The competitive landscape of the Robotaxi industry is undergoing a significant shift.

During recent earnings calls for Pony.ai, WeRide, and Baidu, the operational economics behind Robotaxi fleet expansion became a key topic of analyst scrutiny, with a focus on the costs and financial returns that accompany scaling.

“So far, L4 single-vehicle intelligence can basically sustain continuous operation in limited scenarios. The task now for every company is to evolve from a single vehicle to a fleet, solving the problems of fleet operation and commercialization,” Zhao Chenhui, CTO of CaoCao Mobility's RoboX division, said in an exclusive interview with Wallstreetcn.

In his view, a vehicle's ability to drive autonomously within a confined area is only the starting point for L4 scale-up. Safety redundancy, remote support, energy replenishment and maintenance, order dispatch, and per-vehicle economics are becoming equally critical capabilities.

RoboX is CaoCao's answer to this evolving dynamic. It extends operations from Robotaxi to Robovan, connecting different types of intelligent mobility through the CaoCao Robo OS platform.

By the end of June, CaoCao had deployed 140 second-generation Robotaxis. The company plans to mass-produce its purpose-built Robotaxi model, the Eva Cab, in 2027.

The more immediate challenge is whether this system can convert autonomous vehicles into safe, stable, and replicable mobility capacity.

What Problems Does RoboX Solve?

The core of RoboX is CaoCao's attempt to transform its existing ride-hailing platform into a system capable of managing both manned and unmanned vehicles, as well as passenger and freight mobility.

Zhao believes that L4 single vehicles have largely crossed the “can it drive” threshold. The next hurdle is whether an unmanned fleet can continuously handle order intake, recharging, maintenance, and anomaly resolution. Competition is thus extending beyond the autonomous vehicle itself to encompass the entire chain of vehicle manufacturing, order acquisition, and on-the-ground operations.

According to CaoCao's plan, RoboX currently covers Robotaxi and Robovan, with room reserved for Robobus and Robotruck. The “X” here doesn't correspond to a specific vehicle model, but rather to the various intelligent mobility forms that can be dispatched through a single platform.

To achieve this, CaoCao integrates Geely's vehicle manufacturing and intelligent driving resources with its own order, dispatch, and operations systems.

The company describes this as a trinity of “intelligent custom vehicles, autonomous driving technology, and intelligent operations,” aiming to shorten the chain between vehicle development and operational needs.

The first issue to address is how unmanned vehicles integrate into the existing mobility network. L4 vehicles are constrained by operational design domains and local permits, and early fleets struggle to independently cover all of a city's demand. Peak-time, cross-district, and long-tail orders will still need manned vehicles. If orders lean persistently toward manned vehicles, unmanned vehicles will struggle to achieve sufficient utilization.

Zhao predicts that both types of capacity will operate in parallel for an extended period. Mixed dispatch is not a transitional arrangement, but a supply-demand issue that platforms need to manage long-term.

The CaoCao Robo OS plays this role. Building on the “CaoCao Brain's” matching of passengers, drivers, and orders, it incorporates factors like an unmanned vehicle's operational range, remaining battery, service readiness, and capabilities, and attempts to coordinate the dispatch of Robotaxis, Robovans, and other capacity.

The platform must decide which vehicle type takes an order, when to recharge, and how to balance passenger waiting time, unmanned vehicle utilization, and manned capacity supply.

Zhao also mentioned that the Robo OS will provide a mobility gateway for AI agents, while retaining service memory such as user pickup points, vehicle preferences, and in-cabin settings to improve the next ride experience.

System decision-making also relies on two data streams: driving data from Geely's mass-produced vehicles, used to identify low-frequency and long-tail scenarios, and CaoCao's mobility data, used to determine where vehicles should be at certain times, where to pick up passengers, and whether stops are compliant.

The former influences “how the vehicle drives,” the latter determines “how the order is fulfilled.”

By the first half of 2026, CaoCao Mobility covered 215 cities with an average of 44.6 million monthly active users, providing an order base for the mixed-dispatch network.

Calculating the Per-Vehicle Economics

How many orders a Robotaxi needs per day to break even has long been a critical industry question.

In March this year, Pony.ai disclosed that its seventh-generation Robotaxi achieved monthly per-vehicle breakeven in Shenzhen in February: average daily net revenue of 338 yuan from an average of 23 daily orders, with costs covering vehicle and autonomous driving kit depreciation, energy, maintenance, insurance, remote operations, labor, parking, and network infrastructure.

