BMW’s Silicon Valley-based venture capital arm is investing in Kodiak Robotics, a company that develops autonomous trucking technology.
Kodiak will use the funds to build out a safety case for its self-driving tech stack so it can more quickly commercialize. It will also work on hiring fresh talent and expanding its truck fleet, with a stated goal of at least doubling the number of vehicles it operates each year. The startup currently has 10 trucks in rotation between its commercial route in Texas and its test pilot in Mountain View, California.
The terms of the deal were not disclosed. BMW i Ventures usually invests in companies that can provide solutions to BMW’s current and future business, but Kodiak's CEO and co-founder Don Burnette told TechCrunch that BMW’s investment was purely financial and not strategic, meaning there is currently no technical partnership between the two.
This new investment comes just a week after tire-maker Bridgestone announced a minority stake in Kodiak. The financials behind that deal were not revealed either. To date, Kodiak has publicly announced $40 million in funding from its Series A, and Burnette says the startup has had several additional investments since then.
Burnette also shared the company's plans to achieve driverless operations at scale for less than 10% of what Waymo has publicly raised to date – $5.7 billion – and less than 25% of TuSimple's total existing fundraise – about $1.94 billion, including the money it raised through its IPO. That leaves us with a number roughly around $500 million.
“That’s the total amount of money that we think we need to get to driverless, and that’s because we think we’ve been a much more efficient company up until this point, and we will continue to be much more efficient going forward,” said Burnette.
The BMW i Ventures funding will eventually make up part of Kodiak’s Series B. With this latest investment, the company isn’t trying to further develop its self-driving capabilities or features, but rather it wants to build out its safety case and prove that its system can handle the road with no driver on board, says Burnette.
“We are building toward a Level 4 autonomy system, but we still have a driver in the seat that’s actually monitoring our system at all times,” said Burnette. “Today, we are technically a Level 2 system, which is true for just about everybody else out there.”
Vehicles with a human driver supervising operations such as steering, brake and acceleration support, as well as things like lane centering and adaptive cruise control fit under the Society of Automobile Engineer’s (SAE) definition of Level 2 autonomy. Level 4 means the vehicle can handle all aspects of driving in certain conditions without human intervention.
Kodiak says it’s made progress. In January it announced its Kodiak Driver achieved “disengagement-free deliveries” (meaning the autonomous system didn't have to be switched off for safety reasons) during a commercial route from Dallas to Houston. The company has been running this route out of its Dallas testing and operations facility for two years, and says it’s now achieved a level of maturity where the system can handle anything the highway throws at it.
“We’re doing really complex and advanced maneuvers, not just handling obvious things like merges and cut-ins and heavy traffic, but also more nuanced challenges like identifying vehicles that are pulled over on the side of the road,” said Burnette. “Our system can automatically identify that and then slow down as required by law, or nudge away from that object to give it more space. It can also consider making a lane change if a lane is available, a way to give even more clearance to the stalled vehicle on the side, and this is exactly how humans drive.”
To get to the point where Kodiak can prove its tech is actually safer than a human driver, and thus suitable for operating commercially at scale, the startup has to build up its total miles driven in simulated environments, structured testing on a private closed track, and in real-world driving.
Burnette says Kodiak is the only company that doesn’t designate one type of sensor as "primary," and rather takes a comprehensive approach, meaning it’s not a lidar-first or vision-first company. Tesla’s head of AI Andrej Karpathy recently revealed the company’s new supercomputer that takes a vision-only approach, but Burnette fundamentally disagrees with that method.
“We believe that each of these different modalities have strengths and weaknesses, and our objective is to take advantage of those strengths and cover the weaknesses with other modalities, and so we've created a sensor fusion algorithm that allows us to consider which sensors are advantageous in the moments where they give us the most usable information,” said Burnette.
Kodiak doesn’t use HD mapping either, so its trucks see in real time on the road, which allows the Kodiak Driver to be flexible when it comes to changing road conditions or environments. The system is trained using data collected by Kodiak’s trucks, as well as on scenarios devised by its engineers, and that data is auto-labeled using Scale AI, which is one of the ways Kodiak is able to keep down costs, says Burnette.
Kodiak’s team hails from Google’s original self-driving team, Uber, Lyft and other notable tech companies. Burnette says BMW i Ventures’ investment in the company came after a thorough vetting process in which the firm sent over their autonomous driving experts and dug into the team’s expertise and tech.