All right. We'll get started since it's 10:15 A.M. Thanks, everybody, for joining. We are here with Kodiak AI. Today we have Don Burnette, the Founder and CEO of Kodiak. Thank you for joining us once again this year.
Yeah. Thanks for having me.
We'll kick it off with, since Kodiak is kind of in a new spot of transportation. Obviously, transportation's been a big part of the economy forever, but Kodiak brings a unique spice to that part of the economy. Can you start off with the elevator pitch of where you're built to win in the marketplace and kind of who Kodiak is from a broader sense?
Yeah, sure. A lot has changed in the last year. It's good to be back. The freight industry is massive. You mentioned a major part of the economy. I would say it's the backbone of every major economy, including here in the United States. And the industry faces tremendous challenges. There's rising insurance costs, there's rising labor costs, there's shortage of labor, there's a difficulty of managing labor pools. There's a real growth challenge in the market as well for especially a lot of the larger players. And we believe that autonomy effectively unlocks all of these things. We have an opportunity to reduce costs, we have the opportunity to increase efficiency, we have the opportunity to increase safety within our roadways. And all of these ultimately are essential to a growing economy and meeting the demand of tomorrow or even today.
Kodiak's place is that we're technology developers. We're specifically focused on AI software in the Physical AI space. We want to take the lessons from the data center AI developers and bring those techniques to the physical world. Kodiak has chosen to do that within the transportation space because we think it's the biggest opportunity within physical AI. For those of us who have been working on this for a very long time, it's been multiple decades, and we're finally at a point now where we have safe and reliable technology that we can deploy at scale. For Kodiak, we've been deploying a driverless fleet now for well over two years. It's the largest fully driver-out fleet that we're aware of. The fleet is owned and operated by the end customer. That's a big step within our industry.
We've gone far enough with the reliability of the system that we can actually hand it over to somebody else. There's no other autonomy provider that we're aware of that has stepped into the kind of reliability chain of, we can just hand this to a customer, they own it, they operate it, and it's theirs to work with on a day in and day out basis. We've really achieved that. We did that in our industrial application. Now we're bringing that to our highway customers. We plan to release our driverless solution this year, so next month, by the end of the year. Then we plan to start scaling that with customers starting in 2027 and beyond. We've actually signed up for the first, we have the industry's first committed driverless highway contract, which is very exciting, with DTL Transport. We can talk about that.
Yeah, it's a really exciting inflection point for the company to have spent so many years developing this technology to finally get it across the line, get it into production, cross the safety bar, and actually start to really make a difference on U.S. highways.
Yeah. No, definitely. To kind of say where you're at here with the DTL Transport and the agreement you have with them, if you can kind of outline what that looks like and what that means for that, kind of paired with your partnership with IKEA, if you can talk about those two and how that translates to demand for your services.
For many years, there's been MOUs, promises, letters of intent, handshake deals, general positive sentiment around self-driving. We've seen that from many companies, and we've seen it for a long period of time. There was a period where there was a bunch of reservations made, and what did reservations really mean? There was really no teeth to it. Up until recently, that's kind of where the industry has stood. If you talk to trucking fleets or carriers or folks who actually operate trucks, they will all tell you now that they're excited about the future of autonomy. But there's still a lot of skepticism in the industry about when it will actually be available. There's been a lot of starts and stops. There's been a lot of hiccups along the way, and within the industry broadly.
Generally, making a firm commitment has been something that the customers have been apprehensive to kind of take that step. What it takes is a track record of building trust and being able to show with data that this solution actually is an improvement to your business, that this is an improvement to your ROI, that this is going to somehow either improve your economics or allow you to expand. The step we've taken with DTL Transport is just that. We've gotten to a point where they saw our existing deployments.
They saw the work with our safety case, for our highway deployment, and they saw what we've been doing with other customers such as IKEA, and many others, and they said, "Okay, we trust that Kodiak is going to have the right solution for us, and we're willing to take that leap and sign a binding agreement to purchase these trucks starting in 2028." It's a start. It's 75 trucks, but you have to start somewhere. People love to talk about tens of thousands, hundreds of thousands, millions of trucks, but you have to get to one, you have to get to 10 trucks, you have to get to 100 trucks before you achieve those numbers.
