Innoviz Technologies Ltd. Q4 FY2025 Earnings Call

Ladies and gentlemen, thank you for joining us for Innoviz' Fourth Quarter 2025 Earnings Call. Please note that this call is being recorded. I would now like to turn the call over to our first speaker, Ada Menaker, Head of Investor Relations. Please proceed.

Good morning. I would like to welcome you to Innoviz Technologies' Fourth Quarter and Full Year 2025 Earnings Conference Call. Joining us today are Omer Keilaf, Chief Executive Officer; and Eldar Cegla, Chief Financial Officer. I would like to remind everyone that this call is being recorded and will be available on the Investor Relations section of our website at ir.innoviz.tech. Before we begin, I would like to remind you that our discussion today will include forward-looking statements that are subject to risks and uncertainties relating to future events and the future financial performance of Innoviz. Actual results could differ materially from those anticipated in the forward-looking statements. Forward-looking statements made today speak only to our expectations as of today, and we undertake no obligation to publicly update or revise them. For a discussion of some important risk factors that could cause actual results to differ materially from any forward-looking statements, please see the Risk Factors section of our Form 20-F filed with the SEC on March 12, 2025. Omer, please go ahead.

Thank you, Ada, and good morning to everyone joining us today on the call. 2025 was a pivotal year for Innoviz in terms of customer engagements, production readiness, and end market expansion. We achieved the financial and operational goals that we set for ourselves at the start of the year, growing our market presence and strengthening our financial foundation. Since the last earnings call, we announced that the major commercial vehicle OEM with whom we have an agreement for series production of Level 4 trucks is Daimler Truck and its subsidiary, Torc Robotics. We also announced our next generation InnovizThree with a smaller form factor and lower power consumption for behind-the-windshield integration, the holy grail for automotive applications. Combined with an RGB camera, the InnovizThree is a compact sensor-fusion model that simplifies integration and deployment. The InnovizThree is also more affordable than the InnovizTwo, which increases its potential Total Addressable Market. You can actually see right now, we are being filmed using our InnovizThree LiDAR. I'll tell you more about it later on. Our InnovizSMART, which we introduced last summer for nonautomotive applications, is available for shipment, and we are seeing excellent traction with a variety of customers. We are very pleased with the inroads the InnovizSMART is making in the security market, and we recently announced that an InnovizSMART-based solution has been deployed across several sites in critical infrastructure as an off-the-shelf system comprising the LiDAR, cameras and analytics software. At CES, we demonstrated the InnovizSMARTer, integrated with NVIDIA Jetson Orin Nano to enable edge compute deployments in bandwidth constrained areas, simplified installation, and cloud applications. To meet the demand of our automotive and nonautomotive customers, we have continued to ramp capacity at Fabrinet and expect production this year to be 3x to 4x higher than last year. On the financial front, in 2025, we grew revenue to $55.1 million, more than double the level achieved last year. Our gross margin for the year was 23% versus approximately minus 5% in 2024. Operating expenses for the year were $80.6 million versus $100.8 million in 2024, a 20% decrease. As we kick off 2026, our NRE payments plan stands at approximately $111 million versus $80 million at the start of 2025. We've recognized approximately $45 million of revenues from these NRE agreements in 2025. We have $66 million remaining. We expect to recognize almost all of our existing NREs in 2026 and 2027. And we expect to sign additional NRE payment plans in 2026. As our current programs reach SOPs and we win new programs across automotive and nonautomotive, we expect to see significant growth in LiDAR revenue over the coming years. Longer term, as sales of LiDARs expand, we expect them to make a growing proportion of revenues versus NREs. We believe that sales of LiDARs into the nonautomotive physical AI applications will grow from approximately 1% to up to 10% of our annual revenues in 2026, accelerating further in the coming years. In all, we believe Innoviz is in a very strong position for 2026 and the years ahead as we expect to play a significant role in what could be one of the most important technological advances of the future. After transforming digital workflows through software and large language models, AI is moving into the physical world. It will power vehicles, robots, infrastructures, and machines that must perceive, reason, and act under real-world constraints and in real-time. This transition, often referred to as physical AI, represents one of the largest and longest duration technology opportunities of the coming decades. Physical AI must function in safety-critical environments, tolerate environmental variability, and scale at an infrastructure level. As a result, it requires a fundamentally different foundation with perception at its center. Perception powers world models, models grounded not in text or images, but in physics. And they are meant to emulate complex real-world systems. LiDAR is emerging as the most reliable method for digitizing the physical world into accurate real-time 3D representations, creating trusted world models that can drive machine decisions. Earlier this week, we published Part 1 of a white paper on the role Innoviz is playing, and will continue to play, in the rise of physical AI and how we bring world models to life. I invite you to read it on our website. Part 2 will be out soon. And in March, we will also host a webinar on the subject with Q&A. Please stay tuned for details. As the need for LiDAR as part of this new phase of physical AI becomes better understood, and as the technical requirements become more stringent, many players have been driven out of the market. We expect additional fallout in the future, with the market consolidating around just a few players. Automotive OEMs need behind-the-windshield solutions with lower power and smaller form factors. Nonautomotive applications, especially in the areas of industrial and security, demand enhanced reliability and safety. Innoviz stands ready to meet these challenging customer requirements. With the maturity and production readiness of our InnovizTwo and the smaller size, lower power consumption, and lower cost of InnovizThree, we believe we are well positioned to become the world's premier large-scale supplier of LiDAR solutions, enabling autonomous driving and the rise of physical AI. And now let's jump into the details. Starting with our trucking customer