Pony.ai’s next big ride: ​Replicating success across global cities

For James Peng, co-founder and CEO of Chinese autonomous driving technology company Pony.ai, the robotaxi race has entered a new stage.

Pony.ai has spent the past decade proving that fully driverless cars can operate in real cities, with its robotaxi operations achieving unit breakeven in Guangzhou in November 2025 and Shenzhen in February 2026, marking a significant milestone towards city-wide unit economics.

Now, the company is going through a critical phase centred on whether robotaxis can become a scalable and profitable global business.

Replicating success across cities

In my interview with Peng, he said the biggest challenge ahead is not simply about regulations, vehicle cost or consumer demand.

“It is whether we can integrate technology, mass production, operations, policy and commercial partnerships, and efficiently replicate a model that has already been proven in one city across more cities,” he said.

That, in his view, is the real robotaxi race. “Robotaxi is a complex system. A breakthrough in any one part is not enough to achieve true scale.”

Its commercial autonomous fleet has reached roughly 2,000 vehicles globally as of mid-2026, with a target of more than 3,500 robotaxis across more than 20 cities by the end of this year.

The company has a market capitalisation of about US$2.2 billion as of end October 2026. In the second quarter of 2026, total revenue reached US$36.2 million, while robotaxi revenue hit a record US$12.1 million.

But for Peng, fleet size alone is not the measure of success. “3,500 vehicles is an important milestone for scale, but reaching a certain fleet size does not automatically bring sustainable profitability,” he said.

“What really determines profitability is the number of orders each vehicle receives, service hours and utilisation, fleet density in each city, and whether vehicle costs and daily operating costs can continue to decline.”

From profitable car to profitable network

Peng described Pony.ai’s path to profitability in three stages. The first is making the economics of each vehicle work. “Guangzhou and Shenzhen have already demonstrated this,” he said.

The second is to replicate the model across more cities, allowing the robotaxi business to cover the cost of fleet operations and expansion.

Only then comes the third stage: sustainable profitability at the company level. “Profitability is not about achieving accounting profit in a single quarter because of special factors,” Peng said.

He wants to reach a point where “adding another vehicle no longer means losing money”, while positive unit economics can be repeated across multiple cities.

At the same time, Pony.ai wants recurring revenue from its joint deployment model — bringing together autonomous driving technology, access to rider demand and day-to-day fleet operation — to continue growing and that includes technology licensing and revenue sharing.

“Our path is very clear,” Peng said. “First, make each vehicle profitable. Then make the robotaxi business profitable. Finally, move towards sustainable profitability at the company level.”

That is a much bigger challenge than simply putting thousands of driverless cars on the road.

Pony.ai does not want to own everything

One way the company plans to scale the robotaxi business is through partnerships.

Pony.ai is currently working with Toyota, BAIC and GAC; BAIC and GAC are two of the largest, state-owned automotive manufacturing conglomerates in China. Together with Toyota, Pony.ai plans to produce more than 1,000 Gen-7 bZ4X robotaxis in 2026.

The idea is not for Pony.ai to own every vehicle, manage every fleet or carry the entire cost of expansion itself. “The core idea is to allow every participant in the value chain to play to its strengths,” Peng said.

The economics will differ from market to market, but the basic principle remains the same. “Each party’s return should match the value it creates, the investment it makes and the responsibilities it takes on.” he said.

For Pony.ai, that means earning returns through vehicle and system delivery, technology licensing and operating revenue sharing.

The company does not see itself simply as a robotaxi operator, nor as a passive software company. “If I had to define Pony.ai’s role, I would say we are an autonomous driving technology company centred on the Virtual Driver, with deep commercialisation and operating capabilities,” Peng said.

Pony.ai’s strengths

As autonomous driving technology develops, some hardware and individual algorithms will become more standardised.

What is difficult to copy, Peng said, is not just technology, but an ecosystem. “A fully driverless robotaxi fleet that can operate safely and reliably in complex urban environments, while also achieving sustainable economics — this is still far from becoming a standardised product,” he said.

Going from one test vehicle to hundreds or thousands of commercial vehicles requires vehicle redundancy, hardware and software integration, safety validation, mass production, remote assistance, dispatching, maintenance and daily passenger service.

“These capabilities cannot be obtained quickly by simply purchasing one software module,” he said.

Pony.ai’s advantage, Peng argued, comes from connecting all of these pieces together seamlessly. “Our strength is not just the Virtual Driver, data or operating experience individually,” he said, “It is the closed loop formed by these capabilities.”

Pony.ai has its Level 4 Virtual Driver and PonyWorld world model, while its real-world fully driverless fleet provides a constant testing ground.

