The fleets are grounded. The apps are stagnant. The once-buzzing "Robotaxi Hubs" in downtown Phoenix are now just overpriced parking lots with high-voltage chargers that no one uses.
This is the story of how we hit the Digital Wall. It’s not just a tale of sensor failure or budget cuts; it’s a story of the ghost we chased—the hubris of believing we could replace human instinct with a trillion lines of code. We wanted a servant that would never tire; we built a calculator that didn't know how to look a pedestrian in the eye.
A deeply human look at the autonomous vehicle industry's 2025 inflection point—the empty charging hubs, the laid-off safety drivers, the coning protests, and the realization that we expected a servant and got a very expensive calculator."
tags: autonomousvehicles, ai, transportation, urbanplanning.
Series: Autonomous Systems Deep Dives
The State of Play: 2026
Commercial Presence: Phoenix, SF, LA, Austin, Dallas, Houston, Atlanta; testing in Tokyo and London.
Reality: No longer operating a commercial robotaxi fleet.
This is the story of how we got here—and where we go next. But more than that, it's the story of the ghost we chased: the expectation that we could replace human instinct with code, that we could build a servant that would never tire, never err, never disappoint.
We were wrong. And the ghost is still out there, haunting the empty lots where the robotaxis used to charge.
Part I: The Dream That Drove Us
The Promise of 2025
Five years ago, we were all intoxicated. I'll admit it—I wrote some of those breathless articles myself. "The End of Car Ownership." "Your Morning Commute, Reimagined." "Why 2025 Is the Year Everything Changes."
Every major automaker, every tech giant, every ambitious startup had declared 2025 as the year of full autonomy. Robotaxis would dominate city streets. Car ownership would become obsolete. The morning commute would be spent working, sleeping, or streaming—not staring at brake lights.
We believed it because we wanted to believe it. The future was supposed to be clean, efficient, and effortless. The future was supposed to arrive on schedule.
The projections were intoxicating:
- 100 million autonomous vehicles on roads by 2030 (Intel)
- $800 billion in annual revenue from autonomous mobility services (UBS)
- 90% reduction in traffic fatalities (Various industry claims)
- Zero human intervention required (Literally every autonomy presentation)
VC funding flowed like water. Over $100 billion was invested in autonomous vehicle technology between 2015 and 2023. Cities rewrote zoning codes for autonomous-ready infrastructure. Regulators raced to create frameworks for a driverless future.
We built cathedrals to a god that hadn't yet arrived.
The Geography of Ambition
The commercial footprint told the story of American ambition, a map of places we thought would be transformed:
Phoenix emerged as the proving ground—wide streets, predictable weather, and welcoming regulators. Waymo launched the world's first fully driverless ride-hailing service here in 2020. Cruise followed. The Valley of the Sun would be the valley of autonomy.
San Francisco represented the ultimate challenge and prize. Narrow streets, unpredictable behavior, dense fog, and aggressive regulators. If you could make it here, you could make it anywhere. Cruise and WThe last giant standing, cautiously expanding to Dallas and Orlando.
Los Angeles offered sprawl and scale—the traffic nightmare that autonomy would solve. The 405 at rush hour, the chaos of LAX, the maze of surface streets connecting a hundred neighborhoods.
Austin became the new frontier. Texas welcoming regulations, South by Southwest as a showcase, and a growing tech ecosystem. Still a ghost of its former self after the 2024-25 safety retreat.
Dallas-Fort Worth represented the Metroplex challenge—sprawling, high-speed, with unpredictable weather patterns.
Houston brought humidity, hurricanes, and some of the most aggressive drivers in America.
Atlanta was the Southeast beachhead—Peachtree chaos, interstate spaghetti, and southern hospitality meeting northern engineering.
Tokyo testing meant navigating the world's most complex urban environment—dense, polite, and technologically sophisticated.
London testing meant roundabouts, narrow historic streets, and left-side driving—the ultimate cognitive challenge for systems trained on American roads.
Seven cities in commercial operation. Two global capitals in testing. The infrastructure of a future being built. And now, most of it sits quiet.
The Philosophy of Failure
Instead of listing every company that tried and stumbled, let's group them by how they failed. Because the failure modes tell us more than the corporate histories.
