Tesla Cybercab begins paid robotaxi rides in Austin, Texas

By Jin Yufan | Edited by Wei Jia Originally published by Dingjiao One (dingjiaoone.com)

In early September, Tesla's Cybercab began charging passengers for rides in Austin, Texas. As of September 3, Tesla had registered 420 autonomous vehicles in Texas, including 45 Cybercabs, with the remainder primarily being Model Y vehicles. The Cybercab service currently operates only in limited areas of Austin and remains small in scale.

According to app data and user feedback compiled by Huaxing Securities, Cybercab fares for trips of 3 to 8 miles are currently priced at approximately $1.50 to $2.00 per mile, with dynamic pricing in effect.

Two weeks later, the vehicle was displayed statically at Tesla's China World Trade Center (Huamao) experience store in Beijing.

Static Display in Beijing

On the afternoon of September 17, a crowd gathered around the display. The vehicle features only two seats and a large screen, with no steering wheel, accelerator, or brake pedals. Both doors open upward.

When asked when the vehicle might operate on Chinese roads, a store employee responded that there is no clear timeline. Regarding the automatic doors, the employee gave an example: if a driverless taxi drops off a passenger who forgets to close the door, traditional vehicles might require an operator to handle it. The Cybercab's doors can be opened and closed remotely, potentially eliminating that step as well.

This illustrates the challenges facing Level 4 autonomous driving. While drivers can be removed, tasks such as charging, cleaning, maintenance, and anomaly handling still require significant operational capability. The larger the fleet, the harder it becomes to ignore the associated labor and operational costs.

Meanwhile, major players including Waymo, Baidu's Apollo Go (Luobo Kuaipao), Pony.ai, and WeRide have already launched paid services in select cities. This year, the industry has begun scaling up, and companies are now calculating more specific metrics: how many orders a single Robotaxi can complete per day, how long it can run, how many support staff are needed, when it will break even, and whether a model that works in one city can be replicated in another.

The Cybercab enters this competitive landscape by making a single Robotaxi cheaper, but low cost is only the first hurdle in large-scale operations.

01. Cybercab: Making the Vehicle Cheaper First

The Cybercab has made significant reductions to the Robotaxi concept. The most obvious change is seating: the Cybercab has only two seats. At the Beijing Huamao store on September 17, staff explained that most trips involve only one or two passengers. If rear seats remain empty, they still take up space and add weight. For group travel, the fleet can deploy Model Y vehicles, and in the future, larger models like the Robovan.

The traditional driver's position has also been eliminated. The Cybercab has no steering wheel or pedals, removing the conventional driver control area entirely.

Interior and Design Changes

By removing the traditional driver's area, more interior space is available for passengers. The vehicle features a lower threshold and larger door opening to accommodate wheelchair users. Emergency braking and door-opening buttons include Braille markings.

The perception hardware follows Tesla's vision-only approach, also with reductions. According to public data compiled by Huaxing Securities, the Cybercab has a curb weight of approximately 1.4 tons and a battery capacity of less than 50 kWh. It relies on external cameras for environmental perception, with no lidar or millimeter-wave radar. A Beijing store employee stated that the vehicle uses eight cameras to collect road information, processed by Tesla's AI chip.

Cost Reduction Through Design and Manufacturing

The two-seat configuration, smaller battery, and elimination of the traditional driver area all affect overall vehicle cost. The manufacturing process also follows a cost-reduction logic. The Cybercab uses Tesla's new "Unboxed" manufacturing process. In a July interview, Tesla Vice President of Vehicle Engineering Lars Moravy revealed that the Cybercab production line automation rate has exceeded 90%, with the final assembly stage requiring only about 20 to 25 workstations.

From wiring harnesses and batteries to complete vehicle assembly, this production line differs significantly from the long-established Model 3 and Model Y lines. A store employee compared the Unboxed process to "building with blocks": the front, rear, battery base and seats, left and right body panels, and roof are produced separately before final assembly.

The body materials and painting process have also changed. The employee explained that the Cybercab's exterior uses composite materials that feel relatively soft but still have a rigid internal structure. The vehicle eliminates traditional painting; colors are integrated directly with the material during production. This was compared to "pouring chocolate into a mold"—the material's color determines the final appearance, reducing painting steps and lowering repair costs for minor damage.

