코디악 CEO는 트럭의 자율 주행 구현이 성공의 절반에 불과하다고 말했다

This year is shaping up to be a big one for self-driving trucks. In addition to Aurora’s plan to deploy hundreds of autonomous big rigs and Waabi expanding into robotaxis, you’ve also got Kodiak AI aiming to launch its own fully driverless long-haul freight operation by the end of 2026. While robotaxis may still win all the headlines, driverless trucks are making their own progress, slowly but surely. Kodiak CEO says making trucks drive themselves is only half the battle The company plans to launch fully driverless long-haul freight operations by the end of 2026. The company plans to launch fully driverless long-haul freight operations by the end of 2026. But in a recent interview, Kodiak AI CEO Don Burnette said that deploying autonomous trucks is really only half the battle. He said that while most of his competitors are fussing over details like AI, perception, and mileage milestones, Kodiak is planning for the reality of operating a business. That includes answering important questions like who owns the trucks, how much uptime will be required, and what ultimately gets shipped. “So you start to very quickly realize that it’s not just does the truck operate safely on the road … that’s table stakes,” Burnette said. “What really matters to customers is how efficiently and effectively can I get that truck into and out of my operation … and everything in between. And nobody talks about this either.” Kodiak AI (formerly Kodiak Robotics) was founded in 2018 by Burnette, a veteran of Google’s self-driving car project (now called Waymo), and Paz Eshel. The company is developing self-driving trucks for highway and industrial uses, as well as the defense industry. In 2025, the company’s trucks began making driverless deliveries for Atlas Energy Solutions in the oil-rich Permian Basin of West Texas and eastern New Mexico, and now operates 20 driverless trucks there. Kodiak AI went public through a reverse SPAC merger in September 2025. And now the company is finally ready for the open road. Burnette said that Kodiak operates across multiple verticals, with a emphasis on industrial and off-road trucking, which he views as a major opportunity compared to traditional on-road autonomy. He describes these environments as “unstructured,” because they’re more complex and unpredictable. In this way, he says his trucks have become better prepared for more “structured” environments, like highways. “We plan to pull the driver by the end of the year,” Burnette said. “Remember the product isn’t valuable unless it’s driverless.” But first, Kodiak needs to complete its safety case. This includes extensive data collection, driving virtually in a simulated world, and creating a detailed plan for risk mitigation. Burnette said his team’s background at Waymo helped influence its rigorous approach to safety. Kodiak is also taking a different approach to its business model than some other companies. Unlike competitors who expected OEMs to deliver autonomous-ready trucks, Burnette said that Kodiak is developing an aftermarket solution in partnership with companies like Roush Industries and Bosch, which allows them to produce fully compliant, automotive-grade trucks and scale more effectively once the technology is ready. As such, the 20 trucks that Kodiak has delivered so far are owned and operated by its customers, not Kodiak itself. This is where Kodiak differs from its peers, and even Waymo. Burnette says that when the customer owns the vehicle, they care about key metrics like utilization, uptime, maintenance, and revenue at all times. This creates a much higher bar for reliability and operational performance, he argues. “When a customer owns the vehicle, it has to work,” Burnette said. “Customers are going to expect the truck to work all the time. So you’ve got to hit that bar before you can really sell a truck to the customer.” When the autonomous vehicle developer owns the truck, rather than the customer, they can stage-manage their deployments without worrying about real-world functionality. “Those are their own trucks,” he says of his rivals. “So if they only work one day of the week, or they only work for like five hours of one day, nobody cares … They’re still driverless and they still get to claim victory. But that would never work with a customer. Like, that’s not a real product.” Burnette can be blunt when talking about his competitors. When asked whether Kodiak would expand its product portfolio to include robotaxis, like Waabi recently did, he responds, “Did Wabi have a product before they existed?” He thinks Kodiak is ahead of its rivals in terms of real-world deployment and operational rigor. He suggests many autonomous vehicle companies emphasize impressive technology and visuals, but haven’t crossed the harder threshold of delivering a usable, customer-owned product: “They produce great videos … they’ve got snazzy visuals,” he says of his competitors. He maintains that most companies haven’t solved what he calls the