Pony.ai Says the Tech Is Solved. The Quarterly Loss Says the Business Isn’t.

(SeaPRwire) –   By: Oliver Hawthorne

James Peng said the technology problem for robotaxis is “pretty much already solved.” He said it confidently at the Leaders Forum in Macau on September 8. Within five years, he claims, you could hail a robotaxi just like a normal taxi. The room probably applauded. But here’s the anxiety nobody wanted to address. Pony.ai generated $12.1 million in robotaxi revenue in Q2. The operating loss in the same quarter was $65.7 million. You’re spending over five times what you’re earning on the core product line. The technology may be solved. The business model is very much not. The gap between what Peng is selling and what the numbers show defines the tension in autonomous mobility right now. Every CEO in this space claims the technology is ready. Every balance sheet says the economics aren’t. The question isn’t whether cars can drive themselves anymore. That argument is effectively over. The question is whether anyone can afford to keep doing it long enough for the math to work. Robotaxis are capital-intensive hardware plays disguised as software stories. And hardware plays require revenue density that the current market simply doesn’t provide. Twenty-five rides per day per vehicle is the operational number Pony.ai runs across Beijing, Guangzhou, Shenzhen, and Shanghai. It’s not a bad number. But it’s not a good number either. Traditional taxi operations in dense urban markets typically run at higher daily utilization rates than current robotaxi fleets. A robotaxi fleet needs near-continuous operation to justify the capital expenditure of vehicle ownership, maintenance, software updates, and insurance. Twenty-five rides means substantial portions of fleet time are idle. In a business where each vehicle represents a significant capital outlay, that idle time is existential. Peng says riders are “very diverse across different age groups, different professions.” He says it’s “definitely not something unique, just for the tech-savvy people to try.” That’s accurate but incomplete. Diversity in ridership is necessary. But the real metric is frequency. How many of those diverse riders return? The 700 percent year-over-year revenue growth sounds explosive. But 700 percent growth off a base of roughly $1.5 million means the company is still burning cash. That burn rate requires constant external financing.

The expansion strategy is where the story gets more interesting and more risky simultaneously. In August, Pony.ai announced an expanded partnership with Uber to deploy more than 2,000 robotaxes across Europe. They’ve already launched in Zagreb, Croatia. The South Korea agreement targets 200 robotaxes by 2028. There are also partnerships in the Middle East and Singapore. Peng says he looks for cities with relatively expensive taxi markets and favorable regulatory environments when choosing where to expand. Partnerships with Uber help foster trust and convince governments to embrace a driverless future. It’s a reasonable strategy. But Pony.ai is betting its European expansion on a platform that treats autonomous driving as a portfolio position. That’s not the same as a core commitment. Peng credited China’s “supportive” regulatory environment for the country’s autonomous driving strength. He said the key challenge is that it’s “such a complex system” requiring hardware, software, and talents. China provides a good system to support this kind of innovation, in his view. That assessment has real merit. Waymo dominates the American robotaxi conversation. But Chinese firms like Pony.ai, WeRide, and Baidu are gaining ground almost everywhere else. WeRide and Baidu are also pushing ventures into Southeast Asia and Europe. The global autonomous mobility map isn’t being drawn solely in Phoenix and San Francisco anymore. That’s a structural shift. Peng shared observations about user behavior that reveal both the promise and the rough edges. Female riders trend upward during evening hours because they “feel a lot safer because there are no drivers.” He described the experience as “safe, private, and, also, it’s consistent.” That’s a legitimate competitive advantage in ride-hailing markets where safety concerns suppress demand. But it’s a narrow advantage. Once robotaxes become normalized, safety perception shifts from differentiator to baseline expectation. More tellingly, in a March interview, Peng noted that passengers sometimes forget to close the car door when leaving. Without a human driver, Pony.ai ends up asking nearby delivery gig workers to close the door for them. That detail is small. But it’s revealing. A vehicle that can navigate highways and dense urban streets but can’t close its own door isn’t finished. Edge cases generate real costs. Human intervention, even minor, erodes the margin assumption that makes the entire business case work. Peng also talked about urban planning implications. Most private cars are used for two hours a day, he said. If robotaxes become ubiquitous, we’ll save on parking spaces. He said the implications “will be profound” and “will change the whole urban planning and change the way of our life.” Parking space optimization is a genuine economic benefit. But two hours of daily usage doesn’t automatically translate to fleet efficiency.

The end-game here is straightforward once you strip away the marketing. Robotaxis need three things simultaneously. First, capital density to fund years of operating losses. Second, utilization rates high enough to amortize hardware costs. Third, regulatory frameworks permissive enough to operate without human oversight across entire metropolitan areas. Pony.ai has the first partially covered through continued financing. It’s making progress on the second but not yet at the required level. The third is the most variable factor and depends entirely on local government willingness to greenlight operations. The Uber partnership is the most interesting piece of the puzzle. If Pony.ai can deploy 2,000 robotaxes across Europe through Uber’s existing infrastructure, it solves two problems at once. Demand acquisition and regulatory trust. Uber brings riders, brand recognition, and established regulatory relationships. Pony.ai brings the autonomous fleet. But this arrangement creates a dependency. If the platform strategy shifts, Pony.ai loses its primary European distribution channel. The alternative is building direct-to-consumer infrastructure in each European market, which is enormously expensive and slow. WeRide and Baidu are not idle. They’re also pushing internationally. The autonomous mobility space is heading toward consolidation. Five years from now, the number of independent robotaxi operators will be significantly smaller than today. Peng’s prediction that robotaxis will be commonplace within five years is probably technically accurate. The technology will support it. But “commonplace” doesn’t mean “profitable” or “independently operated.” It may mean “absorbed into the operations of a handful of surviving companies with deep pockets and favorable government relationships.” The real question for Pony.ai and every other player in this space is capital sustainability. Will the funding needed to cover the next three years of losses still be available? Will it be there when the technology delivers on its promise? Peng can say “technology is solved” with genuine confidence. The business case still has more unknowns than knowns. The next five years will separate companies that solved the engineering problem from companies that solved the money problem. Only one of those outcomes produces shareholders.

Author bio: Oliver Hawthorne, a principal correspondent permanently stationed at an international technology review, covering autonomous mobility, hardware infrastructure, and emerging transport economics.