The Detroit Muscle: Cruise (GM) and Argo AI (Ford+VW) represented the old guard's attempt to buy innovation. They poured billions into autonomy, expecting to bolt it onto their manufacturing might. Cruise peaked at a $30 billion valuation. Argo absorbed $3.6 billion before being shuttered in 2022. Their failure was believing that capital could compress time—that money could buy the decades of learning that human drivers accumulate effortlessly.
The Pure Tech: Waymo (Alphabet) and Zoox (Amazon) took the patient approach—build the perfect system, then deploy. Waymo spent 15 years and billions on the most sophisticated autonomy stack in existence. Zoox built a vehicle from scratch with no steering wheel, no pedals. Their failure was believing that technical perfection would win—that if they built it, riders would come. But technical perfection doesn't matter if the economics don't work and the public doesn't trust you.
The Visionary Gambler: Tesla Just "flipped the switch" in Austin, but the fleet is small and the rain still stops the music, leveraging its million-vehicle fleet to gather data. Promised full self-driving "next year" every year since 2016. Their failure was believing that scale alone would solve complexity—that if you gathered enough data, edge cases would eventually disappear. But the long tail is infinite; data alone doesn't tame it.
The Pragmatist: Mobileye (Intel) took the supplier route, selling chips and software to everyone else. Their failure was less dramatic—they're still profitable, still relevant. But they bet that autonomy would arrive incrementally, through driver-assist features that gradually take over. The incremental path turned out to be the only path, but it's not the revolution anyone promised.
Each philosophy failed differently, but they all hit the same wall.
The Patchwork Problem
The regulatory landscape wasn't a foundation—it was a minefield. We called it "The Patchwork Problem."
A vehicle could be perfectly legal in Phoenix, where Governor Doug Ducey's executive orders welcomed autonomy with open arms. But cross into California, and suddenly that same vehicle was effectively a felon, operating without the proper permits from the CPUC.
Texas said "come on down" with no new laws, just permission. Nevada had been first, back in 2011, but never quite became the hub everyone expected. California regulated slowly, carefully, painfully—each permit a hard-won battle.
And internationally? UNECE regulations in Europe created one framework. Japan had its own. China built a national strategy that made American efforts look like a garage project.
The industry begged for federal standards, for a single set of rules that would let them build once and deploy everywhere. Congress did nothing. So companies built for Phoenix and hoped to figure out the rest later.
Later never came.
Part II: The Digital Wall
The First Cracks
2022: Argo AI Shuts Down
The first major signal that something was wrong. Ford and VW had poured billions into Argo AI, expecting a 2021 launch. Instead, Ford CEO Jim Farley announced the company would "absorb" Argo's talent and pivot to L2+ driver-assist rather than L4 autonomy. The reason: "Profitable L4 autonomy is a long way off."
The market didn't panic—Argo was just one player. But those paying attention noted the pattern: two of the world's largest automakers, with unlimited resources and patience, couldn't make the economics work.
2023: Cruise San Francisco Setbacks
San Francisco was supposed to be the showcase. Cruise had hundreds of vehicles operating across the city, including fully driverless operations. Then the incidents started accumulating:
- A Cruise vehicle blocking a fire truck responding to an emergency
- Multiple vehicles stopping in intersections, causing gridlock
- A car driving into wet concrete
- A collision with a bus
Each incident made headlines. Each incident eroded public confidence. Each incident caught regulator attention.
But the worst was yet to come.
October 2023: The Turning Point
A pedestrian in San Francisco was struck by a human-driven vehicle and thrown into the path of a Cruise robotaxi. The Cruise vehicle braked but couldn't avoid contact. Then, in a sequence that would haunt the industry, the Cruise vehicle attempted to pull over—dragging the pedestrian 20 feet.
The California DMV suspended Cruise's deployment permit immediately. The CPUC followed. Cruise recalled its entire fleet. General Motors took a $500 million write-down.
The dream hit the digital wall.
The Nature of the Wall
What made autonomy so much harder than anyone predicted? The wall had multiple layers. But let's talk about them differently—not as engineering problems, but as human ones.
Layer 1: We Asked the Software to Read Vibes
A human driver approaching an intersection with a group of teenagers doesn't just see people—they read the vibe. Are they waiting for a bus? About to jaywalk? Distracted by their phones? About to chase a ball into the street? That judgment happens in milliseconds, based on thousands of hours of social conditioning.
We asked software to do the same thing. We gave it cameras and LiDAR and asked it to understand human intent.