win. Back in September, we announced that we were selected for series production of Level 4 autonomous trucks by a major commercial vehicle OEM. In December, we were very pleased to be able to name the customer, which is Daimler Truck and its subsidiary, Torc Robotics. Under the terms of the engagement, Innoviz will provide multiple LiDARs per vehicle for the customer's L4 Class 8 Freightliner Cascadia platform. We have already begun shipping units to support Daimler's trucking fleet. This partnership positions Innoviz' technology as a critical component in Daimler's truck strategy to bring autonomous trucks to the market. Deployment is planned across highway and regional routes in North America to help fleet operators improve operational efficiency and enhance road safety. Having met the requirements of one of the world's largest commercial OEMs, we believe we are well positioned to gain additional wins in the trucking space. And as we roll out new technologies, including an upcoming ultra-long-range solution, we will continue to work with Daimler to explore additional opportunities based on our full portfolio of products. The agreement with Daimler that we just discussed underscores the traction we are seeing in Level 4 applications. As we said before, the case for Level 4 has become very clear to the automakers. They see the benefit of adding content to a vehicle, eliminating the driver, and deploying the vehicles in fleets. Waymo's success in pushing others forward, and at CES, about two-thirds of the customers and potential customers we met were focused on Level 4 applications. We are working towards Level 4 SOPs with Mobileye, Volkswagen, and Daimler Truck, and we believe we are the LiDAR company with the most significant Level 4 Western SOPs. I just recently toured the VW ID. Buzz production line in Germany, and it was truly impressive. This is the first automotive series production of a Level 4 robotaxi in the world. And it was very exciting to see our suite of nine LiDARs being installed on each vehicle as part of the automated process. We expect fleets of these vehicles in six cities in the U.S. and in Europe this year, targeted to ramp in the second half of the year. We are also progressing on our Level 3 SOPs with the Mobileye Chauffeur and programs such as Audi expected in 2027. In terms of new programs, in addition to the strong interest in Level 4, we are also very excited about the increase in Level 3 activity as well as our continued work with NVIDIA. Level 3 is now viewed as a KPI for upcoming car designs, and there are multiple RFQs for programs aiming for 2028 and beyond. Many of these programs are targeting behind-the-windshield deployments. Several OEMs have recently discussed their Level 3 efforts, specifically mentioning LiDAR as a key component. With significant progress in LiDAR designs and the availability of mature technologies, automakers are exploring their options, and we are poised to compete and win on new RFQs with our solutions. To support our customers' efforts, we've introduced the InnovizThree. Over the last 10 years, the LiDAR space went through two phases. And we are now entering a third phase. In the first phase, there were many players, and the devices were largely proof of concepts. In the second phase, the automotive OEMs needed a mature, durable working product, and the number of players declined. As far as we are aware, there are only two companies that were able to launch a Level 3 product during the second phase, and Innoviz was one of them. The time, development, and effort that Innoviz invested in the first two phases are the foundation of the InnovizThree, designed to enable us to meet the elevated requirements of this next phase. Here, the holy grail of automotive LiDARs is behind-the-windshield installation that doesn't compromise vehicle design or in-cabin environment. The InnovizThree has been designed to meet these challenges with a smaller form factor and lower power consumption. The InnovizThree also offers a lower price point. While the InnovizTwo cost approximately 70% less than the InnovizOne, the InnovizThree offers an incremental 35% cost reduction. This could make it a compelling solution for L2+ applications as well. At CES, we even showcased the InnovizThree combined with the camera. This solution simplifies OEM sensor integration and sensor fusion, streamlining packaging and enabling faster deployment. The LiDAR plus camera can also be used in applications such as drones, microrobotics, and humanoids, further enabling physical AI. And by the way, as I said at the first of the call, we are using InnovizThree to make this call. I think you'll agree with me that this is unlike any other LiDAR image out there. By adding color capabilities directly into our LiDAR, we are giving OEMs a cleaner, more efficient, lower-cost path to multi-sensor perception without compromising vehicle design. While the automotive space remains our primary area of focus, let me touch on our progress with the InnovizSMART, which is now available to order and the newly announced InnovizSMARTer. The InnovizSMART is based on the InnovizTwo platform and optimized for nonautomotive applications such as security, smart city, robotics, and others. One example is an InnovizSMART-based perimeter security solution which has already been deployed in several locations. This is an off-the-shelf system for protecting critical infrastructure and municipal areas. It combines our LiDARs with our partner's PTZ cameras and analytics software. Unlike radar and camera systems, the solution can detect movement behind obstacles such as trees and fences, maintains performance in harsh conditions, and delivers reliable detection of slow-moving or camouflaged targets. The InnovizSMARTer integrates InnovizSMART LiDAR with an NVIDIA Jetson Orin processor to deliver a one-box edge solution for real-time 3D perception, enabling wireless deployment of LiDARs in bandwidth constrained environments. At CES, we showcased the InnovizSMARTer by live streaming a point cloud of the Las Vegas Strip, which was compressed at the edge. The video you've been seeing in the recorded output, as you can see, it flawlessly shows the landmarks, people walking, moving cars, and even the textures of buildings. These are all great examples of what LiDAR can do in terms of helping build a world model for physical AI, and we are seeing a lot of interest in our smart products for a variety of end markets. Now let's talk about our outlook for 2026. Driven by ongoing NRE payments and the ramp of LiDAR shipments, we expect to grow revenues by approximately 27% to $67 million to $73 million. In 2026, we expect up to 10% of our revenues to come from nonautomotive physical AI applications, up from 1%. We expect new NRE payments plans of $20 million to $30 million in addition to our existing plans. We expect to add two to three new programs this year. And now I will turn it over to Eldar to discuss our financials.