We think that this is a great starting point and a testament to the quality, reliability, and safety that the Kodiak Driver system brings to the market, for us to be the first to actually sign a customer in that regards. On the IKEA side, we've been working with IKEA for many years. It's our longest-standing customer today. We work with them on a daily basis, and they're always at the forefront on technology, thinking, "How can we make things easier? How can we make things more efficient? How can we make things cheaper?" That's in their DNA, and it has been for the lifetime of their company.
We're really proud to be working with them, and they're very excited to be the very first commercial driverless load that we carry next month, and we intend to continue to work with them on an ongoing basis into the future.
That's very exciting. If we can also talk about one of your other big customers, Atlas has been one of your biggest proof points in recent history. They've been a very good partner of yours and a good customer. If you can talk about some of the learnings that you have from that and why maybe Atlas was a good place to start at scale.
Atlas is a very forward-thinking trucking company, a little bit less conventional. They operate a fleet of trucks in a place called the Permian Basin. That's the oil and gas fields of West Texas. In a lot of ways, it's very similar to the freight industry that runs on highways. It's just that those road networks happen to be mostly unpaved roads. But they're still roads. There's still intersections, there's traffic. In fact, they're not separated the way highways are, and so you have oncoming traffic, which presents some additional challenges for autonomy. Often, those roads are very, very narrow. It turns out that getting drivers out into the Permian, which is hours away from civilization, you typically have to house people out there because they can't make the trip back and forth each day. Otherwise, it cuts into their shift time.
What's interesting about Atlas as well is that they already operated a 24/7 business, and so their cost base is significantly higher per truck than, for instance, a truckload carrier that operates a one driver to one truck ratio, where they might get seven or eight useful hours out of a day. Drivers are technically allowed to go 11 hours, but pretty much a lot of that time, they're on the clock from the moment they show up to the moment they leave, and so the actual useful driving hours on average is about eight hours per day for a driver. So a truck only gets eight productive hours per day in the over-the-road sense, whereas Atlas actually runs multiple shifts.
They already had this challenge of, "We're maximally efficient with our assets, but our costs are extremely high." The belief was, with autonomy, they could bring those costs down, and they could increase reliability of their fleet. They came to us. What was interesting about the story is that we were focused on highway and defense. We were not looking at this as an application, but they actually saw the work we were doing in the defense industry, and they reached out to us and said, "That looks a lot like what we need, and we've been looking for an autonomy solution for many years, and we haven't found one. Can you come and automate our system?" The jump into industrial applications was really not something that was on our original path. The customer came to us and said, "We need a solution.
Can you build it for us?" We thought, "This is a perfect environment for us to deploy our technology as a starting point because it's lower speed." In lower speed environments, you can build a larger safety envelope around your vehicle, and ultimately, safety is what we want to focus on. That turned out to be not quite true because of the oncoming traffic and the narrow roads and the tight squeezes and all the things that make autonomy extremely hard, but we set out to prove that our technology could get there. We deployed in December of 2024, and then the real work started. This is what people don't appreciate about autonomy, is the unknown unknowns, the things you don't understand about building the actual product. Getting to a driverless launch is seen and it's talked about as the end game, right?
Once you're there, you've solved it. There's really three legs to the stool. There's the technology, which is where 99% of the conversation is usually focused. There's the safety case, which is greatly underappreciated. Safety cases are incredibly complex. It requires PhD mathematicians and statisticians. It requires an intense amount of effort and testing and validation, but we've done that before. The product side, however, outside of robotaxis, where it took them a long time as well, nobody really had a product in the market, and so nobody really knew what it meant to operate that product, and we had to learn the hard way, as did Atlas with us, hand in hand. So, lessons learned, things like, how do the line managers and the site operators deal with queuing in self-driving cars?
Typically, the way it works is you have this loading facility, and in the Permian, everything is pull-through. So there's no disconnecting and connecting trailers. You use a lot of verbal cues and you use a lot of eye contact. "You go over here, you stay over there, follow that person in. I need everybody coming around this way today, not around that way like yesterday." They're changing it. It's dynamic. It's chaotic. For the last, I don't know, 60 years, that's the way it's been. Now you have these self-driving trucks that have none of those capabilities. So, putting the tools in the hands of the operator to efficiently use the product in a way that doesn't disrupt their flow, that was a very, very tough learning. We worked through all of that. Now, we call this the flywheel.
