On September 3, 2026, a Cybercab without a steering wheel or pedals began carrying paying passengers in Austin.
The following day, the National Highway Traffic Safety Administration (NHTSA) launched a compliance investigation into approximately 1,000 related vehicles.
Within 24 hours, the launch of paid services and the initiation of the compliance investigation occurred in succession, laying bare the current state of the industry within this time gap: autonomous vehicles are already capable of hitting the roads to charge fares, yet regulatory access, safety trust, and unit economics have not yet crossed the critical threshold. We judge that the true watershed moment will arrive around 2030, not at the launch event—when cost, policy, demand, and capital align simultaneously, Robotaxis will scale from demonstration fleets of around 10,000 vehicles to millions globally, rewriting urban mobility, employment structures, and value distribution in the automotive industry.
I. Cybercab: A Car Without a Steering Wheel Exposes the Industry's Real Bottlenecks
The Cybercab is the first mass-produced autonomous vehicle designed from the ground up based on first principles to eliminate the steering wheel, pedals, and rearview mirrors. Relying on a pure vision solution with eight cameras and native manufacturing, it drives the target selling price below USD 30,000 and the target operating cost down to USD 0.2 per mile (approximately CNY 0.84 per km). Inside, a 22-inch screen controls the doors, air conditioning, seats, navigation, and entertainment. The few hundred vehicles deployed in 2026 are not the focus in themselves; the key is that removing safety drivers, native mass production, target costs, and production disruption have been achieved simultaneously on the same native autonomous vehicle for the first time.
However, the starting point of commercialization has been quite restrained. As of the launch, there were 314 licensed vehicles on the road in Texas, USA, of which only 45 were Cybercabs, with the rest being modified Model Ys. In Austin, approximately 200 were actually in operation, running from 6:00 to 22:00, and temporarily barred from highways. Within the first week of launch, several incidents were captured on camera: a door nearly caught a child; a vehicle broke down in the middle of the road and, lacking a steering wheel, could not be moved, simply blocking traffic; a 10-minute trip was detoured to take 70 minutes; and the App only displayed the estimated price, not the duration. The day after the launch, Tesla's stock price fell by about 6%, with media estimates calculating a single-day market capitalization evaporation of approximately USD 90 billion.
The focus of the NHTSA investigation is whether the certification complies with the current Federal Motor Vehicle Safety Standards (FMVSS), which still require vehicles to be equipped with steering wheels, pedals, and rearview mirrors. In other words, the primary constraint facing the Cybercab is no longer the autonomous driving capability itself: whether regulations will allow the mass production and road deployment of vehicles without steering wheels, and whether society is willing to entrust safety to machines—these two questions remain unanswered.
Multi-sensor fusion solutions represented by Waymo, leveraging LiDAR and safety redundancy, have already begun small-scale paid operations in multiple cities, but the per-vehicle cost remains high; Tesla, on the other hand, is betting on low cost and rapid generalization with a pure vision approach, using a native vehicle to rapidly drive down the product cost curve.
Source: Public information, Aijian Securities, Beginners in AI, image sourced from the internet, analysis by Gasgoo Auto Research Institute
II. The 2030 Scale Inflection Point: Not Linear Growth, but a Tipping Point After Four Factors Align Simultaneously
2025 is dubbed the first year of paid Robotaxi operations, yet the global market size for the year is still measured in hundreds of millions of dollars, with in-operation fleets totaling around 10,000 vehicles, of which approximately 3,800 are in the US and 5,000 to 7,000 in China and overseas. Technical validation has been completed, but the scale remains very small.
The market size of China's shared mobility (including taxis, ride-hailing, carpooling, etc.) was approximately CNY 841.2 billion in 2025 and is projected to reach CNY 1,526.4 billion by 2030. Among these, China's Robotaxi revenue is expected to reach CNY 158.3 billion in 2030, approaching 20.4% of the ride-hailing market size. In other words, Robotaxis in 2030 will not replace ride-hailing but will capture a 20% share in a still-expanding mobility market, meaning the threshold for economic viability has been crossed.
Source: Public information, Frost & Sullivan, BCG, Aijian Securities, analysis by Gasgoo Auto Research Institute
The reason the inflection point falls around 2030 rather than earlier lies in the gap between per-vehicle economics and company-level economics.
According to estimates by relevant consulting firms, the total cost per kilometer for a Robotaxi without a safety driver is approximately CNY 0.55, a drop of about 72% compared to the CNY 1.93 for traditional fuel-powered ride-hailing vehicles.
Looking solely at the per-vehicle level, this cost advantage already makes financial sense: cities like Wuhan, Guangzhou, Shenzhen, and Abu Dhabi have already achieved a positive per-vehicle economic model (all excluding R&D amortization). However, if this gross profit deducts R&D amortization (estimated by relevant consulting firms to exceed USD 3 billion invested in L4 autonomous driving platforms), headquarters expenses, remote assistance center costs, insurance, and urban infrastructure expenses, Robotaxi companies as a whole have not yet achieved profitability.
To cross the break-even point, scale is also a threshold.
Crossing this threshold requires four factors to be established simultaneously:
Hardware cost reduction, technology convergence, model generalization, and policy opening
1. Hardware cost reduction: The whole vehicle costs under CNY 200,000, the total cost per kilometer is CNY 0.55, and it can be replicated across multiple cities.
