Gasgoo Munich- On February 6, Pony.ai announced a strategic partnership with Moore Threads, a Chinese developer of full-featured GPUs. The collaboration centers on advancing L4 autonomous driving toward large-scale deployment, with both sides working closely on training and optimizing Pony.ai's core technologies—its world model and virtual driver system—powered by secure and reliable AI computing to accelerate both technological iteration and commercial rollout.
Image source: Pony.ai
The partnership is widely seen as a milestone in China's AI industry collaboration. It marks the first time Pony.ai has adopted China's domestically developed AI computing at scale for critical training and simulation workloads, while also signaling Moore Threads' formal entry into the core computing stack of autonomous driving.
Leveraging Moore Threads' MTT S5000 training-and-inference integrated computing cards and the KUAE intelligent computing cluster, the two companies will jointly adapt and validate training for Pony.ai's world model and in-vehicle models. By co-building an autonomous driving computing ecosystem, they aim to create end-to-end synergy across algorithms, data, computing power and applications—shortening development cycles, lowering costs and supporting high-quality growth in smart mobility and logistics.
Backed by its fully self-developed PonyWorld model and Virtual Driver system, Pony.ai has built one of the industry's most advanced L4 autonomous driving capabilities and has become the first company in China to operate fully driverless Robotaxi services across four tier-one cities. Founder and CEO Peng Jun noted that the reinforcement learning–based world model can generate more than 10 billion kilometers of test data each week, creating hundreds or even thousands of high-risk scenario variations. This allows the virtual driver to evolve through repeated training cycles—something human drivers cannot match—forming the basis for safety levels claimed to far exceed those of human driving.
Moore Threads, one of China's few providers of full-function GPUs, has demonstrated rapid innovation through what it calls a “five chips in five years” pace, completing four generations of GPU architecture upgrades. Its portfolio now spans AI, scientific computing and graphics rendering, enabling cloud, edge and device-level applications. The KUAE computing cluster built on the MTT S5000 platform has reached internationally competitive levels in training models with hundreds of billions or even trillions of parameters, while its strengths in graphics rendering support high-fidelity simulation, scenario reconstruction and visualization for autonomous driving.
The collaboration aligns closely with Pony.ai's defined roadmap toward L4 commercialization. By the end of last year, its Robotaxi fleet had grown to 1,159 vehicles, surpassing annual targets, while its seventh-generation Robotaxi system achieved positive unit economics in Guangzhou—an important milestone toward a sustainable business model.
Looking ahead, Pony.ai plans to further scale its Robotaxi strategy, targeting deployment of more than 3,000 vehicles by the end of 2026. The partnership with Moore Threads is expected to materially improve model training efficiency and system performance, providing a strong acceleration engine for that expansion.
