Kodiak AI, Inc. Common Stock (KDK) Q4 2025 Earnings Report, Transcript and Summary

Thank you for standing by, and welcome to Kodiak's Fourth Quarter 2025 Earnings Conference Call. [Operator Instructions] I would now like to hand the call over to Dan Goff, VP of External Affairs with the Safe Harbor. Please go ahead.

Thank you, and welcome to Kodiak's Fourth Quarter 2025 Earnings Call. On the call today are Don Burnette, Founder and Chief Executive Officer of Kodiak; and Surajit Datta, Chief Financial Officer of Kodiak. Our press release and an earnings presentation were issued earlier today and are posted on the Investor Relations section of our website. This call is being broadcast live via a webcast, and a replay will be available on our website after the call. Before we begin, I would like to remind you that during today's call, Kodiak will be making forward-looking statements within the meaning of the federal securities laws about financial performance and future events, including our guidance for fiscal first quarter and full fiscal year 2026 as well as our long-term goals. Actual events or results could differ materially. Please refer to our SEC filings, including our most recent Form 10-Q and the Form 8-K filed with today's press release for important risks and other factors that may cause our actual results to differ from those in our forward-looking statements. Additional information will also be set forth in our annual report on Form 10-K for the year ended December 31, 2025. We disclaim any obligation, except as required by law, to update or revise any financial or operational guidance and long-term goals or our other forward-looking statements, whether because of new information, future events or otherwise. Any forward-looking statements made on this call speak only as of the date of this call. Further, in addition to discussing results that are calculated in accordance with generally accepted accounting principles, we also refer to certain non-GAAP financial measures. For more detailed information on our non-GAAP financial disclosures, including reconciliations to most comparable GAAP measures, please refer to our earnings release, which can be found on our Investor Relations website. I will now turn the call over to Don. Please go ahead.