2. Technology convergence: End-to-End/VLA + World Model + rule-based fallback to achieve the removal of safety drivers, with an accident rate no worse than that of humans.
3. Model generalization: A single model reused across multiple scenarios (Robotaxi, Robotbus, etc.), enabling the opening of a new city within months.
4. Policy opening: Fully driverless permits covering first-tier cities with cross-regional mutual recognition, forming closed loops for liability, insurance, and data.
Source: Public information, BCG, Frost & Sullivan, China Passenger Car Association data, McKinsey, image sourced from the internet, analysis by Gasgoo Auto Research Institute
III. The Social Transportation Landscape After Large-Scale Deployment: Mobility Costs Reset, at the Expense of Job and Urban Restructuring
1. Changes in Mobility Logic
The mobility logic in first-tier cities and among young demographics will shift from owning a car to summoning one on demand. The proportion of private cars in urban mobility will trend downward but will be retained in suburban, long-distance, and personalized scenarios, ultimately forming a mixed transportation system within cities dominated by unmanned sharing and supplemented by private cars.
2. Mobility Costs and Equity
The most direct change lies in pricing and the equity brought by it. The low price of Robotaxis will not only substitute existing demand but also induce new demand; all-weather, fully accessible autonomous vehicles will provide mobility capabilities previously hard to obtain for the elderly, people with disabilities, and those without driver's licenses. When mobility costs approach those of public transportation and availability approaches that of private cars, urban mobility begins to take on the attributes of a public utility.
3. Cities and Roads
Urban space and road operations will also change accordingly. Continuous vehicle operation, centralized parking, and wireless charging will lead to a decline in parking demand and parking lot land use in core areas; algorithms do not get fatigued or drive drunk, so long-term accident rates tend to decrease. However, in the short term, there is a counterforce: deadheading and induced demand may exacerbate congestion in the early stages of popularization, which needs to be absorbed through vehicle-infrastructure cooperation, unified dispatching, and dedicated pick-up and drop-off zones.
4. Employment Patterns
The most realistic impact is on employment. As of October 2024, there were approximately 7.48 million licensed ride-hailing drivers in China, a group that will face transformation; new jobs such as remote safety officers, vehicle operation and maintenance, cleaning, charging, and dispatching will emerge, but the number of jobs will be fewer than the replaced driving positions, and the skill requirements will differ, requiring the integration of social security and retraining. This cost will not be automatically absorbed along with the corporate growth curve and requires public policy to address it in advance.
5. Industrial Patterns
Industrial value will be redistributed. Automakers will shift from selling products to selling mobility capacity, with the profit pool concentrating on LiDAR, domain controllers, drive-by-wire chassis, and operating platforms. Cities will see dedicated Robotaxi stations and pick-up/drop-off zones; cybersecurity, accident liability determination, data privacy, and the widening urban-rural mobility gap caused by autonomous vehicles prioritizing high-density cities are all constraints that need long-term handling.
IV. Modular Production Rewrites Auto Manufacturing: Costs and Capacity Rapidly Escalate, but Supply Will Outpace Demand
Another breakthrough of the Cybercab lies in the factory. It adopts what Tesla calls Unboxed production: while traditional cars are assembled segment by segment on a single assembly line, the Cybercab disassembles the vehicle into modules such as the floor pan, front and rear ends, and side panels for parallel assembly, finally integrating them all at once. Tesla states that this approach is closer to the manufacturing logic of high-volume consumer electronics, with a theoretical takt time of about 10 seconds per vehicle.
This process and the native design without a steering wheel mutually reinforce each other. After eliminating the steering wheel, steering column, pedals, and instrument panel, the number of parts and mechanical connections is significantly reduced. The full drive-by-wire chassis makes module boundaries clearer, naturally suiting parallel assembly; modularity also compresses assembly line length, labor hours, and fixed investment, making the economies of scale steeper, directly supporting its target selling price of under USD 30,000 and target operating cost of USD 0.2 per mile. The same floor pan module can also be derived into models like the Robovan, further amortizing manufacturing costs.
For the automotive industry, this is a paradigm shift.
First, factory-installed native vehicles replace the modification model of adding sensors to mass-produced cars; autonomous vehicles begin to be manufactured like consumer electronics rather than modified equipment;
Second, the barriers to manufacturing integration decrease, and value concentrates on core modules such as software, operating platforms, drive-by-wire chassis, domain controllers, and sensors. The role of automakers shifts from selling whole vehicles to manufacturing mobility capacity for operators.
However, cost reductions and capacity leaps do not automatically equate to demand realization; the ramp-up speed on the supply side is far faster than on the demand side. Tesla's Texas factory has already built an annual capacity of over 125,000 vehicles, with the company's long-term annual capacity vision set at 2 million and ultimately 4 million vehicles, while the clear in-operation fleet target for 2026 is only about 1,000 vehicles, with permits and actual deployments still counted in the hundreds.
Demand and trust are slow variables. Current federal safety standards still require steering wheels and pedals, and the NHTSA compliance investigation has not yet concluded; the handling of steering-wheel-less vehicles in breakdowns and extreme scenarios is still under scrutiny, and operations are restricted by geofencing and time slots;
Substitution rates, wait times, and cross-city permit mutual recognition must all ramp up city by city. The manufacturing side can use the speed of consumer electronics to pull annual capacity to the million-vehicle level, but the demand side can only open up city by city, following regulations, safety validation, and urban order density.