It couldn't. It still can't.
Construction Zones: Every construction zone is unique. Cones placed unpredictably. Workers waving flags. Lane markings painted over. Temporary signals. Human drivers navigate by reading context—the worker with the stop/slow sign isn't just an object, they're an authority figure. Autonomous systems see chaos.
Emergency Vehicles: Sirens echo off buildings, making direction ambiguous. Police officers wave traffic through red lights. Fire trucks block intersections. Ambulances need to pass. Humans interpret intent—the officer waving you through is telling you it's safe, even though the light is red. Autonomous systems see a person in the roadway and a traffic signal in conflict, and they freeze.
Weather: Rain creates reflections that confuse cameras. Snow covers lane markings. Fog scatters LiDAR. Ice changes traction unpredictably. Phoenix was chosen for its predictable weather—but the real world isn't Phoenix.
Layer 2: The Long Tail Is Infinite
Engineers call them "edge cases"—situations that occur rarely but require unique handling. In autonomy, edge cases aren't the exception. They're the rule.
Stationary Vehicles: A stopped car on the highway shoulder is obvious to humans. To an autonomous system, it's an object that could be parked, broken down, or about to re-enter traffic. Classification uncertainty matters. The system doesn't know what it doesn't know.
Debris in Road: A mattress, a tire tread, a piece of lumber. Humans recognize and avoid. Autonomous systems may not classify correctly—or may swerve dangerously.
Cyclist Hand Signals: Not always used, not always consistent, but legally meaningful. Humans interpret. Systems may not.
Law Enforcement Directives: An officer waving traffic through a red light overrides every traffic signal and sensor. Humans understand authority. Autonomous systems see a person in the roadway.
Layer 3: The Simulation Gap
We simulated billions of miles. We proved our systems safe in virtual environments. But simulation is only as good as its models. If you haven't modeled the precise way a pedestrian behaves in the rain at a particular intersection, your simulation miles are worthless.
Some events are so rare they'll never appear in testing, even with millions of miles. The pedestrian who slips and falls. The driver who has a medical emergency. The child chasing a ball into traffic. These are statistically unlikely but must be handled correctly when they occur.
We can't simulate everything. We can't test everything. And yet the real world will throw everything at us.
Layer 4: The Economics Never Worked
The math was always suspect, but we didn't want to look too closely.
The Cost of Sensors: Early autonomous vehicles carried $200,000+ in sensor equipment. LiDAR prices dropped dramatically, but still added $5,000-10,000 per vehicle. For a robotaxi fleet operating 24/7, that's manageable. For consumer vehicles, it's prohibitive.
The Cost of Compute: The computers required for L4 autonomy add another $10,000-20,000 per vehicle. Power consumption affects range. Heat management adds complexity. The economics work for robotaxis but not for personal vehicles.
The Cost of Validation: Proving safety to regulators requires billions of miles of testing—billions of real, not simulated, miles. At current testing rates, that takes decades. No company has that kind of time or money.
The Cost of Teleoperation: Every autonomous vehicle needs remote human monitors for edge cases. Even at one operator per 10 vehicles, that's thousands of operators for a meaningful fleet. Labor costs scale with fleet size—eliminating the economic advantage of removing the driver.
We expected a servant. We got a very expensive, very confused calculator.
Part III: The Reality of the Retreat
The Silence in the Valley of the Sun
In early 2025, if you stood on a corner in Chandler, Arizona, the hum of electric motors was the soundtrack of the future. Waymo vehicles gliding through intersections. Cruise cars waiting patiently at stop signs. The occasional Zoox prototype, looking like a spaceship that forgot to leave.
By December, that soundtrack had been replaced by the familiar rattle of 2012 Honda Civics.
The retreat wasn't a bang; it was a quiet deletion of apps. The "Commercial Presence" in Phoenix, SF, LA, Austin, Dallas, Houston, and Atlanta didn't end because the cars stopped working—they ended because they stopped making sense.
The Ghost Fleets
Drive through a certain industrial park in Mesa, Arizona, and you'll see them. Hundreds of autonomous vehicles, parked in neat rows. Charging cables dangling. Solar panels accumulating dust. These were the backup fleets, the expansion vehicles, the future that never arrived.