All these operational efficiencies, picking up, dropping off, customer support, truck gets stuck, what do they do? Who do they call? How do you handle it? How do you get the truck unstuck? How do you deal with maintenance issues, downtime, uptime? All these things we've kind of worked out. To get to the crux of your original question is, we can now take those learnings and use it as a springboard as we launch our over-the-road business in a way that we think is a huge competitive advantage relative to the rest of the market.
Yeah, I think, I'm sure that was a very big thing to point to whenever you were talking to DTL and IKEA, and selling them on your services. I do want to make sure we understand the Driver-as-a-Service model and kind of how important that is to the Kodiak story, to its scaling here in 2027 and 2028. So, if you can help us understand, one, for those that aren't familiar, what is Driver-as-a-Service, or DaaS, kind of similar to SaaS, but why that's important to you, what it is, maybe some of the economics around it, if you can help there a little bit.
Sure. Yeah. So, Driver-as-a-Service as opposed to transportation as a service, those are the two big buzzwords. Effectively, the difference is, in one, the customer owns and operates the asset, and that's DaaS. In TaaS, you own and operate the asset, and you work on behalf of the customer. So, they pay you to bring your asset to do a job for them, and effectively, you're a contractor for their work. It really comes down to asset-light versus asset-heavy model. There's pros and cons to both, but we firmly believe that Driver-as-a-Service is the superior business model and the one that we've been pursuing from the very beginning. Maintaining an asset-light business, allowing us to focus on what we're good at, which is AI software, as opposed to truck ownership and logistics, which we're not as experienced in.
That's the model that we deployed with Atlas, that's the model that we're going to deploy with DTL, and that's the model that we think is broadly the best one for us to grow and scale our business, and ultimately bring the benefits of self-driving to the industry. The one thing I will Well, it seems obvious, of course. Everybody should just do that, right? The one challenge with that model is that the bar is significantly higher, right? Again, it's exactly as I was talking about in the last answer, where the product suddenly becomes much more challenging. Because when your operations team operates the asset, operates the truck, they're trained experts. They know all these things. They use it on a daily basis. They're familiar with it. They helped you build it. They know the ins and outs. They know all the secrets.
They know all the special knobs to That's not a knob. The knobs to turn and the levers to pull. They know how to get it up and running and healthy and happy, and they can do it really well because they've been doing it for years, and they are the experts. When you hand this to a customer, we talk about partnerships and these contracts as deals between companies, really. You think about it as executives doing a handshake and signing some paper. But what about the end user? The end user is not the CEO, is not the exec team sitting in an office. The end user is the line manager at the dock. That's the person who ultimately has to use this on a daily basis. And quite frankly, those are the folks that may be curious about autonomy, but they don't really care.
They just want to get their job done, and it's got to be seamless, and it's got to be easy. So, the bar is significantly higher when you want to deploy DaaS in terms of how frictionless the process is, and also the reliability. When it's your truck and you can swap it, so if I work for you and you say, "Bring me a truck on Tuesday at 8:00 A.M., and you're going to do this job," you don't know which truck I'm going to bring. I might have had a pool of 10 trucks, and nine of them are down, and the one that's up, I bring to you. And you're like, "Seems great. They showed up with a truck." But you don't know the fact that my efficiency was one-tenth, and that actually I had to scramble just to get that one truck up.
That was all obfuscated. We swept that under the rug, and we dealt with it on the backend. Now, our economics are going to suffer from that, but from a customer perspective, you're none the wiser. Same with robotaxis, by the way. You only care about the robotaxi between when it picks you up and when it drops you off. If you're the only trip of the day, if that asset was otherwise broken, charging, circling the block, doing nothing productive, you don't care. You could care less, right? You only care when it shows up to do your job. But when you flip that, if you owned that robotaxi, you would care a lot about what it does between trips. You'd care about the efficiency. You'd care about the uptime, right? When you buy a product, you expect it to work, and you expect it to always work.