Good afternoon, and thank you for joining us. Before I discuss Kodiak's 2025 Q4 results, I want to frame the incredible long-term opportunity for Kodiak. We believe physical AI represents one of the most significant technology shifts in modern history. Physical AI is transforming how work gets done in a physical world and self-driving is leading this transformation. Autonomous systems will increasingly power how goods and people move, how freight is delivered and how packages reach people's doorsteps. The Kodiak Driver is at the forefront of this revolution. Scale matters, and we are building the system to scale. The key to deploying physical AI is harnessing the power of data center scale AI and then optimizing it to run a low-power compute that can fit on a range of form factors. We've already solved that core challenge by building the Kodiak Driver to be a modular physical AI platform that is commercialized today. The same core system that powers long-haul freight is built to extend into anything that moves from cars to pickup trucks, buses and delivery vans to even other form factors such as drones, construction equipment and humanoids. While we remain focused on our core trucking strategy, this platform leverage has the potential to significantly expand our long-term addressable market. We are building toward deploying the Kodiak Driver on thousands of trucks within the next few years. Our commercial roadmap, manufacturing partnerships and expansion strategy are aligned around that objective. As adoption increases across major freight corridors, autonomous trucks will move from limited deployments to everyday infrastructure. I believe that in the coming years, you will not be able to drive on any highway in the United States without encountering a Kodiak truck. Our objective is clear, removing humans from goods delivery with AI making every decision in real time. But that feature is not a distant concept. It's beginning now. This quarter marked another step forward in deploying safe, scalable, fully driverless operations with no safety driver in the cab and no required human remote monitoring. As the physical AI revolution has accelerated, so has Kodiak, which has been fueled by our industry-leading deployments. In 2025 alone, we advanced our core AI capabilities and expanded our product footprint at a pace that significantly exceeded what we achieved in prior years. That acceleration is translating directly into stronger reliability metrics, higher operational readiness and increasing commercial confidence. Our execution priorities remain disciplined and measurable, safety performance, system reliability, commercial scaling, cost structure improvement and capital efficiency. With accelerating technical progress and a clear path to scale deployment, Kodiak will continue to lead. That leadership is now translating into measurable results, which I would now like to discuss. 2025 was a monumental year for Kodiak. The fourth quarter was our first full quarter as a public company, and we delivered results that exceeded our expectations across all guided metrics, including truck deliveries and cash burn. We have made significant progress across product, commercialization and partnerships, entering 2026 with strong momentum. We remain on track for our long-haul driverless launch in late 2026 and continue to execute against our initial 100-truck commitment with Atlas. Our continued progress extends well beyond West Texas' Permian Basin, where our first driverless trucks are deployed. During the last quarter and early this year, we strengthened our ecosystem through collaborations with blue-chip companies like Bosch and Verizon. We secured a new contract with the U.S. Marine Corps and launched operations between Dallas and El Paso. Together, these milestones demonstrate growing market validation, expanding strategic relationships and increasing confidence in our ability to deliver autonomous trucking at scale. Just over a year ago, we delivered the first ever driverless Class 8 trucks to a customer. On our third quarter call, we outlined a plan to expand the initial 2 truck deployment to mid- to high teens by the end of the year. I'm proud to say that we finished 2025 with 20 driverless trucks in our customers' hands, which represents 100% growth quarter-over-quarter. This represents the largest customer-owned driverless trucking deployment in the world. We also remain the only company operating through an asset-light Driver-as-a-Service model, which generates multiyear recurring revenue, enabling us to scale faster with less capital. As of the end of 2025, those driverless trucks have collectively driven more than 10,700 revenue-generating hours with no one inside the cab. To put that into perspective, assuming an average highway speed of about 65 miles per hour, this would be equivalent to around 700,000 miles or nearly 30 trips around the planet. In addition, these trucks operate safely without continuous remote monitoring, a capability that is critical to achieving attractive unit economics and meaningful scale. Overall, as of the end of Q4, Kodiak's Autonomy System has delivered more than 12,600 loads, representing an 87% increase in loads delivered compared to year-end 2024. These achievements reflect our focused development and disciplined capital-efficient execution that we believe will enable us to achieve profitability and free cash flow earlier as we scale. I want to underscore, scaling autonomous trucking is hard. None of these results magically appear overnight, and there are no shortcuts. But as I've discussed before, AV trucking is not just about technology. To build and scale a physical AI business, you need to execute across three distinct pillars: technology, safety and product. In Q4, we made strong progress across each. I'll begin with technology. As a reminder, the Kodiak Driver is a physical AI-powered virtual driver that integrates with our modular vehicle-agnostic hardware to work seamlessly across different vehicle platforms and driving environments. Kodiak-powered vehicles operate around the clock in sun, rain, dust storms and other inclement weather. By operating without high-definition maps, our system can adapt in real time and perform reliably in complex unstructured environments. This flexibility allows us to deploy the same technology across long-haul, industrial and defense applications and in the future into adjacent opportunities. In 2025, we accelerated our use of AI coding tools and are already recognizing efficiencies from the use of tools such as OpenAI Codex. These AI tools are accelerating our development process, allowing us to do even more with less. We are also executing rapidly on key technical milestones for our deployed driverless product. One great example is our progress on hauling nonstandard trailer types. You may recall that in Q3, we added the ability to haul doubles or 2 trailers at the same time. We recently became the first AV company to pull triple trailers, which when combined with the tractor, weigh over 275,000 pounds or 137 tons and extend more than half the length of a football field. Hauling triples requires extreme precision and enables us to provide our industrial customers higher asset utilization and a more cost-effective solution. We also recently began hauling car trailers, another autonomous industry first. This achievement is only possible because of Kodiak's unique sensor pod configuration. Mounting sensors on the top of the vehicle would interfere with the car trailers, which extend over the cab. We believe that these advanced capabilities put us well ahead of the competition in terms of the range of customers that we can efficiently serve. Our deployment with Atlas in the Permian Basin is helping us to hone our technology in countless ways. We've continued to refine the use of our AI safety agent, which allows the Kodiak driver's AI to identify rare scenarios that can be a challenge for more traditional perception techniques. This new feature allows us to better handle the long tail of complex edge cases and gives us further confidence as we move down the path towards long-haul driverless deployment. For example, the AI agent helped the Kodiak Driver to identify a dust tornado or dust devil in the middle of the road. We've also recently announced technology collaborations and made technology investments that we expect will accelerate our ability to scale and reduce AV hardware costs. Most notably, at CES in January, I had the pleasure of presenting our ambitious vision for industrializing and scaling the Kodiak Driver through our strategic collaboration with Bosch, one of the world's leading automotive suppliers. Together, Bosch and Kodiak are developing a next-generation redundant autonomous platform with integrated hardware, firmware and software designed for automotive grade reliability and manufacturing at scale with our partners at Roush and directly on the OEM factory floor. This partnership underscores our focus on developing physical AI, not hardware, and enhances our ability to scale autonomous trucking. We also continue to focus on meaningfully reducing our hardware costs and are pursuing a comprehensive AV hardware cost-down strategy across multiple major initiatives. First, we recently embarked on a major engineering program designed to lower the cost of our bill of materials. This effort will continue over the course of 2026, is already bearing fruit, and we will have more to share in future quarters. Second, our cooperation with ecosystem partners, Roush and Bosch, helps enable us to manufacture at scale while sourcing multiple components from the same vendor. We expect this effort to provide