Thank you, Omer, and good morning, everybody. Innoviz experienced significant growth and operational momentum in 2025. Revenues were at $55.1 million, a record year for Innoviz. We ended the year with approximately $72.1 million in cash, cash equivalents, short-term deposits and marketable securities on our balance sheet. And we have no long-term debt. Cash used in operation and capital expenditure in the year was approximately $49.3 million. For the quarter, cash used in operation and capital expenditure was in the single digits at $7.3 million, which included proceeds from sales of machinery. This was the second quarter of single-digit burn in the year, demonstrating our commitment to lowering cash burn over time. We believe that our strong balance sheet and continued focus on operational excellence will provide us with the runway to reach customer SOPs into 2027. Now turning to the income statement, our 2025 revenues of $55.1 million more than doubled year-over-year, supported by NREs as well as sales of LiDAR units. Gross margins in the year were approximately 23%, an important step towards profitability. Going forward, we expect that we will continue to see quarter-to-quarter variability in margins due to the revenue mix and customer timing. Our operating expenses in 2025 were $80.6 million, a decrease of approximately 20% from $100.8 million in 2024. This year's operating expenses included $10.7 million of share-based compensation compared to $17 million in 2024. Research and development expenses for 2025 were $56.5 million, a decrease from $73.8 million in 2024. The decrease is primarily related to the allocation of costs associated with sales of NRE and to operational realignment in Q1 of 2025. The year's R&D expenses included $6.2 million of share-based compensation compared to $11.2 million in 2024. 2025 was truly a pivotal year for Innoviz, and I look forward to what is ahead as we ramp up the new product and secure additional program wins across the automotive and nonautomotive space. With that, I'll turn the call back to Omer for his closing remarks.

