for this, and by end of this quarter, higher-end will start shipping the radar to those cars. It will start as a delivery van, a small delivery van that will drive itself fully autonomously, and then it will get to much broader applications like high-end robotaxes in China.

Speaker 1

We'll see if the audience has any questions. If not, I'm happy to continue.

Kobi Marenko, CEO

So basically we are selling chipset for automotive in a high volume in around 80 bucks with more than 50% gross margin, and we are selling radar for other applications in around $1,000 per unit in much lower volumes. Is that a good number for a delivery robot?

Speaker 1

Yeah. Yes. Maybe just one point on that. so you mentioned similarly in your presentation that you've you've kind of shifted your model um you know previously you were only providing chip sets and now you're for non-automotive applications you're building the radars in-house could you just help us frame you know what the catalyst was for that shift was it more um you know just sheer volume at the tier ones or you know economies of scale on your end well we came to the conclusion that the automotive will take longer than expected and in order to survive we must produce the radar by ourselves the tier ones are not willing to take those kinds of volumes.

Kobi Marenko, CEO

It's too low for them, it's thousands, 10 thousands of units that they know to work in millions. And there's no millions of units in this industry, and this is why we had no choice but building our own production line, and we have it. And by the end of Q2, we started shipping full radars from our production line.

Speaker 1

Great, and then maybe just as a last question, And, you know, kind of looking ahead, looking ahead beyond this year, you've cited a lot of your production programs are going to reach scale. You know, how do you foresee kind of the ramp for the company going over the next few years?

Kobi Marenko, CEO

So I think that we're going to aim for reaching the $10 million next year and growing from that to $25 million to $30 million by the next year. It's based on the current level of customers and our ability to add defense clients. Of course, this is also complicated, but we are aiming for one or two defense clients by the end of this year, which will drive our revenues for the second half of next year and mainly in 2018.

Speaker 1

Great. Thank you so much, Kobi. Appreciate your time.