Motional Bets Big on AI ‘Brain’ as It Reboots Robotaxi Strategy for 2026
- Hyundai- and Aptiv-backed Motional has scrapped its old software stack and is pivoting to an AI “foundation model” for driverless taxis.
- The company is targeting fully driverless service in Las Vegas by late 2026, even after layoffs and financial pressure.
Motional is throwing out much of its old playbook and betting that a single, powerful AI “brain” can finally turn robotaxis into a real business.
The Boston- and Las Vegas-based autonomous driving company, backed by Hyundai and Aptiv, has launched a major reset that puts a transformer-style, end-to-end AI model at the core of its self-driving system. After missing earlier timelines and cutting deeply, the company is now aiming for commercial driverless operations by late 2026.
“We made the very hard decision to slow down so that we could actually speed up,” Motional CEO Laura Major reportedly told TechCrunch in an interview published January 11, 2026.
Major framed the shift as a response to both rapid progress in AI and tougher funding conditions across the autonomous-vehicle sector.
“This is about getting to something that’s not just safe,” she said. “It has to be affordable and scalable.”
From Modular Stack to AI ‘Foundation Model’
For years, Motional—like most of the industry—built its self-driving system as a modular stack. One set of machine-learning models tried to understand the scene. Another tried to predict what drivers and pedestrians would do. A third planned the car’s next moves, all wrapped in layers of hand-written rules.
Inside the company, that approach increasingly looked like a dead end.
Engineers described the old system as complex and brittle, according to people familiar with the matter. Every new city or novel edge case meant weeks or months of re-engineering. That might fly in a research lab. It is punishingly expensive for a business that hopes to operate across dozens of cities.
The reboot centers on a transformer-based backbone model, the same class of neural network that underpins today’s large language models. Instead of stitching together a series of separate components, Motional now feeds sensor data into a single, unified network that is trained to see, predict and plan in one continuous loop.
In theory, that end-to-end design can adapt more quickly. Rather than rewriting code for each market, Motional can gather local driving data, retrain the model and redeploy.
“Foundational AI lets us scale with data, not armies of engineers,” one Motional engineer told a person briefed on internal discussions.
A High-Stakes Reset After Missed Deadlines and Layoffs
The strategy shift follows a bruising 18 months for the company.
Motional had previously promised a driverless launch with Lyft, then missed the deadline without much public explanation. In May 2024, it cut roughly 40% of its staff, shrinking from about 1,400 employees to under 600. Aptiv, a long-time backer, pulled back on funding as losses mounted, and Hyundai stepped in with a new $1 billion lifeline to keep the effort going.
Inside Motional, the mood shifted from urgency to something closer to existential worry.
“People were asking, ‘Are we actually going to make it to commercialization, or just end up as another AV footnote?’” recalled one former employee, who asked not to be named because they were not authorized to speak publicly.
Major has tried to present the cuts and delays as a painful but necessary reset. Company insiders say Motional effectively paused near-term commercial plans to rebuild its tech stack around the new AI architecture.
It is a substantial gamble. A cleaner, more flexible system could be an advantage in the long run, but it also means Motional will spend the next two years racing—again—to prove it belongs in a sector that has already consumed billions with little profit.
Las Vegas as the Testbed—and First Real Market
For now, Las Vegas is Motional’s main proving ground and its first true target market.
The company currently runs an internal robotaxi service there using Hyundai Ioniq 5 electric vehicles. Employees can hail rides, but a safety operator sits in the driver’s seat, ready to take over.
The new roadmap lays out three major steps:
- Continue internal operations in Las Vegas while training and refining the AI model.
- Open the service to the public in mid-2025 through a ride-hailing partner—likely an existing partner such as Lyft or Uber—still with a human safety driver behind the wheel.
- By the end of 2026, begin removing safety operators and operating fully driverless rides.
Las Vegas is hardly a gentle training ground. The Strip runs late. Tourists spill into crosswalks. Hotel valets and ride-shares create chaotic loading zones. Big events can snarl traffic with little warning.
That complexity is also the point. An AI-first system thrives on variation and volume of data, and Las Vegas offers both in abundance.
“If you can survive a Friday night near the Strip, you’ve earned your stripes,” joked a city transportation official, who said regulators are tracking Motional’s work but declined to discuss specific permit timelines.
Safety Questions Linger Over a Black-Box Brain
The biggest and hardest question is also the simplest: how safe is an end-to-end AI driver that even its creators may find difficult to fully explain?
Motional says its architecture keeps some modular elements so engineers can probe and stress-test specific behaviors, while still leaning on a core model to learn subtle patterns that hand-coded rules might miss.
But the company has shared little detail about its formal safety case. It has not publicly released hard metrics such as disengagement rates, crash statistics relative to human drivers, or clear thresholds for simulated and real-world miles before pulling safety operators from vehicles.
It also has yet to spell out what happens when the unified model fails. Are there independent backup systems that can override a bad decision, or is everything riding on that central network? Researchers have warned for years that large neural networks, however powerful, can be opaque and difficult to certify in safety-critical roles like driving.
Regulators are likely to push on those issues, especially after a run of high-profile problems at other autonomous-vehicle companies. Cruise, Uber’s former self-driving unit and Tesla’s driver-assistance systems have all faced investigations and public scrutiny after crashes involving automated features.
A safety advocate who has briefed federal officials called Motional’s approach “ambitious but unproven,” adding, “If you’re going to replace a modular system with a giant black box, you’d better have extraordinary evidence that it fails less often and fails more gracefully.”
Racing Waymo, Cruise—and the Clock
Motional’s reset lands in the middle of a reshuffling competitive landscape.
Alphabet’s Waymo continues to run robotaxi services in Phoenix and parts of California. Cruise, backed by General Motors, is trying to rebuild trust and regain regulatory approval after pulling its fleet from U.S. streets following safety incidents and investigations. Tesla keeps expanding its “Full Self-Driving” features through over-the-air software updates and has signaled its own interest in more end-to-end AI.
Motional is operating in fewer markets and with tighter financial constraints than some of its rivals. Its clearest advantage is Hyundai. A major automaker as a strategic partner means Motional’s technology can be designed directly into factory-built vehicles and, over time, potentially into consumer cars.
Major has been open about that longer ambition. Robotaxis are “stop number one,” she told TechCrunch, suggesting the real prize is Level 4 autonomy in personal vehicles sold at scale.
Analysts are wary. They have heard this kind of talk before from multiple players in the space.
“Everyone’s talking about foundation models and end-to-end autonomy,” said one autonomous-vehicle consultant. “The question is whether Motional is early, late, or just bold enough to survive when others can’t.”
The Road to 2026
For riders, an AI-first robotaxi may not feel dramatically different at the beginning. The changes could be subtle: steadier handoffs, smoother lane changes, fewer awkward pauses when a delivery van blocks the street.
Behind the scenes, though, Motional is betting the company on a very different kind of machine driver.
If the new model learns quickly, transfers its skills across cities and clears skeptical regulators, Motional could reemerge as a serious contender in a field many had written off as overhyped. If it falters, this reboot may go down as a final, expensive roll of the dice.
By committing to a single AI brain and a 2026 deadline, Motional is testing more than its own survival. After a decade of missed timelines and diminished expectations, the company is forcing a larger question on the industry and its backers: is this the leap that finally gets robotaxis out of pilot programs and into everyday life, or the point at which investors and the public decide the promise of fully autonomous taxis has been stretched too far?