The reliability has to go up. The SLAs for uptime have to go up. The ease of use has to go up. Everything is significantly harder when you try to deploy a DaaS model, because you are no longer in control of the technology. Somebody else is, and they are not technologically savvy, right? We have to build this for people who do not understand the technology, and it is a much, much harder problem.
That makes a lot of sense. I think there are some other people out there in the market trying to solve similar problems to what you are trying to solve, whether that is autonomous trucking or something else. What differentiates you from competitors, from other offerings out there? I know one thing is that Kodiak is a little more modular than some of the others, so if you could maybe touch on the advantages of that as well.
Ultimately, any customer service business is about flexibility. We have said from day one at Kodiak that we want to be ultimately flexible for customers. So we have built our hardware and software solution to be platform agnostic as the starting point. So we can offer customers a multitude and a variety of platforms. We have now done with the addition of the International LT, which we recently announced, that is going to be our fourth driverless platform that we have brought to the market. We want to give customers choice, and we think that we can offer more choice to customers than our competition based on the fact that our system has been built from scratch to be extremely modular.
Also, if you are a mixed fleet and you operate multiple makes and models of trucks within the fleet, if you have the Kodiak Driver on all those different models, we want the system to be singular and universal so that if you need to do swaps, repairs, replacements, you do not have to have inventory for every specific make and model that you operate. This is where Kodiak's SensorPod solution comes in. The pods are universal, and they can fit onto any vehicle, any platform, any truck. The mounting points need to be obviously specific to that make and model, but those are fixed, and those are just brackets. Once we put the brackets on, now you can actually put your solution onto any vehicle. We think that is a game-changer in the market. So that is one.
The other one is our attitude toward operational efficiency and really capital usage. Kodiak as a company is significantly more capital efficient than our peers. From a burn perspective, we burn significantly less, orders of magnitude in a lot of cases, and yet we are able to achieve the same milestones and supply the market with the same, if not better, quality of product. We think that from a market perspective, from a shareholder and investor perspective, that ultimately is going to lead to more success and a better outcome down the road as we deploy at scale.
I also wanted to touch on something that you touched on earlier, which was the DTL, the autonomous long-haul unmanned trip coming up, right?
What did it take to get to this point? Will the next ones in the future take less now that it is going to be proven out here with DTL, or is it going to be a similar path as we try to implement this further?
I think the cheeky answer is it is a culmination of decades of work that have gone into getting us to this point, right? For me, I have been working on this for over 18 years. For many of my colleagues, very similar long-term development cycles. A lot of technological experience goes into this, a lot of iteration and improvement on the reliability of the system. We are on our seventh generation now of the hardware and software solution, and with every generation that we deploy, we learn more, we make improvements, we harden the system, we make it more reliable. For one, and I did not mention this before, one thing about the Permian that we learned is that it is pretty much the most rugged environment you can put a truck in, I think, in the U.S. I always challenge people to find a more rugged environment.
I don't think there is one. We've been able to harden the system such that it basically doesn't break, even in the most rugged conditions. Now we put it on the highway environment, it's basically like a silk carpet that the system is running on. We have high, high confidence that our system is going to be very reliable. We wanted to make sure that we got all the boxes checked, not just performance and safety, but also hardware reliability and our ability to scale, working with our manufacturing partners. Basically, everything has been built to this point, not just, "Hey, did we cross the safety bar?" Which is generally what companies are racing to do. We wanted to get all the pieces in place. Hardware solution, manufacturing, supply chain, yes, of course, performance, safety case, testing validation.
We wanted to get reps under our belt. We also needed to know about the product. How do we actually put this in customers' hands? What are the challenges they were going to face? We went and learned those lessons already. This is really a culmination of all of the work over the last nine years at Kodiak, everything we've learned so that we can offer the best product, but also so that we can scale quickly and without stumbling along the way as we actually get this product into the market.
That makes sense. Safety has been something you've mentioned a lot already, and BreakPoint is your in-house AI safety simulation, I believe.
Can you talk about how that has helped when you look for deployments, again, like the DTL, when we look for what you're going to do for customers in the future, how does that help with your safety rating and how you drive safety and build confidence in deploying these trucks onto the road?