Efficiency and Lifecycle Costs

Moravy repeatedly emphasized "efficiency" when discussing the Cybercab. This includes not only vehicle energy consumption but also the full lifecycle cost from raw materials entering the factory, through manufacturing, to subsequent charging and operation. For a vehicle designed for high-frequency operation, Tesla has been calculating per-mile costs and total ownership costs from the development stage.

According to Huaxing Securities citing Tesla, the Cybercab's manufacturing cost is below $30,000, with a long-term ideal operating cost of $0.20 per mile, accounting for electricity, depreciation, insurance, and maintenance.

However, the $30,000 cost range is no longer unique to the Cybercab. Baidu stated during its early 2026 financial results call that the current sixth-generation Robotaxi, the RT6, also has a per-unit cost below $30,000. The RT6 is also a production model designed from the ground up for Level 4 autonomy.

Operational Considerations Built Into Design

What makes the Cybercab particularly noteworthy is that it incorporates many unmanned operational requirements into the vehicle design from the outset. Examples include:

• Charging: The Cybercab supports 25 kW inductive wireless charging. When the vehicle parks in the correct position, no manual plugging or unplugging of charging cables is required.
• Interior materials: Seats and flooring use materials that are easier to clean for high-frequency maintenance.
• Passenger monitoring: An interior camera monitors passengers.
• Trunk: The trunk can be opened and closed electronically.
• Remote door control: If a passenger forgets to close the door after exiting, it can be handled remotely.
• Vehicle identification: When a passenger hails a Cybercab, the front light bar displays a color corresponding to the app, making it easier to locate the correct vehicle among a row of driverless cars.

These reductions ultimately affect the manufacturing cost and operational convenience of a single vehicle. After making the vehicle itself cheaper, Tesla's focus shifts to rapidly expanding the fleet.

Commercial Sales and Fleet Operations

Tesla's website in the U.S. has already opened commercial purchase inquiries for the Cybercab. Individuals or businesses can register to purchase single or multiple vehicles for commercial operation. The Robotaxi partnership page is also soliciting cooperation for Cybercab fleet procurement, as well as operational stations and infrastructure.

A store employee specifically mentioned pricing and operational models: the vehicle costs $20,000. After joining the Robotaxi fleet, Tesla handles unified maintenance and operation, and the purchaser shares in the operating revenue. This is more akin to an operational investment than buying a private vehicle for personal use. However, the final price, access conditions, and revenue-sharing model have not yet been officially announced. Whether external buyers are willing to purchase ultimately depends on whether the vehicle can deliver a viable financial return.

02. Robotaxi Operators Begin Calculating Operational Economics

With the Cybercab hitting the road, domestic Robotaxi companies in China are once again being compared. The current frontrunners remain Baidu's Apollo Go (Luobo Kuaipao), Pony.ai, and WeRide.

Each company has different priorities:

• Apollo Go: Has the broadest deployment and largest operational scale.
• Pony.ai: In recent years, has emphasized front-end mass production, per-vehicle economics, and cross-city replication.
• WeRide: In addition to Robotaxis, develops other Level 4 products and has expanded earlier into overseas markets.

In past years, companies frequently reported autonomous driving mileage, operational areas, and testing progress. This year, financial reports and operational announcements increasingly feature metrics such as average daily orders, per-vehicle revenue, and unit economics (UE).

Pony.ai's Per-Vehicle Economics

In March, Pony.ai disclosed per-vehicle revenue and costs for its seventh-generation Robotaxi in Shenzhen. As of the end of February, the average daily net revenue per vehicle over the previous month was 338 RMB, with an average of 23 orders per day. The per-vehicle unit economics (UE) reached breakeven. Guangzhou had already achieved this milestone earlier.

Pony.ai further broke down costs for Dingjiao One. The autonomous driving system bill of materials (BOM) for the seventh-generation Robotaxi decreased by 70% compared to the previous generation, including an 80% reduction in domain controller costs and a 68% reduction in lidar costs. Beyond increased deployment scale, Pony.ai leverages OEM supply chain channels to reduce component procurement costs.