They're not going anywhere. No one wants to buy a used robotaxi—too much custom hardware, too many sensors that need calibration, too much uncertainty about software support. So they sit. Ghosts waiting for a resurrection that may never come.
The Empty Hubs
In San Francisco's Dogpatch neighborhood, the Cruise facility sits quiet. The garage doors are down. The charging stations are dark. The only sign of life is the occasional security guard making rounds.
In Austin, the East 6th Street staging area—once buzzing with activity during SXSW—is now just another parking lot.
In London and Tokyo, the test vehicles have been repatriated or mothballed. The international ambitions, the global footprint—reduced to PowerPoint slides in archived investor decks.
The Human Cost
Meet Marcus. He was a safety driver for Cruise in San Francisco. Eight hours a day, five days a week, he sat behind the wheel of a vehicle that was supposed to drive itself. He wasn't supposed to touch anything, just monitor. Just be ready.
"It was the most boring job I've ever had," he told me. "You're watching the road, watching the screen, waiting for something to go wrong. Most days, nothing did. But you couldn't look away. Couldn't check your phone. Couldn't relax. Just sit there, alert, for eight hours."
Marcus was laid off in November 2023, when Cruise paused its operations. He was one of thousands.
Meet Elena. She was a remote operator for a different company, monitoring up to 10 vehicles simultaneously from an office in Dallas. When a vehicle encountered something it couldn't handle—an ambiguous situation, a construction zone, a confused pedestrian—it would signal for help. Elena would review the video, make a decision, and guide the vehicle through.
"It felt like playing the world's most stressful video game," she said. "Except if you lost, someone could get hurt."
Elena kept her job longer than most—her company pivoted to trucking, where remote operators are still needed. But she knows the writing is on the wall. Eventually, they'll automate her too.
Meet David. He was a mapping technician, part of the army of workers who drove every road in every city, creating the high-definition maps that autonomous vehicles need to navigate. When operations paused, David's contract wasn't renewed.
"I spent two years driving the same 50 miles of highway, over and over," he said. "Checking for changes. Updating lane markings. Noting new construction. It was tedious, but it paid well. Now I'm delivering food for DoorDash."
The human cost of the retreat isn't counted in billions of dollars of write-downs. It's counted in thousands of workers who believed they were building the future, only to find themselves building someone else's resume line.
Who Was Actually Riding?
Before the wall hit, data from 2024-2025 pilot programs showed a distinct demographic divide in who actually used these services:
| Demographic Group | Adoption/Usage Rate (2024) | Primary Sentiment |
|---|---|---|
| Early Tech Adopters (Aged 18-34) | 62% | Enthusiastic/Experimental |
| Senior Citizens (Aged 65+ in Phoenix) | 28% | High Trust (Mobility Independence) |
| Urban Commuters (San Francisco) | 41% | Mixed (Frustrated by Gridlock) |
| Black & Hispanic Communities | 19% | Low Trust (Safety & Surveillance concerns) |
The trust gap wasn't technical—it was social. In many Black and Hispanic neighborhoods in Houston and Dallas, the presence of sensor-laden vehicles was often viewed through the lens of increased surveillance rather than a transport revolution.
"I don't want a car recording everything I do," one Houston resident told a local news crew in 2024. "I don't care if it drives itself. Who's watching the watchers?"
The industry never had a good answer for that question. They talked about safety, about efficiency, about the future. They didn't talk about who owned the data, who could access the footage, who decided where the cars went and didn't go.
The social wall was as real as the technical one.
The Coning of San Francisco
Perhaps the most visible symbol of public resistance was the "coning" phenomenon.
In 2023 and 2024, San Francisco residents discovered that placing a single orange traffic cone on the hood of an autonomous vehicle would cause it to freeze. The sensors detected an obstacle, couldn't determine what it was, and entered a protective state—hazards on, transmission in park, waiting for remote assistance.
What started as a prank became a protest. Cones appeared on robotaxis across the city. One night, a single individual was credited with coning 50 vehicles. The videos went viral. The message was clear: we don't want you here.
The industry called it vandalism. The public called it self-defense.
Neither was entirely wrong.
The Patchwork Problem in Practice
The regulatory landscape that had seemed manageable during the boom became a nightmare during the retreat.
A company that wanted to restart operations couldn't just flip a switch. They needed to reapply for permits in every jurisdiction, often facing new restrictions, new requirements, new fees. California demanded detailed incident reports and safety cases. Texas wanted proof of insurance and little else. Arizona had gone quiet, its early enthusiasm cooled by years of incidents and public pushback.