Safety is really challenging. There is no handbook for autonomy safety. There is no certification standard. There is no government standard. There is no ISO standard. You cannot look it up. How do I prove safety for an autonomous vehicle? Well, it is a non-deterministic system, meaning we do not control all the variables in the environment. We have to make assumptions and make predictions and look at statistical analysis to show whether, in various cases, the truck is going to perform a certain way. We have had to borrow from other industries. We have gone to aerospace. We have gone to nuclear. We use a solution called a probabilistic risk assessment, PRA, which you can look up on Wikipedia. This is not something we invented. But it has really been utilized by NASA and some other industries where you do not get a lot of testing before you have to deploy.
If you are building a rocket ship, you can do some simulation, but there is not much testing you can do until you try to launch the rocket. That is it, effectively. Then you have to learn very quickly from a very limited amount of data. SpaceX is a great example of this. But the similar approaches that SpaceX uses to quantify their risk are exactly what we do here at Kodiak. These are all very complicated topics. They are highly mathematical. They require a lot of mathematical knowledge. But also, this is an area where we are having to invent it ourselves. Again, there is no prescription on how to do it, and every company is sort of working on this, and it is, I would argue, as difficult as developing the technology itself.
BreakPoint specifically is a technology we developed in-house that allowed us to go and find the corner cases, to find the edge cases. Everybody likes to talk about the edge cases, and the question is, how do you test for edge cases? Well, you simulate, right? We use simulation. Everybody uses simulation. You can simulate millions and millions of miles, billions and billions of miles. But it turns out that almost all of the miles that you drive and most of the miles that you simulate are going to be mundane. They are going to be boring. They are going to be "easy" cases. Even hard cases are "easy" when you are talking about finding the true rare edge cases. BreakPoint was specifically a technology to allow Kodiak to go and search the space to look for edge cases in an intelligent way.
It is an AI solution that finds pockets of difficulty within this huge 10,000+ dimensional space that cannot be searched through brute force and is not practical or feasible to test in any kind of traditional way. We go and we find what we call challenging difficulty islands within the space. It finds where the truck kind of is not performing super well, and then it goes really deep into those directions very efficiently. We can test a much higher magnitude of edge cases without burning as much dollars and compute as it would have taken to just brute force our way through billions and billions and billions of miles, hoping to find challenging cases.
That was a big unlock from a technological perspective that allowed us to use our statistical calculations to say, "Okay, in these extremely rare conditions, we know that the Kodiak Driver is going to be safe, or at least it's going to perform along some metrics that are as good, if not better than what a human can do in those cases." Yeah, go ahead.
What's the liability framework? The system behaves as expected and makes the least bad decision, but the least bad decision still results in some kind of injury or fatality. Who bears that? Who bears that? You or the operator? Has that even been settled yet?
Just for the webcast, the question is on liability and situations in which Kodiak might be involved.
Yeah, in the U.S. judicial system, the answer is everybody gets sued, right? Everybody carries liability insurance, no matter who you are. We look at this from a slightly different perspective. Accidents will inevitably happen. Again, if you look at deployments, like Waymo gets into accidents all the time, right? It's not even eventful anymore. The question is, how do you assign fault, and how do you prove defensibility, and then what data do you have at your disposal? Part of the challenge in the trucking industry is that there's a lack of data, and it becomes a he said, she said type of situation. And when there's severe accidents, juries tend to be more sympathetic towards victims. One of the benefits that self-driving vehicles have is that we have constant monitoring, 360 degrees around the vehicle, many, many times a second through lidars, cameras, radars.
We're tracking every vehicle. We can reconstruct the accident into incredibly minute details and basically say, "Here's what the steering wheel of that other vehicle was doing. Here's what the brake and gas pedal were doing." We can lay out the evidence. It becomes a more objective perspective than a subjective one. That alone is a game-changer within the liability framework. Then the next thing is, our trucks are always driving defensively, as you posed in the original question, and like you assume the best. That tends to knock down the most severe of the accidents. So, if you're minding your own business, driving in the right-hand lane, going the speed limit, being very defensive, maintaining maximum distance around all the vehicles and other objects around you, things tend to go better than they do if you're being a more aggressive driver or an inattentive driver.