However, the vehicle itself is only part of the equation. Pony.ai emphasizes that operational costs are roughly as important as the total vehicle BOM. Per-vehicle operating costs include charging, ground support, remote safety operators, insurance, maintenance, network traffic, and parking. If only the vehicle is made cheaper without corresponding reductions in operational costs, the cost advantage remains limited.

Labor and Utilization

Labor is one area where changes are most visible. While drivers have been removed, people have not been completely eliminated. Pony.ai explains that special situations such as passenger emergencies or traffic controls may still require remote operator intervention. Ground operations staff handle charging, cleaning, maintenance, and inspections.

When Guangzhou achieved UE positivity last year, one remote assistant covered approximately 20 Robotaxis. Technologically, this ratio can now reach 1:30. Ground operations staff are also beginning to serve multiple vehicles. Pony.ai states that one operations staff member can currently complete charging, cleaning, and inspections for dozens of vehicles within one hour.

Beyond labor, vehicle utilization directly affects the payback period. Pony.ai told Dingjiao One that improving per-vehicle economics requires not only reducing vehicle and operational costs but also increasing daily operating hours and order volume. After returning to the operations depot, Robotaxis can now autonomously find charging spots, even in tight spaces. Some tasks that previously required a person to drive the vehicle back and find parking are being automated.

Scaling Up: Lessons from Delivery Vehicles

Robotaxi fleets in individual cities currently number only in the thousands. For insights into larger scale, one can look at autonomous delivery vehicles, which have already reached tens of thousands of units.

Xin Shiqi (Neolix) CEO Yu Enyuan told Dingjiao One that the company currently operates approximately 27,000 unmanned vehicles. At the hundred- and thousand-vehicle scale, teams focus primarily on autonomous driving technology. At the ten-thousand and even hundred-thousand scale, charging, swapping, maintenance, repair, and safety management become significantly heavier. "The focus shifts from pure technology to a combination of technology and operations," Yu said.

Neolix's current person-to-vehicle ratio is approximately 1:50. Yu stated that if this can reach 1:100 or higher in the coming years, it would be close to ideal. Since delivery orders vary more widely in length, Neolix focuses more on operational mileage. Yu explained that Qingdao currently has nearly 2,000 Neolix unmanned vehicles forming a formal citywide operational network, with nighttime orders roughly equal to daytime orders. When vehicles can run continuously day and night, utilization improves, significantly impacting payback periods and return on investment.

Profitability Remains Elusive

While passenger and delivery vehicles track different metrics, they share one common requirement: vehicles should remain idle as little as possible.

Achieving positive per-vehicle unit economics is still far from overall company profitability. In the second quarter of this year, Pony.ai's Robotaxi revenue increased nearly sevenfold year-over-year, and WeRide's total revenue grew more than 80%, but both companies remained unprofitable for the quarter. Research and development, computing power, vehicle expansion, and new market investment continue to require significant spending.

Cross-City Replication

Beyond making vehicles in one city more efficient, the next step is expanding fleets to more locations. Pony.ai told Dingjiao One that while the same autonomous driving system can now adapt more quickly to different markets, each city still requires new license applications, compliance filings, and operating permits. Special road rules, traffic signs, user travel habits, pricing standards, and service details must also be adjusted locally.

One approach is to delegate certain tasks to partners with existing systems. In March, over 100 GAC Aion Robotaxis entered the Ruqi platform. GAC provides the vehicles, Pony.ai supplies the seventh-generation autonomous driving system, and Ruqi handles order dispatch and daily operations. OEMs have existing production and supply chain systems, while mobility platforms already have users, dispatch capabilities, and offline operations.

A professional working in intelligent driving product planning at an automaker told Dingjiao One that as Level 4 continues to develop, automakers may further participate in unmanned vehicle operations beyond manufacturing and selling vehicles. His assessment is that while technology continues to advance, many automakers "are not yet ready for the identity shift."

For companies that have already achieved positive UE in Shenzhen and Guangzhou, the key question is whether efficiency can continue to improve as scale expands. Only with sustained data improvement can Level 4 evolve from an operational business in a few cities into a replicable business model.

03. Who Is Responsible When There Is No Driver?

At Level 4, the changes go beyond the absence of a driver in the vehicle—the allocation of driving responsibility also shifts.

The industry professional mentioned above told Dingjiao One that one of the biggest changes from Level 2 to Level 4 is driving responsibility. At Level 2, primary responsibility still rests with the human. At Level 4, "responsibility definitely lies with the manufacturer, with the automaker." At the same time, requirements for vehicle operating range, fault handling, and safety redundancy increase significantly.

He also noted that many of the main models and data accumulated during Level 2 and Level 3 can continue to be used, but Level 4 requires additional systems that were unnecessary for private passenger vehicles. For example, how to handle extreme situations remotely is a new design consideration.

The Cybercab has eliminated the steering wheel and pedals, and the vehicle does not retain traditional driving controls for passengers. After an incident occurs, sufficient data must be retained to reconstruct the situation.

The "Black Box" for Autonomous Vehicles

Tianjin Bool Technology provides autonomous driving data recording systems for automakers—essentially a "black box" for unmanned vehicles. Technical Director Zhu Peitian told Dingjiao One that autonomous driving systems record data from perception, prediction, planning, decision-making, and control stages. However, when determining responsibility, simply looking at the system's commands is insufficient.

For example, if the system records a deceleration request, the actual deceleration achieved depends on real physical data such as wheel speed. If there is a discrepancy between the control command and the vehicle's actual behavior, the determination of the accident's cause may differ.

This data also serves another purpose. Zhu explained that data from both the vehicle and the cloud is used to identify edge cases, which are then used to train and refine the autonomous driving system. An anomaly recorded once may become training data for the next algorithm iteration.

System Reliability at Scale

As fleet sizes increase, the systems that record this data must themselves remain stable over the long term. Bool Technology Product Director Xin Shubin noted that during project testing, issues with device clock synchronization were encountered. If any node's clock behaves abnormally during the synchronization of timestamps across image acquisition devices, electronic control units, domain control systems, and other hardware and software nodes within the vehicle system, data from multiple sources cannot be aligned to the same time reference, causing timing errors.

Some issues may not surface during small-scale testing. After vehicles operate at high frequency over extended periods, data recording equipment must also contend with storage aging and device reliability. Xin believes such systems must meet reliability requirements for the entire vehicle operational lifecycle.

Regulatory Hurdles

Yu Enyuan told Dingjiao One that when unmanned vehicles in a city exceed 1,000 units and operate 24/7, safety situations arise that did not occur at smaller scales, increasing demands on both technology and operations.

The Cybercab currently faces regulatory hurdles. U.S. motor vehicle safety standards are designed for vehicles with drivers, steering wheels, and pedals. By removing these components, Tesla must demonstrate that the Cybercab still complies with existing federal standards.

On September 4, the National Highway Traffic Safety Administration (NHTSA) launched an investigation. On September 15, it requested additional certification materials from Tesla, with a response deadline of September 30. The core question: how can a vehicle designed without a steering wheel prove it meets current safety requirements?

The industry professional believes that many current standards for intelligent connected vehicles primarily govern how vehicles are tested, approved, and what safety requirements they must meet. At Level 4, with no driver in the vehicle, new standards and requirements will emerge for how vehicles interact with roads and traffic systems.

Path to Commercialization

The Cybercab has already begun paid operations in Austin, Texas, but only within a limited area. There is no clear timeline for when it might operate on Chinese roads.

For Robotaxis to truly enter a city, they must also address their relationship with existing mobility services. As unmanned vehicle fleets expand, the income and employment expectations of taxi and ride-hailing drivers will be affected. Previous discussions about autonomous driving deployment in cities such as Wuhan have also addressed how to coordinate old and new business models and how to smoothly transition driver groups.

Therefore, how quickly Robotaxis can scale is not determined solely by technology. The number of vehicles deployed and the area covered must also consider urban transportation, employment, and how existing mobility services adapt to these changes.

All photos except where noted were taken by Dingjiao One.