Zhao believes that each company's cost structure and primary operating region differ, so the order volume needed for per-vehicle breakeven also varies.

A Robotaxi eliminates the in-car driver, but adds costs from autonomous driving kits, vehicle redundancy, remote support, and roadside assistance. Vehicle procurement, custom development, and recharging costs also differ among players.

He also noted, “How many orders it takes to break even per vehicle also depends on which city the vehicle operates in. Average order values differ across cities.”

The per-vehicle breakeven calculation boils down to whether revenue from paid orders can cover vehicle depreciation and operating expenses. Beyond order volume, revenue per mile, average order value, and deadhead rates all affect the equation.

Zhao stated that the focus for 2026 is to use existing vehicles to validate autonomous driving, passenger interaction, mixed dispatch, and backend support, then feed those results into the Eva Cab.

The Eva Cab aims to change the cost structure starting from the vehicle itself. While ordinary passenger cars are designed around the driver, a purpose-built Robotaxi is centered on passengers and the operator.

According to Zhao, the Eva Cab eliminates traditional driving controls, redesigns passenger space with sliding doors and high-durability components, and supports automatic battery swapping. Features not suited for operational vehicles will be streamlined.

The purpose-built design seeks to reduce aftermarket retrofitting, lower maintenance complexity, and extend the lifecycle of high-frequency operational components.

Without a driver, passenger vehicle confirmation, pickup point recognition, door anomalies, and in-cabin responses must all be managed by product and backend systems. The sliding doors and passenger interaction upgrades Zhao mentioned are also designed to cut the “last few meters” of time spent picking up passengers.

Green Intelligent Mobility Hubs handle recharging and maintenance during periods when vehicles are not actively taking orders.

Tasks previously performed by drivers, such as recharging, cleaning, and inspection, are now centralized at these stations. The land, equipment, and personnel costs of these stations need to be spread across the fleet. Under CaoCao's plan, these hubs can share battery-swap infrastructure with manned custom vehicles to reduce duplicate investment.

CaoCao also plans to reuse existing customer service, fleet management, and order systems, while adding remote support, roadside assistance, and service preparation processes for unmanned vehicles.

How the “Double 100,000” Target Will Be Achieved

CaoCao has set a goal of deploying 100,000 Robotaxis and 100,000 Robovans cumulatively by 2030. As of the end of June 2026, the company had 140 second-generation Robotaxis deployed. The next phase of expansion depends on the Eva Cab entering mass production in 2027 as planned, and replicating the safety, dispatch, and operational processes validated in Hangzhou to other cities.

As fleet size grows, the growth rate of personnel and station investments needs to stay below the fleet growth rate to dilute per-vehicle operating costs.

According to Zhao's roadmap, the “double 100,000” target will proceed through three phases: model validation, mass production of purpose-built vehicles, and cross-city replication.

The industry is exploring different expansion approaches. Pony.ai uses a joint deployment model, while WeRide emphasizes a lightweight asset strategy overseas, with automakers, mobility platforms, and local operators sharing vehicle and operational investments.

CaoCao, leveraging Geely's manufacturing system and its own mobility platform, directly participates in vehicle definition, deployment, and fleet operations. Expanding both vehicles and infrastructure simultaneously increases capital requirements and raises the bar for asset operating efficiency.

The Robovan is the other half of the “double 100,000” goal. In July, CaoCao launched the first commercial Robovan operations in Changsha, proposing models such as vehicle sales, leasing, and robotics-as-a-service.

Zhao believes passenger and freight transport can share autonomous driving, dispatch, recharging, and maintenance systems, spreading underlying investments.

Going overseas also requires solving local operational challenges.

“When we go abroad, the first challenge we encounter may not be technical, but rather how to handle local operations and local compliance,” Zhao said.

Vehicle certification, licensing, insurance, accident liability, and data compliance must all meet local requirements. Zhao noted that through Geely's existing international dealer network and safety and certification systems, CaoCao has advantages in obtaining overseas vehicle access, and has already begun partnerships with local companies in the Middle East.

For CaoCao, moving from 140 vehicles to a much larger scale ultimately requires spreading the depreciation and operating investments from fleet growth across a continuously increasing base of paid orders.