Our Kodiak Driver's always paying attention, and it never gets drowsy. So, we have all of those factors to our benefit. That being said, when things inevitably go wrong, we also carry a tower of insurance, product liability, like any product liability insurance tower would, as do our customers for their product. But what we're seeing in the market, what I can actually give you as an example, is that with our existing deployments, we're already on parity from an insurance cost perspective with manual drivers. We believe that those costs of insurance is going to come down over time as we prove that we can eliminate some of these, particularly what we call nuclear verdicts within the industry. That takes time to prove, but we believe that we will be able to prove it out. Thanks for the question.
As we look at Waymo's and other autonomous driving vehicles already out there, are there any learnings from the core systems that you can bring to potentially inform your decisions going forward within
Kodiak. Yeah, it goes down to those three pillars, right? There's technology, there's safety, and there's product. I think there's lessons to be learned from across the spectrum. From a technology perspective, the industry is actually very open, right? A lot of people publish results. There's papers coming out every year. There's new improvements in AI. There's new model structures. There's new training techniques and methodologies. We, of course, are staying at the forefront of all this technology and utilizing all the best practices that people are bringing to conferences and showing the world. Of course, we're developing our own IP and pushing in our own directions there. Those are things that we can definitely learn, and we have learned as an industry to make the product actually better. On the safety case, there isn't as much collaboration, unfortunately. I think there should be.
I think it would be great to have more collaboration on safety certification standards. We're working on that behind the scenes with NHTSA and other government bodies, and with our peer companies. On the product side, absolutely, right? You can look at the robotaxis and see some of the challenges that they've gone through on pickup and drop-off, on usability, on in-vehicle experience, and you can draw lines and extrapolate from those learnings that they've had to go through and they've put their money into learning, and we can just kind of take the end result and say, "Okay, this is something that they've done, so we should probably read into that." Trucking and robotaxis are slightly different worlds, or very different worlds, so there's only so much. I think our own experience in deploying with Atlas is probably the best learning that we can leverage.
As you earlier mentioned, capital efficiency is one of the things that gives Kodiak a leg up. Can you walk us through maybe the capital efficiency where you have financing options to support growth into 2027? Right now, you are still ramping, kind of scaling the business offerings in your DaaS solution, so right now, you're burning a little bit of cash, but talk about the financing options that you have to help support the scaling of this.
Well, I'm limited in what I can say here other than we've been open and upfront about the fact that we'll need to raise more capital. We've recently said we're targeting free cash flow breakeven in 2029 as we exit with about 5,000 trucks deployed in 2029. We have a plan, from everything from manufacturing, technology, customer demand, et cetera, to get to that point. We will need to raise capital, and we're looking at opportunistic ways to raise capital, everything from strategic investments, partners, customers, to traditional financial institutions, in both the equity and the debt markets. So, there's different ways to capitalize the company. A lot of it is CapEx, honestly. We call it success-based CapEx. The more you build, the larger you scale, the more it costs to build things.
Certainly looking into equipment financing and other kind of asset-backed financing is something that we're working on. I think that's probably all I can say right now.
Yeah. With the last minute we have here, can you just talk about what you want us, sell-side analysts and investors, to take away from this meeting as we think about the investment case for Kodiak going forward?
The space is massive. The opportunity is massive. I think there's no more question, or rather, I will say there shouldn't be any more question as to whether or not autonomy is the future of transportation. Seeing is believing. If you haven't been in a self-driving vehicle, I encourage you to go ride in a Waymo and experience it for yourself. I promise you that after you have, you will believe that autonomy is the future of all transportation, both commercial and passenger. That is an unfathomably large opportunity that we sit at the precipice of. The last piece then is, who is going to take this market? How do you slice and dice it? I think ultimately there, that comes down to trust.
My goal is to build trust with the investment community, to continue to hit our milestones, to continue to beat our guidance, to continue to deliver on the things that we've said we're going to deliver on. So far, as we've gone public, we've hit every single milestone. We've delivered on everything we said we would, and we intend to do that. I really want to build that trust with you all. I think that we are in the driver's seat, pun intended, to go and capture this massive market.
That's great. Thank you, Don.
Thank you.