benefits beyond just 2026. Further, we anticipate AV hardware, including sensors, compute and actuation to become further commoditized, making the incremental hardware costs associated with autonomy marginal. These efforts are critical as reduced hardware costs are a key driver to achieving profitability and generating positive free cash flow over time. Lastly, we have made key investments in our off-board infrastructure required for autonomous operations at scale. Our commercial connectivity agreement with Verizon is now enabling reliable connectivity through 5G and LTE, supporting remote assistance, over-the-air software updates and centralized fleet management. This connectivity is foundational to our DaaS model, ensures our customers have full-time visibility into their autonomous fleet and enable safe and reliable 24/7 autonomous operations. Moving on to the safety pillar. We continue to make meaningful progress toward closing our long-haul safety case. Our autonomy readiness measure rose to 84% as of February of 2026. During Q4, we did much of the underlying functional safety work that allows us to validate the safety of key truck platform technologies, thereby demonstrating that technologies we developed for industrial driverless deployment are sufficiently robust for high-speed long-haul operations. In parallel, we established manufacturing, installation and verification processes with our partners to support both our next-generation industrial and long-haul platforms. This work relied heavily on the operational insights we have gained from our full year of driverless operations with Atlas. One of the most exciting tools driving our safety case development is a new cutting-edge technology we call Breakpoint. Breakpoint, which we built leveraging AI, enables us to test the Kodiak Driver against millions of realistic scenario variations while intelligently searching for the most challenging conditions. This enables our team to prioritize engineering workflow based on risk priority. Using Breakpoint, we have been able to preemptively discover and resolve extremely rare edge cases that we've not yet observed in our thousands of hours of real-world operations. This ability to precisely identify the focused work that we need to do to complete our safety case is a key contributor to our capital-efficient approach. Lastly, in February, we began testing at a new track, the American Center for Mobility proving ground in Southeast Michigan. The ACM proving ground is one of only a few tracks in the U.S. large enough to bring long-haul trucks up to highway speeds. Our structured testing at the track will allow us to evaluate rare scenarios under near real-world conditions at highway speeds. This kind of testing is critical to us completing work on our initial long-haul safety case and launching in late 2026. Lastly, I would like to briefly discuss our progress across the product pillar, where our technology, safety case and operations combined to give our customers a great experience. As I previously mentioned, we've completed a full year of real-world driverless operations with our customer, Atlas Energy Solutions, one of the largest and most sophisticated sand logistics providers in the Permian Basin. This has yielded tremendous benefits, including the ability to pressure test the core components of our autonomous driving platform, which are shared across different driving environments. This generates a flywheel effect, allowing us to leverage these features and our learnings across long-haul and defense deployments. We continue to learn from this relationship and deployment. For example, the Kodiak Driver collects data that enables us to optimize predictive maintenance, which maximizes the uptime for our customers' trucks, which they deeply care about. Lastly, we've gained significant insights in how to optimize daily operational decisions, how to train up and team with customers, how trucks should behave at launch and land and how to optimize workflows. All these operational decisions have an outsized impact on throughput and are leverageable across verticals. We put a lot of focus on predictable performance in these areas and build playbooks and processes that will help us continue to scale in the Permian and beyond. Turning to our commercial pipeline. We have made significant progress across all three verticals. I'd like to start with the long-haul vertical. As we discussed on our first earnings call, we've been delivering freight using our fleet since 2019. We continue to scale these operations thoughtfully as customer demand and fleet capacity evolve. We've made significant commercial progress in long-haul operations over the last few months with active long-haul deployments with leading logistics companies, including J.B. Hunt, Werner and Martin-Brower. We also recently launched a new pilot with a major Fortune 500 private fleet hauling between Dallas and Houston. Today, we announced that we've launched a new route with Martin-Brower, logistics partner to some of the world's leading restaurant brands between Dallas and El Paso. This additional route with Martin-Brower brings our total weekly operational lanes to 8 and adds an additional route that stretches beyond a single hours of service. Our systems independence from HD maps has allowed us to quickly and easily add this lane. This new route represents a core tenet of our strategy of increasing our penetration with our existing customers. Our industry-leading customer base has an aggregate fleet size of over 125,000 trucks. This deep customer base means that even with moderate penetration into our current customer fleets, we will have all the demand that we can service in the coming years. I'd like to move on to defense, where we've made amazing progress over the last few months. We believe the policy environment is rapidly improving for dual-use physical AI developers like Kodiak. We're already starting to see the benefits of the Pentagon's increased focus on working with commercial first developers. In February, we onboarded Chet Gryczan, a leader with nearly 20 years of experience in the defense ground vehicle ecosystem as our new Vice President and Managing Director of Defense to help us capture all these new opportunities. Additionally, in February, we announced that we were awarded a contract with the U.S. Marine Corps to integrate the Kodiak Driver into the ROGUE-Fires carrier vehicle. This is an uncrewed ground platform designed to support distributed operations, expeditionary fires and force projection in contested environments in the Indo-Pacific region. Of course, the ROGUE-Fires work leverages the same production-ready autonomy platform that underpins our core autonomous trucking business, reinforcing the strength and scalability of Kodiak's approach across both commercial and defense markets. Kodiak was recently selected to participate in two high-profile U.S. Army demonstration events, the Army's xTech Overwatch demonstration in Texas and the Defense Innovation Unit's Project G.I. For Project G.I., a major autonomy experiment, which focused on contested logistics in the Indo-Pacific, we were selected as one of just 12 companies chosen for more than 400 submissions. These exercises highlight the growing demand within the Department of War for commercially developed autonomy solutions. For Project G.I., Kodiak deployed our autonomous Ford F-150s to Hawaii, where we conducted operational demonstrations in support of the 25th Infantry division, a key Army unit responsible for operations in INDOPACOM. The exercise provided soldiers the opportunity to interact directly with Kodiak's autonomy technology in realistic operational conditions, demonstrating how our off-road autonomy system can enable distributed operations and more resilient logistics in complex terrain. These events underscore the military's increasing interest in production-ready dual-use autonomous systems and reinforce Kodiak's position as a flexible autonomy partner capable of supporting missions ranging from tactical mobility to large-scale logistics. Looking ahead, we see strong opportunities to extend our commercial long-haul trucking autonomy platform into defense logistics applications at scale. Lastly, we continue to explore opportunities in the industrial trucking vertical, which includes oil and gas, mineral transportation and logging transportation. Industrial trucking represents a promising opportunity for Kodiak, given that operators in remote industrial locations face even greater difficulties recruiting and retaining drivers than long-haul carriers. We continue to have productive conversations with major industrial trucking operators in the U.S. and also see significant opportunities abroad, specifically in Australia, Canada and the Middle East. In summary, the experience we've gained over the past year with driverless operations has established a durable foundation for our next phase of growth. By running our system around the clock, we've built the technical maturity and organizational muscle memory required to scale. That experience compounds, each deployment benefiting from the capabilities and processes we've already developed. This positions us well as we prepare to launch driverless long-haul operations in late 2026. Looking ahead, we will remain focused on disciplined execution, expanding driverless deployment in a way that is both safe and sustainable. I want to give a huge thanks to the team for executing at a high level as we scale this business. 2025 was a fantastic year, and we expect 2026 to be even stronger as we continue to accelerate our deployment and continue to deliver on our milestones. And now over to Surajit.