The patchwork that had seemed like a minor inconvenience during growth became a fatal barrier during recovery.
Part IV: The Ghost in the Machine
What We Actually Built
We asked software to do something humans do with instinct—read a vibe, interpret intent, make judgment calls in ambiguous situations. We gave it rules and asked it to handle situations with no rules. We trained it on perfect data and deployed it into a world that is fundamentally imperfect.
The ghost in the machine isn't a bug. It's the gap between what we wanted and what we got. It's the expectation that we could replace human judgment with calculation.
We expected a servant. We got a very expensive, very confused calculator.
The Ghost in the Interstate
The I-80 corridor through the Sierra Nevada mountains became an unexpected laboratory for human-AI trust. Truck drivers, the humans most affected by automation, developed sophisticated mental models of when to trust autonomous features and when to override.
Research from this corridor revealed something profound: trust is built through consistent, interpretable behavior. When an autonomous system behaves predictably, humans learn to trust it. When it behaves unpredictably—even if safely—trust erodes.
The I-80 case study illuminates the dynamics of trust between humans and autonomous systems—lessons directly applicable not just to vehicles, but to any AI system that interacts with humans.
Lessons for Banking and Beyond
The autonomous vehicle industry's struggle holds profound lessons for every field pursuing AI autonomy—including banking and finance, which we've explored extensively in this series.
The Edge Case Problem: In banking, edge cases are fraud patterns that don't match training data, customers with unique circumstances, economic conditions not seen before, regulatory changes that redefine requirements. The long tail is equally long.
The Black Box Problem: Autonomous vehicle perception systems can't fully explain why they classified a plastic bag as a deer. Banking AI can't fully explain why it flagged a transaction as fraudulent or denied a loan application. Explainability matters for trust and regulation.
The Human-in-the-Loop Imperative: Autonomous vehicles need remote operators for edge cases. Banking AI needs human underwriters for complex decisions. The loop isn't going away—it's just moving.
The Simulation Gap: Banks test models on historical data, assuming future resembles past. But 2008 happened. COVID happened. Inflation happened. The future never matches the training data.
The Trust Deficit: Just as the public feared autonomous vehicles they didn't understand, customers fear AI banking decisions they can't explain. Trust is the ultimate currency, and it's earned through transparency.
This parallel between autonomous vehicles and automated underwriting is explored in depth in our analysis of what the algorithm really sees in insurance underwriting. The same principles apply.
The Architecture of Partnership
The most successful autonomous systems aren't those that eliminated humans—they're those that optimized the human-machine partnership.
Remote Assistance: When an autonomous vehicle encounters an edge case, it calls for help. A remote operator reviews the situation, provides guidance, and the vehicle proceeds. Humans handle the long tail; machines handle the routine.
Teleoperation: For complex scenarios, remote operators can take direct control. The vehicle becomes a robot with human intelligence. This isn't failure—it's graceful degradation.
Fleet Management: Humans monitor fleet health, optimize deployment, handle customer service, manage incidents. The machines drive; the humans orchestrate.
Continuous Learning: Every edge case handled by a human becomes training data. The system improves. The long tail shortens, one case at a time.
The architecture of this human-technical partnership is detailed in our piece on fraud detection tools and the human-technical bridge. The patterns are identical.
The principles of autonomous system design, including graceful degradation and human handoff protocols, are explored in The Architecture of Autonomy: Where Code Meets Humanity.
The challenges of late nights and hard handovers in automotive AI mirror the challenges of any 24/7 autonomous system.
The 1,100-mile autonomous trucking run from Bakersfield to Denver demonstrated what's possible when the environment is controlled and the edge cases are manageable.
And as we learned from telematics data analysis, behind every sensor reading is a human story—a principle equally vital in finance.
The digital dialogues between connected vehicles and infrastructure mirror the data exchanges between borrowers and lenders in modern finance.
Part V: The Road Ahead
What Remains
Commercial Presence: Phoenix, SF, LA, Austin, Dallas, Houston, Atlanta; testing in Tokyo and London.
This infrastructure doesn't disappear. The vehicles are parked, but the maps remain. The permits are inactive, but the relationships persist. The talent is reassigned, but the knowledge endures.
