When the Biden administration announced expansive new chip export controls this month, its main target was the Chinese military. 

A host of other organizations have been caught in the crossfire, however — including Pony.ai, an autonomous vehicle (AV) startup founded in California by two prodigious Chinese-born engineers. Whether the company can eventually succeed in bringing self-driving cars to the mass market may now depend not just on its technological prowess, but on its political acumen too. 

This week, The Wire looks at Pony.ai: the self-driving car startup’s emergence and warm welcome in China, and the challenges it faces amid U.S.-China technological decoupling. 

CALIFORNIA UPSTART

Pony.ai was founded in 2016 by James Peng and Lou Tiancheng, two engineers who formerly worked in Chinese internet search giant Baidu’s AV division. Peng holds a doctorate from Stanford and spent over a decade at Google and Baidu; Lou received his PhD at leading Chinese university Tsinghua, and is a two-time winner of Google Code Jam, a top international programming competition. The new company established a R&D center in Guangzhou, capital of southern Guangdong province, a year later. 

“You can’t talk about autonomous vehicles technology in China without talking about California when it comes to R&D,” says Michael Dunne, chief executive of ZoZoGo and ​​a longtime Asian auto industry consultant. The state is the U.S.’s epicenter for AV development, thanks to its concentration of industry specialists, relatively permissive regulatory environment and fair weather, making for ideal road testing conditions.

Pony.ai received its first permit to test its cars on public roads in 2017; a permit to test in Beijing came a year later. Pony.ai completed the third most testing miles on public roads among all AV companies in California in 2021, after industry leaders Waymo (backed by Google) and Cruise (backed by General Motors). 

The company’s investors include Sequoia Capital China, IDG Capital, 5Y Capital and Toyota, which invested $400 million in 2020. It was valued at $8.5 billion following a recent funding round in March. The company has raised $1 billion to date.

INDUSTRY DOUBTS

While some $100 billion has been invested into AV companies since 2010, according to McKinsey, a consultancy, there is growing skepticism that firms can ever put a truly autonomous car on the road. Public confidence in self-driving cars has declined in the U.S., thanks in part to embarrassing videos of cars flummoxed by navigational issues or involved in minor accidents. AV software remains vexed by endless bugs, due to the myriad road-related variables they need to account for.

Pony.ai hasn’t been immune. After a fleet car veered into a road center divider and traffic sign last year, California’s Department of Motor Vehicles suspended the company’s driverless testing permit. Another investigation this year found “numerous safety violations” on the driving records of Pony’s safety drivers, leading to the suspension of its other permit to test AVs with an on-board driver. 

China’s regulatory environment is moving in the opposite direction. “In the U.S., city officials’ first thoughts tend to be: what’s the risk of accidents? How trustworthy is this technology, and what does it mean for labor?” says ZoZoGo’s Dunne. “Whereas in China…city officials want autonomous vehicle technology to work, and they want to be the first city in the nation to host those [AVs].”

Some Chinese cities have already made themselves more hospitable to self-driving cars. Compared to the confusing streets of San Francisco, for example, Beijing’s AV testing district has designated lanes and pick-up and drop-off zones. Dedicated infrastructure like improved lane markings and cell towers make it easier for cars to navigate. 

“This is what China does really well,” says Dunne. “If [they] can’t build the software to account for all of the infinite complexities of the real world, then change the real world and make it something that’s more manageable.” 

In August, the Ministry of Transport released draft rules for self-driving cars, making China one of the first countries to release national regulations: At least eight major cities are currently trialing robotaxi services. Pony.ai has public road testing permits from all four of China’s tier one cities (Beijing, Shanghai, Guangzhou and Shenzhen). It was also one of the first companies, alongside Baidu, to receive a robotaxi operating license from Beijing’s municipal government in April this year.

COLLATERAL DAMAGE

Pony.ai’s biggest challenge will arguably be analyzing the data it collects from its vehicles’ drives. Autonomous vehicles rely heavily on artificial intelligence, which quickly interprets huge volumes of data generated by the car’s sensors and cameras, using powerful chips. Chipmakers have designed bespoke chips for use inside AVs: Pony.ai uses Nvidia’s DRIVE Orin chip, which has found a core market among Chinese electric vehicle makers.

In October, the Biden administration announced export controls on the sale of advanced computer chips used for AI. That followed a move in August to restrict the sale of specific advanced chips made by Nvidia and AMD, which Nvidia warned could reduce sales by $400 million in the third quarter alone. 

Pony.ai risks taking a beating from such restrictions. While a spokesperson for Nvidia confirmed to The Wire that export controls do not currently affect sales of its DRIVE Orin and other “automotive-grade in-vehicle solutions”, its A100 and H100 chips, widely used by Chinese AV companies, are specifically blocked from sale in China.

A spokesperson for Pony.ai declined to comment on whether U.S. export restrictions have affected the company. 

One way that Pony.ai could sidestep the fallout is by performing some of its data processing abroad. Doing so, though, would risk running afoul of China’s strict data security and localization regulations, given that AVs collect mapping data, photos and videos of other vehicles and their surroundings.

Some expect the dire situation for China’s AI-dependent firms could encourage authorities to make exceptions. 

“If a Chinese business decided that it was really necessary for them to transfer large amounts of data abroad for the purpose of accessing stronger computational capabilities that they’d be denied [at home] due to the U.S. limits, that would seem like a reasonable motivation,” says Graham Webster, editor-in-chief of the Stanford-New America DigiChina Project, which analyzes Chinese digital policy.

---