Published on Feb. 21, 2026

Pony.ai, one of the three leading robotaxi developers in China, has for the first time adopted Chinese AI computing technology to train its models. On February 6, 2026, the company announced a strategic partnership with Moore Threads, the Shanghai-based chipmaker, the Chinese news portal Gasgoo reported.

This new partnership is not replacing Pony.ai's longstanding cooperation with Nvidia, but rather adds a second, dual track to reduce the company's reliance on US-made computing.

The Chinese autonomous driving provider has been using Nvidia DRIVE Orin SoC (system-on-chip) technology for many years. Since 2022, it has been mass-producing autonomous driving controllers built on Nvidia's DRIVE Hyperion infrastructure.

Pony.ai will use Moore Threads' integrated computing cards and its KUAE intelligent computing cluster to train its world model and in-vehicle models, using the Chinese-made technology for simulation workloads.

While there is currently no replacement for Nvidia, this new business for Moore Threads still represents progress for China's efforts to build a domestic ecosystem where AI developers, chipmakers and automotive companies actively explore their own synergies.

“This strategic partnership with Pony.ai marks the first deep collaboration between domestically developed full-function GPU computing power and leading autonomous driving algorithms", Moore Threads CEO Zhang Jianzhong was quoted on the website of the Guangzhou Municipal Science and Technology Bureau.

Seen through the lens of the AI race between the United States and China, the partnership between Pony.ai and Moore Threads can be understood as another example of China's strategic pursuit of domestic substitution, slowly reducing Washington's leverage over its AI industry amid frequent US restrictions on exports of Nvidia AI chips to China.

The partnership is also a prime example of the pragmatism with which China is pursuing this upgrading of its own industries. Partnerships with US and other foreign companies remain intact and welcome. The Chinese remain open to partnerships and investments in their country, learning about new technologies in the process, while aiming to build up domestic capabilities step-by-step.

Zhang Jianzhong, also known as James Zhang, the founder and CEO of Moore Threads, is a former vice president at Nvidia and general manager of its China operations. Now, empowered by Beijing's wish to counter Washington's trade and technology wars, he has become a dollar billionaire building exactly the kind of chips he used to sell for Nvidia.

Pony.ai likewise is a very international-minded company, which has been founded in Silicon Valley. Its founder, Peng Jun, an expert in autonomous driving, has worked at Alphabet in the past before moving on to Baidu and then founding his current company, which is backed by investment from Baidu. Now, the AI race is also generating a lot of new support for his ambitions at home.

Leaving the tech rivalry aside and turning to the technological and market implications of this new cooperation, its significance lies in its potential to speed up the deployment of L4 autonomous driving.

The GPUs from Moore Threads will now further empower Pony.ai's reinforcement-based world model, which can already generate 10 billion kilometres of test data each week, simulating hundreds of high-risk scenarios, the algorithm developer said in its press release.

In the end, real world-data is what really counts, and in that context Pony.ai's fleet of robotaxis is important. By the end of last year, its fleet on the ground had grown to 1,159 vehicles in several Chinese cities. Until the end of this year, the company wants to expand this fleet to 3,000 vehicles.

"The partnership with Moore Threads is expected to materially improve model training efficiency and system performance, providing a strong acceleration engine for that expansion", Gasgoo commented. At the same time, Pony.ai is also speeding up the production of robotaxis together with its partner Toyota, which is one of its shareholders.

A number of factors are currently joining forces that make analysts predict a rapid growth of robotaxi fleets and the large-scale adoption of L4 autonomous driving in passenger cars in China in the near future: progress in large AI models, the fast electrification of the whole automobile industry and the willingness of Chinese consumers to embrace emerging technologies and hail a driverless taxi on their smartphone apps.

A prognosis published by UBS foresees up to 300,000 driverless taxis operating in China's four top-tier cities as early as 2030, and a total of 4 million robotaxis all over China by the late 2030s.