Pony.ai and Moore Threads: Self-Driving Cars Move to Chinese Hardware

While Western giants argue about autopilot safety, China has decided it is time to stop depending on outside supplies of computing power. Pony.ai, one of the most visible players in the robotaxi market, has announced a strategic partnership with Moore Threads.

If you don't follow the semiconductor market, Moore Threads is the company trying to build a full-fledged Chinese alternative to NVIDIA. Together, they will now work to refine Level 4 (L4) autonomy, where a car handles the road without human involvement in virtually any conditions. The deal is not about simply buying graphics cards.

The companies plan to build what they call a closed loop of "algorithms, data, compute, applications." Pony.ai is moving the training of its systems onto MTT S5000 accelerators and KUAE intelligent computing clusters.

This is a serious challenge for software: algorithms honed for years on the CUDA architecture must be adapted to new, if ambitious, Chinese hardware. But the game is worth the candle, because it gives Pony.ai direct access to optimization at the silicon level.

The announcement paid particular attention to so-called "world models." In autonomous driving, this is currently the dominant trend. Instead of merely reacting to obstacles, a neural network learns to predict the physics of the surrounding world and the behavior of other road users.

Such models demand enormous computing resources. Moore Threads claims its clusters can handle the task no worse than Western counterparts, while delivering significant cost savings as a robotaxi fleet scales. Why does this matter right now?

The autonomous driving industry has hit a glass ceiling on cost. To bring L4 to streets en masse, the price of the car's "brain" and the cost of training it must fall radically. Switching to local solutions such as the MTT S5000 is not only a matter of political security and protection from sanctions, but also an attempt to make the technology economically viable.

If Pony.ai can confirm the claimed specifications, we could see a mass exodus of Chinese tech companies from NVIDIA platforms to domestic solutions. The Pony.

ai and Moore Threads tie-up also shows that the time of pure software companies in AI is passing. The future belongs to those who control the whole technology stack. Pony.

ai stopped being just a startup with good coders long ago; it is now an organization that influences chip development and server architecture. For Moore Threads, this is also a critical moment: if its chips withstand the load of real driverless traffic in megacities, questions about the competitiveness of Chinese GPUs will answer themselves. The bottom line: Pony.

ai is betting on import substitution at the very heart of its technology. Will Moore Threads deliver the same stability as NVIDIA, or will we see L4 development delayed by the teething problems of new hardware?

*Meta has been designated an extremist organization and is banned in Russia.