Thank you, Eldar. Before I wrap up the call and open for Q&A, I want to recap some of our recent developments. In 2025, we reported record revenues of $55.1 million, more than twice the previous year. We improved our gross margin and reduced our cash burn. We were selected to supply LiDARs to Daimler Truck for their autonomous trucking platform and made significant progress on our L3 and L4 programs. We continue to see a lot of interest in L4 and see a step-up in engagements in Level 3. We introduced the InnovizThree, which we believe meets the holy grail of automotive requirements for behind-the-windshield installation. We are building a strong presence in smart applications with our InnovizSMART and InnovizSMARTer. 2025 was a truly pivotal year for us, and we look forward to continued momentum in 2026. The transition to physical AI has begun, and perception is a foundational layer in bringing world models to life. Physical AI cannot tolerate ambiguity, and LiDAR is critical to the establishment of ground truth. LiDAR offers accurate, reliable data, and critically, it meets privacy requirements. We believe Innoviz is uniquely positioned at this inflection point. I look forward to telling you more about this later in March when I will host a webinar to discuss our technology and its role in the implementation of physical AI. Please stay tuned for details. And with that, operator, let's begin the Q&A.

Mark Delaney with Goldman Sachs.

Yes. I was hoping you could give more color on the guidance for 2 to 3 new wins in 2026? And in particular, I'm hoping for color on where you think those wins could come from and how close you think Innoviz is on converting on those new opportunities.

Yes, sure. So Mark, there are several programs that we are currently active on that are on Level 4, actually, that we expect to converge. On top of that, there are a few Level 3. So those are the ones that we are, I would say, more tangible, and we see good cadence and progress, and generally, we're in a good position for. Other than that, there are Level 3 programs which we are competing on. And they are split between InnovizTwo and InnovizThree. The InnovizTwo is based on programs that are trying to launch earlier. They kind of give advantage to an already automotive-grade product that is going to production. And there, obviously, the InnovizTwo is a very mature product that we can offer. For those that need a product that is designed for behind-the-windshield, this is where we are offering the InnovizThree. And there, I believe that we are also in a good position, but those decisions probably will take a bit more time, maybe towards the second half of the year. But overall, those are the activities that we were pointing at.

And could you also speak specifically to where Innoviz stands at converting on the SoDW with a top 5 auto OEM and the potential for that to convert into a series production award?

So we completed the SoDW, and we're in discussions with the OEM about the next steps. It's unclear yet what it will convert into and when.

Okay. And then lastly, guidance for revenue for 2026, I believe, assumes that 10% of the revenue comes from the non-auto market. Could you talk about how well covered that is in terms of bookings? In other words, have you already secured the design wins and orders to support that revenue outlook outside of the auto market this year? Or is there still work to do to achieve that?

Part of it is booked through our ongoing effort in the security market. This is where we're seeing quite high demand and a good fit. We went through several RFQs competing on different opportunities and came out with the upper hand due to the very impressive performance that our LiDAR offers. While many LiDARs that are active in the nonautomotive market are relatively low resolution and range, and therefore, were not really suitable for that market, our LiDAR, due to its unique high resolution and long range, we were able to unlock that market. And those are very premium markets in terms of its safety level applications. So everywhere we go, we see a very good appetite for what we're showing. We're also working on an ultra-long-range LiDAR that I hope to be able to share a bit more soon. That will take us even further in that market. And of course, we are still also in discussion with a wide range of applications that are already being served by different LiDARs but see the benefits of using a LiDAR which is automotive-grade with high resolution and resilience to dirt, etc. So it's growing. It's growing fast. It's also, I would say, it's fun in a way, after working only with automotive customers, it's fresh air sometimes. But generally, it brings a good vibe to the team seeing the variety of different opportunities and the excitement we’re seeing from our customers.

Jash Patwa with JPMorgan.

Congratulations on all the progress this quarter. I was having a hard time distinguishing between the LiDAR and camera feeds, so appreciate that intuitive demo. I wanted to start with a high-level question on the technology landscape. With AI disintermediation risk front and center for many investors, could you share your perspective on how LiDAR technology might be insulated from or even benefit from AI advancements? Relatedly, do you see any risk that AI-driven improvements in alternative sensing solutions could ultimately limit the longer-term Total Addressable Market for automotive LiDAR? And I have a follow-up.

