Less than two years ago, Motional found itself facing an existential moment in the autonomous vehicle race.
The self-driving company—formed through a $4 billion joint venture between Hyundai Motor Group and Aptiv—had fallen behind on its plans to roll out a fully driverless robotaxi service with Lyft. Aptiv exited as a financial backer, forcing Hyundai to inject an additional $1 billion to stabilize the business. A series of layoffs followed, including a major restructuring in May 2024 that cut roughly 40% of staff. From a peak workforce of about 1,400 employees, Motional shrank to fewer than 600. At the same time, rapid advances in artificial intelligence were reshaping how autonomous driving systems were being built.
The message was clear: adapt quickly or risk becoming irrelevant.
Motional chose to reset.
The company confirmed to TechCrunch that it has overhauled its robotaxi roadmap, rebuilding its self-driving system around an AI-first architecture and committing to launch a commercial, fully driverless service in Las Vegas by the end of 2026. As part of that relaunch, Motional has already begun operating robotaxis with human safety drivers for employees. A public version of that service—still with a safety operator—is expected to follow later this year through an unnamed ride-hailing partner. Motional already works with both Lyft and Uber.
By the close of the year, the company plans to remove the safety driver entirely, marking the start of its true commercial driverless operations.
During a presentation at Motional’s Las Vegas facility, President and CEO Laura Major explained that the reset was driven by both opportunity and necessity.
“AI was advancing incredibly fast,” she said. “We had a safe driverless system, but there was still a gap when it came to affordability and the ability to scale globally. We made the tough call to pause commercial efforts in the short term so we could move faster in the long term.”
From Traditional Robotics to AI Foundation Models
The reboot required Motional to move beyond its earlier robotics-led approach. While AI was already embedded in the system—powering perception, tracking, and semantic understanding—other parts of the software relied heavily on rules-based programming. According to Major, that mix resulted in a sprawling and complex stack made up of many separate machine learning models.
Meanwhile, breakthroughs in AI—particularly transformer-based architectures originally developed for language—were being applied to robotics and autonomous driving. Those advances paved the way for large-scale foundation models and, ultimately, the explosion of tools like ChatGPT.
Motional began rethinking how its technology could take advantage of these developments. Engineers worked to unify smaller, task-specific models into a single end-to-end backbone while retaining the individual models for developer-level access. Major described the result as a hybrid approach that combines flexibility with scalability.
“This is essential for two reasons,” she said. “First, it makes it much easier to generalize to new cities, environments, and scenarios. Second, it allows us to do that in a cost-efficient way. If traffic signals change from city to city, you don’t need to rebuild everything—you collect data, retrain, and the system can operate safely in that new location.”
Early Signs of Progress on the Road
TechCrunch was given a chance to experience Motional’s updated system during a 30-minute autonomous ride through Las Vegas. While a single demonstration can’t fully validate a self-driving platform, it can reveal how far the technology has come—and where it still struggles.
The ride showed meaningful improvement. The Hyundai Ioniq 5 navigated away from Las Vegas Boulevard and into the busy pickup and drop-off zone at the Aria Hotel, an area notorious for congestion, pedestrians, and unpredictable vehicle behavior. The autonomous vehicle carefully maneuvered around a stopped taxi, shifted lanes multiple times, and passed through crowds of people, cars, and large obstacles like planters.
This represented a notable change from Motional’s earlier operations in Las Vegas, which were limited to partial autonomy. Previously, human drivers were required to take control in parking lots and hotel pickup areas—locations that were excluded from autonomous operation.
There are still visible gaps. The in-vehicle rider interface remains unfinished, and while no disengagements occurred during the demo, the vehicle was cautious when navigating around a double-parked Amazon delivery van, taking its time to pass.
Even so, Major maintains that Motional is now on a viable path toward safe, scalable, and economically sustainable deployment. Hyundai, she added, remains committed to supporting the effort over the long term.
“The bigger picture isn’t just robotaxis,” Major said. “That’s the first major milestone and an important one. But the ultimate goal is Level 4 autonomy in personal vehicles—where the system handles all driving without human intervention. I think that’s where every automaker would like to end up.”
thetechtrident.com 