From driverless to unmanned operation, Robotaxi is starting to recalculate its business model.

The competitive landscape of the Robotaxi industry is shifting.

At the recent earnings conferences of Pony.ai, WeRide, and Baidu, the operational costs behind the expansion of Robotaxi fleets became a topic of discussion among analysts during this earnings season. They were concerned about the costs and commercial returns behind the fleet expansion.

"To date, L4 vehicle intelligence has basically achieved continuous operation capabilities in limited scenarios. What each company needs to do now is evolve from single vehicles to fleets, solving fleet operation and commercialization," said Zhao Chenhui, CTO of Cao Cao Mobility's RoboX business unit, in an exclusive interview with Wall Street Insights.

In his view, the ability of vehicles to complete autonomous driving in limited areas is only the starting point for the large-scale deployment of L4; safety redundancy, remote support, energy replenishment and preparation, order scheduling, and per-vehicle economy are becoming equally important capabilities.

RoboX is Cao Cao's solution to this change. It expands its business from Robotaxi to Robovan and connects different types of intelligent transportation capacity through CaoCao Robo OS.

As of the end of June, Cao Cao had deployed 140 second-generation Robotaxi vehicles. The company plans to mass-produce the natively developed Robotaxi model, Eva Cab, in 2027.

At this stage, the more critical question is whether this system can transform driverless vehicles into safe, stable, and replicable transportation capacity.

01 What does RoboX aim to solve?

At the heart of RoboX is Cao Cao's attempt to transform existing ride-hailing platforms into systems capable of simultaneously managing manned and unmanned, passenger and freight capacity.

Zhao Chenhui believes that L4 autonomous vehicles have basically crossed the threshold of "whether they can be driven." The next challenge is whether the fleet of autonomous vehicles can continuously accept orders, recharge, maintain operations, and handle anomalies. Therefore, competition extends from the autonomous vehicles themselves to the entire chain of vehicle manufacturing, order processing, and offline operations.

According to Cao Cao's plan, RoboX currently covers Robotaxi and Robovan, with space reserved for Robobus, Robotruck, and other forms. The "X" here does not correspond to a specific vehicle, but rather refers to different intelligent transportation capabilities that can be accessed by the same platform.

Cao Cao therefore combined Geely's vehicle manufacturing and intelligent driving resources with his own order, scheduling, and operation system.

The company summarizes this as a three-pronged approach: "intelligent customized vehicles, intelligent driving technology, and intelligent operation." This aims to shorten the chain between vehicle development and operational needs.

The first issue to address is how to integrate autonomous vehicles into existing transportation networks. Level 4 vehicles are constrained by design and operational conditions and local permits, and early fleets may struggle to independently cover the entire demand of a city. Peak-hour, cross-regional, and long-tail orders still require manned vehicles; however, if orders consistently favor manned vehicles, autonomous vehicles will struggle to achieve sufficient utilization.

Zhao Chenhui believes that the two types of transportation capacity will coexist for a long period of time. Mixed order dispatch is not a transitional arrangement, but a supply and demand issue that the platform needs to deal with in the long term.

CaoCao Robo OS plays this role. Building upon the "CaoCao Brain's" system for matching passengers, drivers, and orders, it further incorporates the operational range, remaining battery power, readiness status, and service capabilities of driverless vehicles, and attempts to uniformly schedule transportation capacity such as Robotaxi and Robovan.

The platform needs to decide which type of vehicle to accept orders, when to replenish its capacity, and how to maintain a balance between passenger waiting time, autonomous vehicle utilization, and manned capacity supply.

Zhao Chenhui also mentioned that Robo OS will provide a capacity entry point for AI intelligent agents and retain service memories such as user pick-up point, vehicle type and cabin settings, using this preference information to improve the next ride experience.

The system's decision-making also relies on two types of data: driving data accumulated from Geely's mass-produced vehicles, used to discover low-frequency and long-tail scenarios; and travel data from Cao Cao, used to determine when and where a vehicle goes, from which entrance it picks up passengers, and whether it can park legally.

The former affects "how the car is driven", while the latter determines "how the order is completed".

As of the first half of 2026, Cao Cao Mobility covered 215 cities with an average of 44.6 million monthly active users. Its existing mobility network provides the order base for hybrid dispatching.

02 How to calculate the economics of bicycles

How many rides a Robotaxi needs to complete per day to break even is a long-standing concern in the industry.

In March of this year, Pony.ai disclosed that its seventh-generation Robotaxi achieved monthly break-even in Shenzhen in February: the average daily net income per vehicle was 338 yuan, with an average of 23 rides completed per day. The cost included vehicle and intelligent driving kit depreciation, energy, maintenance, insurance, remote operation, labor, parking and network infrastructure, etc.

