Pony.ai announced a next-generation autonomous driving domain controller built on NVIDIA DRIVE Hyperion, powered by dual NVIDIA DRIVE AGX Thor chips connected via NVIDIA NVLink. The new platform delivers a combined maximum computing performance of 4,000 FP4 TFLOPS (floating-point operations per second), enabling the company to scale its robotaxi operations and expand its domain controller business to customers worldwide.

What Is a Domain Controller and Why Does It Matter?

A domain controller is essentially the brain of an autonomous vehicle. It processes data from multiple sensors, fuses that information together, and makes real-time decisions about steering, acceleration, and braking. The more computing power a domain controller has, the faster and more accurately it can handle complex driving scenarios. Pony.ai's new system is designed to handle multi-sensor fusion, full-scenario perception, and high-complexity scenario understanding while meeting the demanding safety requirements of Level 4 (L4) autonomous driving, which means the vehicle can operate without human intervention in most conditions.

The collaboration between Pony.ai and NVIDIA dates back to 2017. In 2022, Pony.ai launched its first in-house automotive-grade computing unit powered by single or multiple NVIDIA DRIVE AGX Orin chips, which were deployed in its sixth-generation robotaxis. By 2025, the company began mass production of the world's first L4 robotaxi domain controller equipped with four NVIDIA DRIVE AGX Orin chips, which now powers its seventh-generation (Gen-7) robotaxis.

How Is Pony.ai Scaling Its Autonomous Vehicle Business?

Pony.ai's strategy combines hardware innovation with software expertise. As one of the few L4 autonomous driving companies globally with full-stack, in-house capabilities across both software and hardware, the company uses a software-hardware co-design approach to optimize performance, safety, and cost. This means engineers work together from the start to ensure the computing hardware and driving software work seamlessly together.

• Robotaxi Fleet Expansion: Pony.ai aims to expand its robotaxi fleet to more than 3,000 vehicles and its geographic footprint to more than 20 cities globally by the end of 2026, building on recent unit-economics breakeven achievements in two major Chinese metropolitan markets.
• Domain Controller Business Growth: Shipments of Pony.ai's "Fangzai" domain controller surged by more than 500 percent year over year in 2025, with customers spread across dozens of countries including Germany, the United Kingdom, South Korea, Japan, and Switzerland.
• Diverse Application Portfolio: Beyond robotaxis, Pony.ai's domain controllers are being deployed in low-speed delivery, robosweeping, logistics, mining, autonomous shuttles, and other robotics and intelligent mobility applications.

The new NVIDIA DRIVE Hyperion platform supports flexible single-chip and multi-chip configurations, enabling deployment across a wide range of autonomous applications. This modularity means customers can choose the computing power they need for their specific use case, whether that is a low-speed delivery robot or a full-scale robotaxi.

What Makes the New NVIDIA DRIVE Hyperion Platform Different?

The key innovation is the use of NVIDIA NVLink, a high-speed interconnect that allows two DRIVE Thor chips to communicate with extremely low latency. Think of it as a super-fast highway between two brains, allowing them to share information and coordinate decisions almost instantaneously. This architecture delivers significant gains in AI computing performance and energy efficiency compared to previous generations.

The platform is engineered to further enhance safety redundancy, system robustness, and deployment flexibility. In autonomous driving, redundancy is critical; if one computing system fails, a backup must be ready to take over instantly. The new design addresses this requirement while also supporting the latest AI models that Pony.ai and its customers are developing.

"Our collaboration with NVIDIA has supported several critical milestones in Pony.ai's autonomous driving journey. The next-generation domain controller built on NVIDIA DRIVE Hyperion will be a key enabler for the continued evolution of our L4 autonomous driving products and help accelerate large-scale commercialization," said Dr. James Peng, Founder and CEO of Pony.ai.

Dr. James Peng, Founder and CEO of Pony.ai

NVIDIA's automotive leadership team echoed this sentiment. Rishi Dhall, Vice President of Automotive at NVIDIA, noted that autonomous driving systems are rapidly increasing in complexity, driving the need for scalable, high-performance compute platforms. He highlighted Pony.ai's long-standing partnership with NVIDIA and expressed confidence in the new platform's ability to support advanced L4 autonomous driving applications.

"Autonomous driving systems are rapidly increasing in complexity, driving the need for scalable, high-performance compute platforms. Pony.ai has been a long-standing NVIDIA DRIVE customer and ecosystem partner, and we're pleased to see them build their next-generation domain controller on NVIDIA DRIVE Hyperion with DRIVE Thor accelerated compute to support advanced L4 autonomous driving applications," explained Rishi Dhall, Vice President of Automotive at NVIDIA.

Rishi Dhall, Vice President of Automotive at NVIDIA

Why Does This Matter for the Autonomous Vehicle Industry?

Pony.ai's announcement signals that the autonomous vehicle industry is moving beyond experimental prototypes toward real-world commercialization. The company has already achieved unit-economics breakeven in two major Chinese cities, meaning it is operating robotaxis profitably on a per-vehicle basis. This is a significant milestone that many competitors have not yet reached.

The expansion of Pony.ai's domain controller business to customers across Europe, Asia, and the Middle East demonstrates that autonomous driving technology is becoming a platform, not just a single-company product. Other companies can now license Pony.ai's hardware and integrate it with their own software, accelerating the adoption of autonomous vehicles across different industries and geographies.

The 4,000 FP4 TFLOPS computing performance represents a substantial leap in processing capability. To put this in perspective, this level of computing power allows the system to process vast amounts of sensor data from cameras, lidar, and radar in real time, enabling the vehicle to perceive its environment with high precision and make split-second driving decisions. The energy efficiency improvements also mean that robotaxis can operate longer on a single charge, reducing operational costs.

As autonomous vehicle technology matures, the competition is shifting from who can build the best software to who can build the most efficient, scalable, and reliable hardware-software systems. Pony.ai's investment in in-house domain controller design, combined with NVIDIA's cutting-edge computing platforms, positions the company to compete effectively in this evolving landscape. The announcement of the NVIDIA DRIVE Hyperion platform represents a critical step forward in making autonomous vehicles a practical reality for millions of people worldwide.