Thank you, Don, and good afternoon, everyone. I am pleased to share Kodiak's financial results for the fourth quarter and full fiscal year 2025. 2025 was a transformative year for Kodiak, culminating in a strong Q4 that exceeded our expectations across all guided metrics. We continue to demonstrate disciplined execution, improved operating leverage and continue to realize benefits of an asset-light business model. We ended Q4 FY 2025 with 20 customer-owned driverless trucks, exceeding the mid- to high teens range, which we had previously guided. This outperformance was driven by successful deployment with our existing industrial customer. Q4 revenue was $1.1 million, representing 37% quarter-over-quarter growth. This was primarily driven by an increase in Driver-as-a-Service revenue generated by 100% quarter-over-quarter growth in customer-owned driverless trucks. This performance underscores the strength of our business model. As the Kodiak Driver is deployed, we benefit from scalable multiyear recurring revenue while continuing our historical discipline on capital requirements. GAAP operating loss for the fourth quarter was $39 million. Non-GAAP operating loss for the quarter, which excludes stock-based compensation, totaled $30 million, primarily due to continued investment in R&D and operations support for our industrial deployment. We incurred capital expenditures of $10 million primarily to purchase AV components that we deploy on our customers' trucks. Turning to cash flow. We outperformed our guidance for Q4 with free cash flow of negative $34 million, less than the negative $36 million to $38 million range, which we had previously provided. This outperformance was driven by improved operating leverage and continued prudent spending and was partially offset by an increase in AV hardware CapEx. We ended 2025 with cash and cash equivalents and marketable securities of $121 million. This number includes increased liquidity from the debt refinancing, which we completed at the end of the fourth quarter. The successful financing upsized the facility to $30 million, lowered our interest rate and extended the maturity profile to early 2030. Importantly, this facility requires no principal repayments until 2028, providing improved liquidity and increased flexibility to operate and scale our business. For fiscal 2026, we expect free cash flow to be in the range of negative $160 million to negative $170 million. The free cash flow burn is expected to be driven by capital expenditures to purchase AV hardware that we deploy on our customers' trucks and R&D investments as we prepare to launch long-haul driverless operations in late 2026. These investments are expected to be partially offset by increasing DaaS revenue and ongoing improvements in AV hardware unit costs as we continue to be disciplined in our spending. For the first quarter of fiscal 2026, we expect to end the quarter with driverless trucks in the high 20s. Q1 FY '26 free cash flow is expected to be in the range of negative $36 million to negative $38 million as we continue to invest in R&D, operational support and incur capital expenditures to purchase and deploy AV hardware. Based on our current operating plan, we have liquidity into Q4 of FY 2026. We continue to be focused on strengthening our liquidity and supporting the next phase of growth, particularly as we build our customer pipeline and launch driverless commercial long-haul operations in late 2026. As previously stated, we will continue to be disciplined and opportunistic regarding financing options with respect to timing and market conditions. In summary, Q4 marked a strong finish to 2025. We delivered strong revenue growth, exceeded our truck deployment and free cash flow guidance and maintained a disciplined approach to capital allocation. Our financial priorities remain unchanged: grow high-quality recurring DaaS revenue; invest prudently in R&D, including safety validation and verification to support long-haul driverless operations; drive down AV unit hardware costs; improve operating leverage through scale and efficiency; and raise capital to maintain a strong balance sheet. We are entering 2026 with momentum, a clear execution plan and a capital-efficient business model designed to scale, drive towards profitability and generate free cash flow over time. Operator, please open the line for questions.

