Earlier this year, a team of engineers flew to Detroit on a secret mission to make Jeep Grand Cherokees drive themselves.
A few weeks later, prototypes were convincing enough for Jeep owner Stellantis to take investors on test rides around Motor City's suburbs and commit to rolling out the technology across America.
The self-driving know-how was developed by Wayve, a British startup that is emerging as an unlikely front-runner in a hard-fought race with U.S. behemoths Tesla and Google owner Alphabet to bring autonomous vehicles to the masses.
Stellantis said last month it would use Wayve's "AI driver" across its brands -- which also include Chrysler, Dodge and Ram -- from 2028. The deal followed a similar agreement with Nissan that envisaged a potentially earlier launch in Japan.
The technology promises hands-free navigation on both highways and urban roads as long as the human driver stays attentive and ready to take over as a backup. Tesla pioneered this approach with the feature it now sells as "Full Self-Driving (Supervised)."
But while Tesla bakes that technology into its own vehicles, Wayve thinks selling its systems as an off-the-shelf solution to other automakers could be where the real money is in self-driving.
"Not everyone wants to buy a Tesla," said Wayve co-founder Alex Kendall. "Our opportunity is to bring this technology to every other automaker."
On a test drive in one of the Grand Cherokee prototypes last month, a Stellantis engineer took his hands off the wheel and the vehicle navigated through stop signs and traffic lights around the automaker's sprawling campus in Auburn Hills, Mich. The ride was smooth, save for a split second when the Grand Cherokee had trouble finding its lane amid faded road markings.
A look in the vehicle's trunk revealed how the automaker jury-rigged a run-of-the-mill sport-utility vehicle to run with Wayve's technology. Dozens of wires snaked around three electronic boxes, one of them connected to an old-school car battery for supplemental power.
Stellantis is sourcing parts that eliminate all those wires and boxes for the vehicles it plans to sell commercially.
Like Tesla in Texas, Wayve also has more futuristic ambitions to do without the human driver altogether. The startup is launching a robotaxi trial in London with Uber this summer ahead of a planned global rollout that will include U.S. cities. The vehicles will initially require a safety driver.
Wayve was founded in 2017 by Kendall, a now 33-year-old New Zealander, and a fellow robotics Ph.D. at the University of Cambridge. Their work on machine learning and intelligence had convinced them that AI could solve the problems of self-driving systems being developed at the time.
In the early years of vehicle autonomy, Tesla, Alphabet's robotaxi unit Waymo and others meticulously coded driving instructions into software. The systems attempted to classify objects and scenarios -- is the thing 20 feet ahead a plastic bag, or a toddler? -- and apply human-written rules for how the vehicle should respond.
But progress was slow, largely because it was hard to demonstrate that such systems could be trusted to handle the near-infinite number of unusual situations vehicles might encounter.
In late 2023, Tesla pivoted to an AI-based system similar to Wayve's, where driving decisions are made by a model -- akin to those behind chatbots such as ChatGPT -- trained on reams of driving videos.
Around that time, Wayve started to attract wider attention. In 2024, it raised $1.05 billion in a funding round led by SoftBank and supported by Microsoft, Nvidia and Uber. This year it got a further $1.5 billion at an $8.6 billion valuation, including from Stellantis, Mercedes-Benz and Nissan and chip makers AMD, Arm and Qualcomm.
"I've always believed that intelligence would be the way that robots work in the future," said Kendall. "But the market's conviction around that, I think, has come together really in the last couple of years."
The seed of Stellantis's bet on Wayve was sown on a test ride in London, when the technology wowed the carmaker's chief technology officer, Ned Curic, with its humanlike way of edging into flowing traffic.
When Curic later took responsibility for Stellantis's automated-driving program in early 2025, he shelved the company's internal development, which followed the traditional rules-based approach, and started scouting for AI-savvy partners.
Curic challenged Kendall to show his technology would work on an unfamiliar Jeep in unfamiliar Michigan.
It performed "immensely well," Curic said, even in a blizzard and after his team asked Wayve to shift sensors from an unsightly roof rack to the wing mirrors.
Still, the "end-to-end" AI approach championed by Wayve -- where a single model takes inputs from sensors and delivers driving decisions as outputs -- remains controversial, particularly when it comes to removing human backup drivers.
Unlike with chatbots, verification and compliance are key considerations in the safety-first automotive industry. To their critics, end-to-end AI models are "black boxes" that lack the transparency required to comply with strict regulations.
Robotaxi pioneers Waymo in the U.S. and Baidu's Apollo Go in China, which already operate without backup drivers, use hybrid approaches that blend AI models with high-definition maps and traditional rules that function as guardrails.
Supporters of AI systems tout their greater capacity to generalize from one driving situation to another, as human drivers can, making them easier to roll out at scale than hybrid systems.
Wayve's approach doesn't require elaborate mapping of streets before it can drive around them. Last year, the company went on a tour of 506 cities across the world to prove the point, albeit with a safety driver behind the wheel.
Kendall said Wayve had spent years working out how to make its AI driver compliant and that its solution had been accepted by the safety experts at Stellantis, Nissan and other automakers with which it is in talks.
A similar debate focuses on the sensors required to run a self-driving vehicle. Wayve, like Tesla, says its technology can do without the expensive laser-based lidar units that most of their peers rely on. Its London test vehicles feature only cameras and radar.
But traditional automakers are more conservative when it comes to deploying the technology at commercial scale. Stellantis and Nissan are both developing robotaxis using Wayve's AI driver that include lidar in the sensor suite.
While Stellantis used six cameras for the jury-rigged Grand Cherokees, the vehicles available for purchase in 2028 will sport 10 cameras and five radars, giving Wayve's AI more data to rely on.
"We feel it's best for us to make sure the system is consistent, safe and seeing everything," said Doug Wellman, who leads assisted driving at the automaker's U.S. headquarters.
The flexibility of Wayve's AI-first approach could attract more traditional automakers looking to share the investment burden of keeping up with Silicon Valley's advances in vehicle autonomy. The startup says it is in talks with most of the world's top-10 players.
Kendall said being based in London, far from the autonomous-vehicle mainstream, helped Wayve stick to its technological guns.
"I want to be building something contrarian," he said. "We've completely gone against the grain of the Silicon Valley bubble for self-driving and I think we might have been swallowed up if we were starting there."
Write to Stephen Wilmot at stephen.wilmot@wsj.com and Christopher Otts at christopher.otts@wsj.com
(END) Dow Jones Newswires
June 28, 2026 05:30 ET (09:30 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments