Pony AI Turns Up the Compute, Cuts the Driver

Pony AI's NVIDIA-powered domain controller hits 4,000 TFLOPS, as the company targets 3,000 robotaxis across 20 cities by end of 2026

6 May 2026

Autonomous computing just shifted up a gear. On 25 April 2026, Pony AI unveiled a next-generation domain controller developed with NVIDIA, delivering up to 4,000 FP4 TFLOPS of onboard compute power for fully driverless commercial vehicles. Built on NVIDIA's DRIVE Hyperion platform and powered by two DRIVE Thor chips linked via NVLink, the system is engineered to handle multi-sensor fusion, full-scenario perception, and the complex real-world conditions that Level 4 autonomy demands.

Commercially, the ground is already moving. Shipments of Pony AI's predecessor "Fangzai" controller surged more than 500% year-over-year in 2025, reaching customers across Germany, the UK, South Korea, Japan, and Switzerland. Breakeven on robotaxi unit economics has been achieved in two major Chinese cities, a milestone that separates credible commercialization from subsidized pilots. Targeting a fleet of more than 3,000 vehicles across more than 20 cities globally by end of 2026, Pony AI is pressing the advantage hard.

Powering this ambition is a hardware architecture built for scale. NVIDIA's DRIVE Hyperion platform is becoming the de facto compute spine of the autonomous driving industry, adopted across multiple AV operators and OEMs. For Pony AI, integrating DRIVE Hyperion into its next-generation controller aligns its own robotaxi platform with an open ecosystem capable of supporting third-party deployment across logistics, mining, shuttles, and robosweeping, diversifying revenue well beyond passenger rides.

Challenges remain real. Regulatory approval for driverless operations in the United States and Europe is slow and jurisdiction-specific, while safely validating rare edge-case scenarios continues to test even the most advanced systems. Progress, though, is no longer incremental.

With production-grade hardware now clearing the compute thresholds that L4 autonomy requires, the gap between today's pilots and tomorrow's scaled driverless networks is closing. Pony AI and NVIDIA are building the infrastructure to cross it.