[Operator Instructions] Our first question comes from the line of Andres Sheppard of Cantor Fitzgerald.

This is Anand on for Andres. Congrats on the quarter. So I was wondering what the pathway towards long-haul launch this year looks like? And maybe what are the milestones you're looking at to get there as we're focused on the autonomy readiness measure as you ramp up to reach that point?

Thanks, Anand. Well, as we've discussed and described previously, the safety case framework consists of a series of claims, and those claims pertain to various aspects of the Kodiak Driver, be it hardware, reliability, functional safety, performance of intended functions, handling of various scenarios and cases that we encounter on the roadway. And really, it's kind of a mechanical process of testing in simulation, testing in the real world, testing on test tracks that allow us to close those claims over time. I think we laid great foundation work in Q4, as we said during the prepared remarks for a lot of the infrastructure work that we need to do. And now it is just really a kind of turning the crank process to close the remaining claims in the safety case. There are no specific kind of feature milestones, if you will, if that's what you're looking for, that remain for us to complete that process. I'll just remind you that this is something that we've already gone through multiple times with our industrial launch. We are already driving vehicles with nobody in the cab on a day-in and day-out basis across various weather conditions and various environmental conditions as well. So we need to simply bring in the additional scenarios and cases that we encounter at higher speeds, and we need to effectively check the box to ensure that all of the performance that we expect to be there is actually there. And so you should really think about closing the safety case as a testing framework as opposed to a feature development or implementation framework. And this is a very manual and tedious process because, as you can imagine, there's lots of scenarios to close, and there's lots of I's to dot and T's to cross. And so we'll continue to update you throughout the year on our progress towards that. But as we said in the remarks, we remain confident in our time line that we've set out, and we're looking forward to the launch later this year.