This Level 4 Virtual Driver refers to the full suite of software, hardware and artificial intelligence that acts as the complete chauffeur for a self-driving vehicle, requiring no human intervention within specific geographic areas or conditions.

According to industry data, only about ten to 15 major autonomous vehicle operators globally have successfully secured Level 4 driverless operational licenses.

Also, working with multiple automakers, Peng added, can strengthen rather than weaken that advantage because it shows Pony.ai can adapt its technology to different vehicle platforms.

Overseas markets will test the model

Pony.ai is expanding across Europe, the Middle East and Asia, with its latest Uber partnership potentially bringing more than 2,000 robotaxis to Europe.

China, Peng said, has provided an unusually demanding environment in which to build the business. “China’s first-tier cities have high road density, complex traffic participants and concentrated travel demand,” he said.

“At the same time, the market has very high requirements for cost and operating efficiency. If a robotaxi system can achieve all-day operations, a good user experience and positive unit economics in such an environment, it provides a very solid foundation for entering other markets.”

But Peng is clear that Pony.ai cannot simply take the China model and replicate it overseas. “What can really be replicated is not a particular set of traffic rules or the exact same operating model,” he said.

“It is the underlying technology and methodology.”

That includes the Virtual Driver’s ability to generalise across environments, vehicle integration, safety validation, remote assistance, fleet dispatching, charging and maintenance.

His approach is therefore to keep the technology core evolving while working with local mobility and operating partners.

“China validates our capabilities,” he said. “Overseas deployment must be built on localisation.”

Why Singapore matters to Pony.ai

Pony.ai entered Singapore with transport group ComfortDelGro and began autonomous vehicle operations in March 2026. In June, its service in the Punggol area became available to the public through ComfortDelGro’s Zig app.

For Peng, however, Singapore’s value goes well beyond the number of robotaxi rides it can generate. “Singapore’s strategic value has never been simply about the number of mobility orders,” he said.

The city-state offers high safety and regulatory standards, mature public transport, advanced infrastructure and an environment that can support the deployment of new technology. Operating in Singapore also gives Pony.ai a very different set of conditions to test, to adapt and to tailor to the Singapore market.

“These include left-side traffic, a tropical climate, a high-density urban environment and how robotaxis can connect with the MRT, buses and community transport,” Peng said. This means that the driverless software must be tuned to operate in heavy rain and heavy traffic environments and see the dynamic road situation from a driver seat that is on the right side of the car.

“These experiences help us further improve the generalisation and localisation capabilities of our Virtual Driver in different regions.”

Singapore can also serve as a bridge to the wider Southeast Asian ecosystem. “Many regional companies, mobility platforms, investment institutions and technology talent are gathered in Singapore,” he said.

Its partnership with ComfortDelGro, he added, is therefore not just about understanding Singapore. It can help Pony.ai build a wider network of partners and operating experience across Southeast Asia.

As to how many robotaxis must Pony.ai have before it will decide whether Singapore is ready for the next step? Peng said there is no single number. “First, safety and service reliability must be verified through continuous real-world operations,” Peng said.

Then comes economics.

Pony.ai will monitor vehicle utilisation, fleet density, remote-assistance efficiency, maintenance and overall operating costs as the service expands.

And finally, the wider ecosystem must be ready, including regulations, local operations, vehicles, charging and maintenance infrastructure, and integration with public transport and mobility platforms.

“The current Punggol service through the Zig app provides a very important real-world environment to evaluate these indicators,” Peng said.

“If safety performance, user acceptance and operating efficiency can continue to improve as operations and scale expand, we will discuss with ComfortDelGro, regulators and other relevant parties the possibility of a larger fleet, more routes and a wider service area.”

Node to the world

For Pony.ai, Singapore could therefore become more than a robotaxi market. Peng sees it as a potential “high-quality technology validation market”, a trusted regional demonstration market and an important node connecting Pony.ai with Southeast Asia’s wider ecosystem.

“I hope Pony.ai can become a broad autonomous driving and physical AI platform centred on the Virtual Driver,” Peng said.

That does not mean expanding into every possible application. “We will still focus on transportation and logistics scenarios where we can fully use our technology strengths, achieve scale and build sustainable commercial value.”

“Capital can accelerate development, but it cannot replace accumulated technology capabilities and real-world operating experience,” Peng said.

Pony.ai’s next chapter is therefore not simply about putting more driverless cars on the road. It is about turning those cars into a global business, one city at a time.

Related: China’s autonomous driving race shifts from hype to high stakes | How China’s self-driving cars could redefine Europe’s tech rules