Sure. More than happy to talk about this. Physical AI was practiced already in the last 10 years in terms of autonomous driving using LiDARs to practice and develop AI to allow the car to drive autonomously. It is actually only in the last couple of years where AI made significant progress and became available to many other sectors. What we're seeing is that while for a very long time, AI focused on digital content, whether it's text or images, where the generative AI was only trying to predict the next pixel or the next word, you see a desire to go into the physical world where you can use AI to understand the world better, analyze it, and also possibly even predict it by being able to train those world models. The missing factor or the missing link, as I try to point out in my white paper, relates to the fact that all of those world models are based on simulation data that are also generated by AI. That creates a big problem when your AI is trained by data that is also provided by AI; you are creating a very big error that is inflated, and you create a bias in your models that are targeted to create models that try to predict how the real world acts. What we're trying to point out is that by using a high-resolution LiDAR, you can actually connect to those world models and those digital twins that companies like NVIDIA are talking about with the SOC, where you connect those digital twins to life. The LiDAR will bring life to those models and enable the development of AI on the real world. And this is where we see today AI being practiced in different industries, whether it's industrial, security, or maritime. Basically, everywhere, you'll see - in the future, you will have LiDARs that are operated, whether it’s outside our houses, in cars, or by robots that will use LiDAR within our houses. So that data will allow AI to better understand how the world really acts. There’s obviously a lot of sensitivity when it comes to collecting all of that data. But I see LiDAR as if we are connecting the world model to the Internet, in a way that we are feeding it with real-time accurate information. And we see that more and more people are aware of AI capabilities and how it can be deployed in many applications, not only autonomous driving. And that's what we are currently working with, with different partners. We will elaborate more on our AMA, Ask Me Anything session.

Awesome. I appreciate all the color there. Just as a quick follow-up. Congratulations on the Daimler Truck announcement. I think you already touched on this in your prepared remarks, but I'm curious if you could peel the onion further on why Innoviz was selected as the preferred short-range solution but not for the long-range variants. Could you provide any feedback from the negotiation process? And were there specific technology considerations or limitations that led the OEM to pursue a dual sourcing strategy?

Sure. With full transparency, at the time that Daimler Truck was evaluating options for a long-range solution, we were not even in the process, for whatever reasons. Only when the short-range opportunity came up did they become aware of our solution. I can share with you that the relationship is very strong, and we are currently discussing further expansions where Innoviz technologies can benefit Daimler and Torc on their mission. As we were mentioning earlier, we are working on an ultra-long range LiDAR that would actually be very efficient and valuable, not only for the security market, but also for the truck market. So this is our first engagement, and I believe it will grow.

Colin Rusch with Oppenheimer.

As you get deeper into the physical AI space, can you talk a little bit about any sort of need to invest in incremental sensor fusion capabilities as well as functional safety, and what the demand for functional safety might look like for you guys, and how long you would expect it would take for you to bring that to market?

Actually, the requirements we're seeing from customers regarding functional safety are well aligned with what we have already achieved for the automotive market. What gets people very excited is our resilience of the sensor. The InnovizTwo sensor, for example, is very resilient to weather conditions and dirt. Those are the situations where an autonomous truck or car or taxi, you wouldn't want it to be in a blind spot whenever something happens, such as dirt on the window. That's a very unique proposition that Innoviz is providing. And that's part of those that want to install LiDARs on different infrastructures, they want to see available. Other than that, as I was referring to at the beginning of the talk about the InnovizSMARTer, that's our next-generation InnovizSMART where we embed processing power to the edge, which helps in equipping the LiDAR in different end points that allow us to transmit the data with compression and connection to the cloud. Once you are able to connect the LiDAR to the cloud from different points and create a world model, it's quite easy to connect it to other platforms such as the one that NVIDIA is developing in order to train your models and build on top of them. We understand the requirements for that flow. We're developing the tools that are needed. It does not require changes to the LiDAR design itself. And as you saw, we are also offering a LiDAR and camera fusion out of the box. So basically, we have all of the design already set up for such an infrastructure.

Great. And then my follow-up is really around customer engagement. Certainly, there was an enormous amount of activity at CES. And I'm sure there are a lot of folks sampling. I just want to get a sense of the scope and scale of the customer engagement for you guys right now. Like working with the distributor is helpful, but I'm sure you've got a wide range of folks that you're engaged with directly. I just want to get a sense of how that's grown over the last year, and really in the last couple of quarters from a kind of numbers perspective. Any color you can provide us in terms of the sampling activity would be helpful.

Sure. So obviously, the answer is split between automotive and nonautomotive. In the automotive market, we are well connected with all of the customers and engaged, I would say, periodically on a weekly basis with all of the customers, most if not all. When it comes to the nonautomotive, we see a huge step up in our leads and engagements. Just to give you a reference, two weeks ago, we had a big conference at Innoviz, where we invited around 80 security consultants from Israel when it comes to the airports, harbors, train stations, etc. We did some sort of exposure visit where we showed them the technology. I think we probably left that event with tens of opportunities. The following week, we visited a defense conference and left the conference with, I think, a few tens of leads. So this is a market where we are growing our awareness, I might say, where we are getting in touch with customers that still require some exposure to the capabilities of the technology. Right now, we feel that in the defense market and security market, we are already having quite nice exposure as we are getting leads already, I would say, passively reaching out to us. We are growing our business development in the States, while we already have a nice position in Europe and Asia. So we are growing our footprint in this domain. I think that’s why we are expecting 10x growth between last year and this year. I actually think it's a modest assumption for what I believe our opportunity within this market, having that we have a good product in production, automotive grade, that is actually solving problems that others don't.

Great. Just wanted to dig in further into the comments on increased L3 activity on Slide 9. And specifically, we've seen a few announcements where the level 3 hardware is going to be equipped as standard fit across all vehicles. I'm curious if in your discussions and your pipeline, are you seeing a similar trend there? Or is generally, Level 3 still going to be sort of at initially low attach rate and then growing from there?

No. Actually, what we're seeing - and this is also something that I think will be deeper in our second part of the white paper - is that the LiDAR space over the last 10 years went through two phases, where the first phase was mostly prototyping and customers were becoming educated about what they need. In the second phase, you've seen several programs already going into production, but only actually two of them. I think the third phase that we see the automotive space entering is where the programs are targeting higher volumes. They see that the pricing of LiDARs has come down where it can enable it. The installation behind-the-windshield removes one of the last frictions as far as I see it. Because I saw a lot of friction in the past where LiDARs were either in the grill and not optimized in terms of height or on the roof in terms of vehicle design. There was back and forth between the design team and the engineering team of the car company, where they were really struggling on how to make it work across all of their models. Being able to deploy it behind-the-windshield removes quite a big friction in that regard. Bringing a LiDAR to behind-the-windshield requires a very significant size reduction and power reduction because the flux of sun that you need to absorb behind the windshield is quite high. The temperatures that you need to be operating at are higher. And obviously, the performance needs are more challenging due to that innovation that is slightly expected from the window. So that’s where I see the complexity, or I would say, the moving target of the LiDAR requirements keeps moving. Over the last eight years, we've seen 200 LiDAR companies don't meet the requirements of the first phase. The last 50 didn't meet the requirements of the second phase. And now we see the third phase where the requirements are elevated again to be behind-the-windshield. To be honest, I'm not familiar with any technology other than the one that we're using that is a better fit for that, with some other technologies not being valued for that. So I'm excited about behind-the-windshield, because I know Innoviz is able to serve that market. I believe we'll be able to position ourselves as the leader here.

Terrific. And as a follow-up, and I apologize if I missed it earlier in the call, on Slide 5, with the growth in the non-auto LiDAR and physical AI, can you just mention how we should think about the impact to gross margins and ASPs for the company over the next few years?

Sure. I mean, obviously, those markets, just to give you some perspective, when it comes to the defense market or the security market, the technologies are priced at around $10,000 per sensor. Obviously, the sensor that we are offering is significantly better than the ones that are used today due to our ability to see beyond trees, beyond fences, small objects, and obviously offer many more features that were not even possible to ask for. So the Average Selling Prices in these markets are higher. Since it's related to safety, the need is clear, especially in the landscape of the world that we live in. This is why we focused on those markets early on because we also saw that that's a market that other LiDARs are challenged with because they don't meet the range, they don't meet the resolution, and that's where we can come in without too much resistance. And of course, we can also penetrate other markets that are served today by LiDARs, such as ITS, airports, and trains. Today, I had a discussion with someone from the team who was talking with me about discussions with the electricity company and with the water company. It feels like a monopoly, like we are starting to cover all of the infrastructure around us. It's a world of opportunities when it comes to physical AI. What we're seeing is that the market today, in contrast to automotive, where we are not displacing the radar or the camera, we are coming in on top, in this market, we see the analysis talking about a $10 billion market size in a few years. When we talk about physical AI, we are looking at the Total Addressable Market of cameras and radars because we believe we can replace them. Those TAMs are quite significant. Meanwhile, we are providing a better value than them.

