Robotaxi Is Actually Starting to Make Money—So What's the Next Revenue Stream?

A street shot of an Apollo Go driverless car. Image source / Wikimedia Commons.

In 2026, Apollo Go has already turned a profit in the single city of Wuhan.

This is the first fully driverless Robotaxi platform in China to publicly announce single-city profitability. The driver's seat is empty, the steering wheel turns on its own, it stops on its own at traffic lights, and in 20 minutes it takes you from outside your office building to the subway entrance. The safety driver that the industry could not get around for the past decade has finally been removed.

But being able to remove the person is the tipping point for this business.

For more than a decade, autonomous driving could not get around the safety driver in the driver's seat. As long as the safety driver was there, costs could not come down. In 2026, that barrier has been crossed: core cities including Beijing, Shanghai, Guangzhou, Shenzhen, and Chongqing have successively opened commercial operation permits for fully driverless Robotaxis, allowing vehicles without a safety driver in the main driver's seat to charge the public for passenger rides on public roads in central urban areas. The fully driverless fleets of leading platforms have already surpassed 1,000 vehicles, and in core operating areas, the cost per mile is already lower than that of traditional ride-hailing with a human driver.

The industry has truly moved from "demos with safety drivers" to "fully driverless, chargeable, and profitable." As early as May 2025, Goldman Sachs asserted that the per-vehicle gross margin of Robotaxis in first-tier cities would turn positive in 2026. Now it looks like the prediction is coming true.

The numbers have to be broken down by specific player to be seen clearly.

Just talking about a tipping point is too vague. Look at the latest data from several companies (as of October 2026):

Apollo Go (Baidu): Covers 28 cities globally; cumulative autonomous driving mileage exceeds 350 million kilometers (240 million+ fully driverless); peak weekly orders exceed 350,000; profitable in the single city of Wuhan. Data source / Baidu Apollo and Apollo Go public disclosures.

Pony.ai: Q3 total revenue of $25.4 million (up 72% year over year); Robotaxi revenue of $6.7 million (up 89.5% year over year, up 338.7% quarter over quarter); fleet of 961 vehicles, targeting 3,000 by year-end; gross margin doubled to 18.4%. Data source / Pony.ai 2026 Q3 financial report.

WeRide: More than 1,600 autonomous vehicles deployed globally (nearly 750 Robotaxis); obtained licenses in eight countries and operates in 30+ cities across 11 countries; Guangzhou Q3 order volume increased 4x quarter over quarter, Beijing increased 8x. Data source / WeRide 2026 Q3 operating announcement.

Tesla Cybercab: The production version officially began operation in Austin, Texas, in the United States on September 3, 2026, with unsupervised autonomous driving mileage already exceeding 1 million miles; after scaling, the cost per mile can be pushed down to $0.2. Data source / Tesla official and public reports.

Looking further out, there is an even bigger number: Frost & Sullivan predicts that China's Robotaxi market will reach $39 billion by 2030 and climb to $179.4 billion by 2035, accounting for more than half of the global total. Source / Frost & Sullivan industry report.

The tipping point has arrived so fiercely because three things matured at the same time.

A close-up shot of a driverless car's lidar and perception sensors. Image source / Wikimedia Commons.

Technology matured first. End-to-end large models became mainstream, allowing cars to drive based on "experience" like veteran drivers, greatly improving their ability to handle complex intersections and unexpected situations. Safety is no longer a single point of dependence either—lidar, cameras, and millimeter-wave radar provide multiple hardware backups for one another, and steering and braking are fully dual-redundant, so if one system fails, another immediately takes over. There is also vehicle-road-cloud integration, where roadside equipment gives vehicles super-line-of-sight "X-ray vision," the cloud monitors all vehicles in real time, and remote intervention is possible in extreme situations. Only with these three layers of protection stacked together did they dare to remove the person.