Nearly two years ago, Motional was at a crossroads for self-driving cars.
The company, which was born out of a $4 billion joint venture between Hyundai Motor Group and Aptiv, has already missed a deadline to launch an unmanned robotaxi service with partner Lyft. The company lost Aptiv, one of its financial backers, prompting Hyundai to step up an additional $1 billion investment to keep the business going. Several rounds of layoffs, including a 40% layoff in May 2024, have reduced the company’s workforce from a peak of about 1,400 to fewer than 600. Meanwhile, advances in AI were changing the way engineers develop technology.
Motional had to evolve or die. I paused everything and chose option 1.
Motional told TechCrunch that it is taking an AI-first approach to its self-driving system and rebooting its robotaxi program, pledging to launch commercial driverless service in Las Vegas by the end of 2026. The company has already opened up its robotaxi service to employees, driven by human safety operators. Later this year, the company plans to roll out the service to the general public with an unnamed ride-hailing partner. (Motional has existing relationships with Lyft and Uber.) By the end of the year, human safety operators will be phased out from robotaxis and true commercial driverless services will begin, the company said.
“We saw that with all the advancements that are happening within AI, there is tremendous potential. We also saw that while we have secure unmanned systems, there is a gap in reaching affordable solutions that are universal and scalable globally,” Laura Major, Motional’s president and CEO, said in a presentation at the company’s Las Vegas facility. “So we have made the difficult decision to pause commercial operations and slow down in the short term.”
This meant moving from a traditional robotics approach to an AI-based model-based approach. AI was essential to Motional. Motional’s self-driving system used individual machine learning models to handle perception, tracking, and semantic reasoning. However, more rule-based programs were also used for other operations within the software stack. The individual ML models then formed a complex web of software, Major said.
Meanwhile, AI models originally built for languages have begun to be applied to robots and other physical AI systems, including the development of self-driving cars. This transformer architecture enabled the construction of large and complex AI models, ultimately leading to the emergence of ChatGPT and the proliferation of its usage.
tech crunch event
san francisco
|
October 13-15, 2026
Motional looked for ways to combine these smaller models into a single backbone, enabling an end-to-end architecture. Major explained that Motional also maintains a smaller model for developers, which gives Motional the best of both worlds.
“This is very important in two ways: One, it makes it easier to generalize to new cities, new environments, and new scenarios,” she said. “The other thing is to do this in a cost-optimized way. For example, the next city you go to might have different traffic lights, but you don’t have to redevelop or reanalyze them. You just collect the data and train the model so it can operate safely in that new city.”
TechCrunch took a 30-minute self-drive around Las Vegas to see Motional’s new approach firsthand. A single demonstration cannot accurately evaluate an autonomous driving system. However, you can pinpoint weaknesses and differences from previous iterations and assess your progress.
It was progress to see my Hyundai Ioniq 5 drive autonomously from Las Vegas Boulevard to the Aria Hotel pick-up area. These busy areas are notorious in La Vegas, and my experience was no different, as the self-driving car slowly circled a stopped taxi, dropped off passengers, changed lanes and back again, passing dozens of people, giant flower pots, and cars along the way.
Motional previously worked with partner Lyft to operate a ride-hailing service in Las Vegas using vehicles that handle some rides autonomously. Parking lots, hotel valet parking, and app pickup and drop-off areas were not included in these operations. A human safety operator is always at the wheel, taking over navigation through crowded pick-up and drop-off areas in parking lots and hotel lobbies.
There is still progress to be made. The graphics that riders will see inside the vehicle are still being developed. And while the vehicle never disengaged during my demo ride, meaning a human safety operator would take over, it moved slowly around a double-parked Amazon delivery van.
Still, Major insists Motional is on the right path to safe and cost-effective deployment. And Hyundai, its majority owner, will be involved in it for the long term, she said.
“I think the real long-term vision is to, you know, apply Level 4 to people’s personal cars,” Major said, referring to the term that means the system handles all driving without any expectation of human intervention. “Robo-taxi, it’s the primary destination and it’s going to have a huge impact. But ultimately, every OEM will want to integrate it into their vehicles.”