Thank you for a really great update, Omer and Eldar. Maybe one quickly for Eldar. And maybe you said this on the call: What do we think about OpEx kind of the run rate? Are we sort of going to be in the same level as we move forward? Or should we expect some meaningful changes for any reason, up or down, I guess, in '26?

Until now, the company has shown its ability to be very, I would say, efficient and modest in the way it conducts itself. Going forward, I think we will continue with this method of running the company. So I'm not expecting any dramatic change in that respect.

Okay. And then more broadly, maybe Omer can talk about this too. Does the drivetrain matter? If you look at automotive and trucking, it feels like a lot of the innovation for L3 and L4 robotaxis have been on EV drivetrains. I'm not sure if that matters. Maybe that's just newer designs and they're more willing to integrate new technologies. But are the opportunity sets different in sort of ICE vehicles and sort of ICE production plans?

No, I would say the following: Several years ago, the OEMs when they had to pick or design their future vehicle, they don't develop it on a yearly basis. They do it on a cadence of 4 to 5 years when they set a certain design, start working on it, and eventually cut out of its different brands, but eventually, the platform itself serves multiple vehicles. Now, several years ago, many of them picked Level 3 and EV as kind of like what they need to do on their next vehicle. In a way, I believe that the LiDAR market or the Level 3 market in the Western market was suffering due to the challenges that came with the EV transition. Some of those platforms were delayed or even canceled. I was surprised at the beginning of the year when we were meeting several OEMs that came up, and we were talking about RFQs for Level 3. It became clear that their strategy related to EV or ICE is behind them in terms of they've made their decisions, and that turmoil is settled now. From that sense, Level 3 is back on the table because they want to launch it because Level 3 was always on their map, but because of issues they had with the platform of the EV, I think the LiDAR market was slightly delayed. So on the technology part, there is no correlation between EV or ICE to a LiDAR. You can operate autonomous driving on any type of vehicle. You just need the platform to be developed.

Got it. Okay. That's helpful. And so it sounds like '26 is setting up great for maybe a meaningful jump in terms of commercial revenues versus its percentage contribution in '25. I think in your initial comments, you said that production would double or triple, Omer? I missed that. I wanted to get...

The production is going to be up 3x to 4x. Our production site at Fabrinet is ramping up. We are preparing towards the SOP of Volkswagen and Mobileye, where we are expecting fleets of vehicles in 6 cities in the U.S. and Europe. On top of that, there are other customers like Mobileye that are going to follow that in terms of SOP. So definitely, our production ramp-up is going to increase our sales of LiDARs, and that would also add to our sales in the non-auto market.

Yes. Well, that's a very exciting outlook, I think, and a great trend point. Just one last point on the NREs that you have, the $111 million in total. Could you just give us a count on how many customers made up that group of NRE contributors, if it was double digits or single digits?

We're currently supporting multiple programs. I'm not sure what you mean about customers, but we are currently in parallel supporting 4 or 5 programs, whether it's different programs within Volkswagen, different programs with Mobileye, and Daimler Truck. This is the number of programs that we are currently committed to and working towards an SOP.

Okay. And then last question. Of the new NRE opportunities, would any of those be outside of automotive or trucking? Do you think they could fall into the physical AI category?

I think our assumption right now is that this will still come from the automotive market. That's our assumption right now.

Thank you. It's a very exciting outlook for '26. So thanks for the update.

There are no further questions. I'm handing the call over to Omer for closing remarks.

Thank you very much for taking the time to listen to our end-of-year earnings. We are experiencing a very exciting time where we're seeing LiDARs being used on so many different opportunities. As I was saying during the Q&A, the vibe within the team is very high due to the excitement that we see in our customers' eyes. So thank you very much, and see you soon.

Thank you very much for your participation. This concludes our call. The session will now be closed.