Zhao Chenhui believes that different companies have different cost structures and main operating areas, and therefore the number of orders required to reach the break-even point for each vehicle also varies.

Robotaxis eliminate the need for a driver, but increase costs associated with autonomous driving kits, vehicle redundancy, remote support, and on-site assistance. The costs of vehicle procurement, custom development, and refueling also vary among different companies.

He also stated, "The number of rides needed to break even for a single vehicle also depends on which city the vehicle operates in. The average order value also varies from city to city."

The break-even point for a single vehicle is whether the revenue from paid orders can cover vehicle depreciation and operating expenses. In addition to order volume, passenger mileage, average order value, and empty-run rate also affect this calculation.

Zhao Chenhui stated that the focus in 2026 will be on using existing vehicles to validate autonomous driving, passenger interaction, hybrid dispatching, and backend support, and then importing the results into Eva Cab.

Eva Cab attempts to change the cost structure starting with the vehicle itself. While conventional passenger cars are developed around the driver, native Robotaxis is centered on passengers and operators.

According to Zhao Chenhui, the Eva Cab eliminates traditional driving control components, redesigns the passenger space, adopts sliding doors and high-durability components, and supports automatic battery swapping; configurations unsuitable for commercial vehicles will be simplified.

The native design aims to reduce aftermarket modifications, lower maintenance complexity, and extend the lifespan of frequently used components.

Without a driver, passenger vehicle confirmation, pick-up and drop-off point identification, door anomalies, and cabin responses all need to be handled by the product and back-end systems. The sliding doors and passenger interaction optimizations mentioned by Zhao Chenhui are also aimed at reducing the time spent on the "last few meters" of passenger pick-up and drop-off.

The green intelligent access island is responsible for recharging and preparing vehicles during periods when they are not accepting orders.

Previously, drivers performed tasks such as recharging, cleaning, and inspections separately. These tasks are now centralized at the depots, and the land, equipment, and personnel costs for these depots need to be shared by the fleet. According to Cao Cao's plan, the green intelligent access islands can share battery swapping facilities with manned customized vehicles to reduce redundant construction investment.

Cao Cao also hopes to reuse existing customer service, vehicle management and order systems, while providing remote support, on-site rescue and preparation processes for unmanned vehicles.

03 How to implement the "Double 100,000" strategy

Cao Cao has proposed deploying a total of 100,000 Robotaxis and 100,000 Robovans by 2030. As of the end of June 2026, the company had deployed 140 second-generation Robotaxis. Further expansion requires Eva Cab to enter mass production as planned in 2027, and the safety, dispatch, and operational processes validated in Hangzhou to be replicated in other cities.

When the number of vehicles increases, the growth rate of personnel and station investment needs to be lower than the growth rate of the fleet in order to reduce the operating cost per vehicle.

According to the timeline given by Zhao Chenhui, the "double 100,000" target will go through three stages: model verification, mass production of original vehicles, and cross-city replication.

The industry is exploring different expansion strategies. Pony.ai adopts a joint deployment approach, while WeRide emphasizes an asset-light strategy overseas, with automakers, mobility platforms, and local operators sharing the costs of vehicles and operations.

Cao Cao leverages Geely's manufacturing system and its own mobility platform to directly participate in vehicle model definition, deployment, and fleet operation. The simultaneous expansion of vehicles and infrastructure will increase capital requirements and place higher demands on asset operational efficiency.

Robovan is the other half of the "double 100,000" target. In July, Cao Cao launched the first batch of Robovan commercial operations in Changsha and proposed models such as vehicle sales, leasing, and robots as a service.

Zhao Chenhui believes that passenger and cargo transport systems can reuse autonomous driving, scheduling, refueling, and maintenance systems, thus spreading the underlying investment.

Going overseas also requires solving the problem of local operations.

Zhao Chenhui said, "When we go global, the first problem we encounter may not necessarily be a technical issue, but rather how to conduct localized operations and how to achieve localized compliance."

Vehicle certification, licensing, insurance, accident liability, and data compliance must all meet local requirements. Zhao Chenhui stated that, leveraging Geely's existing international dealer network, safety and certification systems, Cao Cao has an advantage in obtaining vehicle access overseas and has already begun collaborating with local companies in the Middle East and other regions.

For Cao Cao, moving from 140 vehicles to a larger scale ultimately meant factoring the depreciation and operational costs of the increased vehicles into the ever-growing number of paid orders.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.