Got it. Appreciate the color. And as a quick follow-up, I wanted to touch a little bit on defense with the recent ROGUE-Fires contract with the military and demos with the Pentagon. I was wondering if you could talk to us a little on how we should see the near-term opportunities with the Army or DoD and what type of contracts maybe we can focus on materializing or if you can quantify this opportunity a little bit further for us?

Sure. Thanks for the question. Yes, I mean, we're really excited about defense. Hopefully, that came across in the remarks, and I'll just reiterate that here. We've been talking about the exciting opportunities within the defense space for the last several years. And as you can imagine, it's -- the rhetoric is heating up. Congress and the Department of War have made it very clear that they want to field more technology from commercially mature companies. And that's certainly something that Kodiak delivers on the autonomy side, uniquely so, might I add. And so we do see an increase in partnerships and contracts being awarded in the future. The ROGUE-Fires contract was a really exciting one because it brings in a new branch of the military. Previously, we've worked with the Air Force and the Army. And now we're getting to demonstrate the great capabilities of the Kodiak Driver to the Marine Corps. And so as we broaden our experience, broaden our demonstration footprint, we are going to see more and more traction towards defense. As we previously -- as we've stated in the past, it's very difficult to give timelines or any kind of guidance towards defense or defense contracts. And so we're not prepared to do that today. That being said, we do expect these relationships and opportunities to accelerate throughout 2026, and we're very excited about the opportunity within defense more broadly.

Our next question comes from the line of Mike Latimore of Northland Capital Markets.

Congrats on getting to 20 vehicles there. That's great. You need to get to 100% on the ARM metric before you launch on highway long haul. I guess, is there a definitive lag timeframe between getting to 100% and then being able to commercially launch? Or can you kind of do it simultaneously? Just trying to get a sense of the timeframe between once you get to that metric and then you can -- and then the time to launch.

Yes, I don't want to split hairs and talk about minutes, hours, days, et cetera. But effectively, yes, getting to 100% by our definition and our metric means that we are definitively ready to launch. And so you can think about those happening effectively in parallel.

And then on the 20 vehicles, live vehicles, active vehicles, what is the exit ARR rate on those?

Yes. Mike, thanks for the question. We are not specifically guiding on ARR. However, we are exiting 2025 with approximately mid-single-digit millions of annualized recurring DaaS revenue that does not factor in any other revenues from Kodiak Express or any upside from defense.

Got it. And so the pricing that you've envisioned is holding as expected?

Yes. Yes, we're not going to just comment specifically on pricing and specific contracts, but we are still executing as per our prior plan.

Our next question comes from the line of Itay Michaeli of TD Cowen.

Just a first question. As you've expanded the driverless trucks and the hours of paid driverless operations, I'm curious whether the ODD has also expanded. Are the trucks doing kind of new different routes, more complex? And maybe just also just talk about any kind of rates of remote assistance and interventions, things of that nature as well.

Yes. Thanks, Itay. Yes, absolutely. It has expanded. I mean you have to remember, for us, these are trucks owned and operated by the customer. So we don't control the trucks. We don't determine where they operate. We don't tell the operator -- sorry, we don't tell the customer where they can operate these vehicles. They choose the routes, they choose their operations. And the Permian is a 75,000 square mile area. And obviously, Atlas being one of the more sophisticated operators in the region, they utilize a conveyor belt called the Dune Express in order to move sand from the mine into some of the polygon regions that are most attractive to their customers and then they use trucks to carry from the endpoint of the conveyor belt to the end well sites. Those well sites might be 10, 20, even 40 miles away, and they change all the time. So every couple of weeks, the routes are changing. So there's new routes. The sites are only used for a handful of weeks at a time. It could be anywhere from 2 to 4 weeks for a well site to be in operation. And so the routes are continuously changing. Some routes are more difficult functionally than others. And so yes, the ODD continues to shift on a regular basis for us. We're serving multiple well sites with our vehicles and some routes are incredibly narrow. I would say the most difficult part of operating in the Permian is the complexity of the terrain in terms of the bumpiness, the amount of potholes, the unevenness. It's just very -- it's very harsh on the hardware. It's very harsh on the vehicles. And then the narrow lanes is what makes it very difficult because these are bidirectional lanes with lots of traffic. And so passing other vehicles in close proximity on a regular basis where you're kind of going back and forth all day can be very, very challenging. And so for narrow lanes, that's where the ODD is, I would say, most difficult. And then you throw weather on top of that, and that adds additional complication. But yes, as we expand, we obviously see new things, see new challenges and increase the diversity of which the system has operated. In terms of disengagement metrics, there are no disengagements for a driverless vehicle. There's nobody there to disengage. I hope that is clear. In terms of like remote assistance usage, we haven't publicly disclosed specific usage numbers. That being said, the primary use of remote assistance for the Kodiak system is in and around pickup and drop-off locations where you have really nuanced fine control of the vehicles in an environment that is changing constantly, and that's primarily where the customer generally requests assistance for these vehicles.