Seven cities of experience. Two global capitals of testing. Millions of miles of data. Thousands of edge cases documented and addressed. The foundation remains.
The Pivot Points
Waymo continues limited operations, focusing on safety above all. The leader now moves cautiously, methodically, sustainably. No more promises of rapid expansion. Just incremental progress.
Cruise rebuilds under new leadership, new oversight, new humility. The aggressive challenger learned the hardest lesson. The path back will be long.
Tesla pushes forward with its vision-only approach, leveraging its massive fleet. FSD improves incrementally. True autonomy remains elusive, but the driver-assist features get better every quarter.
Zoox continues testing in Las Vegas, patient under Amazon's ownership. The purpose-built vehicle waits for the technology to catch up to the vision.
Mobileye powers everyone else's driver-assist systems, gathering data, improving algorithms, waiting for the moment when L4 becomes viable.
Aurora focuses on trucking, where the economics work and the environment is simpler. The path to profitability is clearer.
What We Learned
The Long Tail is Real: Edge cases aren't exceptions—they're features of the real world. Any system that can't handle them can't be trusted.
Humans Are Not Going Away: The most successful autonomous systems are those that optimize human-machine partnership, not those that eliminate humans.
Trust is Earned, Not Engineered: No amount of technical sophistication compensates for lost trust. Every incident matters. Every headline erodes confidence.
Economics Matter: However elegant the technology, it must deliver value. The capital requirements of L4 autonomy may exceed any possible return.
The Social Wall is as Real as the Technical One: Communities will resist being lab rats. They will cone your vehicles, protest your expansions, and vote for restrictions. You can't engineer your way out of a social contract problem.
The 2030 Outlook
Where will we be in five more years?
Gradual Expansion: Robotaxis will return, but slowly, cautiously, in limited geographies with perfect weather and simple road networks. Phoenix first, then Sun Belt cities, then cautiously into more complex environments.
Trucking First: Long-haul highway autonomy will arrive before urban robotaxis. The economics are clearer, the environment simpler, the regulatory path more defined.
Human-Machine Teaming: The most successful systems will be those that optimize the partnership—machines handling routine driving, humans handling edge cases remotely.
Consumer Features: Advanced driver assistance will become standard, then expected, then required. Every new vehicle will include highway pilot, traffic jam assist, automated parking.
Public Acceptance: It will take a generation—literally—for autonomous vehicles to feel normal. The children of 2025 will grow up with driver-assist features and accept them as natural.
Conclusion: The Ghost We're Still Chasing
The ghost in the interstate isn't a technical problem—it's a human one. It's the gap between what machines can do and what humans need. It's the space where trust lives or dies. It's the recognition that autonomy without humanity is just automation—and automation without trust is just a machine waiting to fail.
The robotaxi dream of 2025 hit a digital wall. But walls can be climbed, tunneled under, or built around. The industry that emerges on the other side will be different—humbler, wiser, more realistic. It will build systems that work with humans, not instead of them. It will earn trust slowly, through consistent performance, not promises. It will deliver value incrementally, not revolution overnight.
And one day, maybe not in 2025 but in 2030 or 2035, we'll look back and realize that the wall wasn't failure—it was the teacher we needed.
The ghost is still out there. But now we know: it's not a ghost in the machine. It's the ghost of our own expectations, haunting the empty lots where the robotaxis used to charge. And until we learn to build with humility, with partnership, with genuine respect for the humans we claim to serve—that ghost will keep haunting us.
This analysis is part of our series on autonomous systems. For deeper dives into related topics:
- Automated Insurance Underwriting: What the Algorithm Really Sees
- Fraud Detection Tools: The Human-Technical Bridge
- The I-80 Ghost: Trust, Terror, and the Algorithm That Can't See Ghosts
- The Architecture of Autonomy: Where Code Meets Humanity
- Late Nights, Hard Handovers: Automotive Transportation AI
- 1,100 Miles of Autonomous Trucking: The Bakersfield to Denver Run
- Telematics Data Analysis: The Human Story Behind the Sensors
- Digital Dialogues: How Connected Cars Negotiate the Road We Share
Disclaimer: This article is for informational purposes only and represents analysis of public information about the autonomous vehicle industry. Specific company strategies and commercial decisions are subject to change. The individuals quoted are composites representing real worker experiences.
Last Reviewed: March 2026
Top comments (0)