That's very helpful, Don. And then just as a follow-up, I'm hoping just to go a bit more into the road map for the Bosch collaboration. Maybe how we should think about the opportunity over time for you to generate some future BOM savings and kind of other efficiencies from that important partnership.

Yes. Thanks for that question. This is a really exciting announcement for Kodiak. Obviously, CES was a really great event for us. Bosch being a fantastic partner. We're already in deep collaboration, working with them on future generations of the Kodiak Driver where we believe we can bring additional efficiencies through single supplier and mature supplier relationships like Bosch. We haven't provided any specific references to costs or timelines for that work, but we definitely look forward to continuing on our public disclosure with Bosch at future events, and we'll have more to say later on.

This is Surajit. Just to add on to what Don mentioned, Bosch is part of our scaling and driving down BOM cost strategy. So as Don mentioned in his prepared remarks, we are executing on three levers, I would say. So we have made some engineering design enhancements, which are already yielding some safety -- sorry, savings results. Second, we -- as we increase the scale of the production and drive cost optimization, and Bosch and Roush, they fit in our overall global supply chain organization. Obviously, Roush fits in from providing high-quality assembly and Bosch from being able to manufacture at scale and source multiple components from the same vendor, that would allow us to drive down BOM costs materially over time.

Our next question comes from the line of James McIlree of Chardan Capital Markets.

Regarding the activity with Atlas, prior to recent events, the oil prices and the Baker Hughes rig count were declining. And I'm curious if that has had an impact on the deployment schedule with Atlas. Is 100 trucks still a viable goal with them?

Yes. Jim, thanks for the question. We remain committed to delivering the remaining 80 trucks on that initial 100-truck order over the next few quarters. The timing of those deliveries will largely be aligned with the customers' fleet planning and deployment schedule, but we do expect the rollout to be back-end weighted in the year. And we'll continue to work closely with the customer to ensure that the trucks are integrated efficiently into their operations as necessary. As you know, the oil market is somewhat cyclical and fluctuations do occur. Of course, things are changing literally by the day. At the same time, at the scales that we're talking about, our trucks are not materially affected by those types of fluctuations. And so we remain committed to the 80 trucks on the customers' fleet planning deployment schedule.

Great. That's helpful. And then as far as other industrial customers announced this year, is that a high likelihood? Is that a moderate possibility? Can you kind of frame the potential for other industrial customer announcements this year?

Yes. So thanks for that question. I mean the pipeline -- what I can say is that the pipeline is very strong. And we're talking to customers all over the world. As we mentioned in the prepared remarks, there's a strong pipeline in Australia. Obviously, the mineral resource industry there is one that fits our model perfectly because it's extremely remote, very difficult to find labor. That labor is incredibly expensive, and these are operations that run effectively continuously in remote regions. And so this is definitely an area where we're interested in expanding. Please understand that the complications with going to Australia is not trivial. And so we're working through the logistics there. And then, of course, the Middle East is very top of mind for us and others at the moment, both from a defense perspective, but also from an industrial perspective. And we're talking to several folks in that region, and we'll have more to announce when the time is right.

Our next question comes from the line of Walter Piecyk of LightShed.

First, I'm going to flex because I'm doing this question from the back of a Tesla robotaxi. Unfortunately, I have a safety attendant in this one. I have not gotten one without a safety attendant. So all the time, all the time. But for you guys, I just want to follow up on the Bosch question. Is there any issue in terms of exclusivity in terms of you guys -- obviously, there's another company out there that has shifted to upfitting themselves. Do you get kind of first bids on this stuff? And I guess if there's anything that's -- you have the cost-saving opportunities that are built into the production grade integration like sensor suite optimization or compute platform, is that stuff that you guys keep yourselves or that they can sell to others at the same time?

So one of the great things about the relationship with Bosch -- so there is no exclusivity, as you mentioned. However, we think that there's strength in the industry by providing companies like Bosch with guidance and experience really when it comes to developing these systems. And as we mentioned in the release with Bosch, this is not exclusive to only our upfit model with our integration partner in Roush, but this is also a system that can be deployed to OEMs for line-sign integration. And that, I think, is something that we're very excited about. This is a very close collaboration. It doesn't imply exclusivity. And so in theory, others could have access to the technologies that we're working on. But we think this is a rising tide raises all boats type of situation, and we're excited to work with one of the largest automotive suppliers in the world on a technology that we think is fundamental to scaling our business. And so we think that this is a pure upside in working with Bosch, and it's not exclusive to either upfit or OEMs. This is really a works-for-all situations type of collaboration.

Got it. And then on the -- first off, pat on the back on the $13 million, $14 million, whatever it is, R&D, obviously, super lean. But at 85%, like I guess the question is for that last 15%, I guess one of the kind of the bear case you could throw out there is like, easy to get to 85%, maybe it is, maybe it isn't. That last 15%, we're going to see a massive ramp in R&D, OpEx, whatever it is. Can you kind of address those concerns to the extent they're out there in terms of increased capital intensity to get 85% up to the point where you can really get this thing going?

I'll say a few things, and then I'll hand it over to Surajit on more of the financials. I think it's a very good point that you bring up. I think this is something that we're incredibly proud of being a capital-efficient lean company. We always have been. That's in our DNA, and it has been for the 8 years that Kodiak has been in existence. And so we don't expect that to change. A couple of things, I think, that are worth mentioning, as I mentioned in my remarks, AI is really accelerating progress when it comes to development, right? This is happening around the industry, but Kodiak has fully embraced the use of AI and generative AI tools for coding and process assistant. And what we're seeing is the ability to accelerate a lot of the work that would have otherwise taken significantly more resources on the R&D side, and we can get that done much more capital efficiently than we would have been able to otherwise. And we've been seeing the fruits of that over the last several quarters, right? That's not something that's just going to turn on this year. But obviously, as those models become more powerful, you will see a rapid acceleration in our ability to make progress at a much lower cost. And then the second piece I would say is that we did mention the expected free cash flow on a yearly basis. And so we don't fundamentally expect that to increase significantly. Maybe I'll turn it over to Surajit to address that.

Don, before you turn it over to him, just a quick follow-up on as the AI helps you now versus what it was before. Could you put a timeline on that? Because obviously, like there seems to be an acceleration, obviously happening in autonomy right now because of AI. How long ago was it that if someone was developing AI, they wouldn't have had that benefit. So they wouldn't have had that getting from 85 to 95 at a much more efficient place. Is that like 1 or 2 years ago, 5 years ago? How recent has the evolution of AI really helped the development costs for a company like yours?

I would say that's within the last year. So if you look at the last -- early 2025, these technologies were just starting to take hold. People didn't really understand them well. We didn't really understand how to use them effectively at large scale -- for large-scale development. So it's important to distinguish between a solo person at their computer working on some code versus an entire repo that's read and understood by an organization. So you have to use the tools a little bit more carefully, and it's taken some time to figure that out. But once you have, they really come of age and they're a rapid accelerant of progress. So yes, in the last 2 quarters, I would say, last 6 months, you've really seen an acceleration in that development, and I think that acceleration will continue throughout 2026.

Walter, this is Surajit here. Just on the question of investments. As Don mentioned earlier, most of our incremental R&D investments to get to long-haul is mostly on safety validation and systems. And that will require some incremental investment, but we expect most of that to be offset by the increasing revenue we are getting from our industrial customer as we deploy more trucks. So that's, I think, our beauty of our model being able to operate across three verticals. We are able to leverage the other verticals to offset some of the increased expenses in the other areas.

Thank you. And ladies and gentlemen, that is all the time we have for questions, and that does also conclude today's conference call. Thank you for participating. You may now disconnect.