China's BYD Launches Chip for Autonomous-Driving Technology

“BYD unveiled an autonomous-driving chip, as the Chinese automaker known for its competitively priced electric vehicles seeks a new growth driver in technology.

 

The company launched the Xuanji A3 chip, a 4 nanometer semiconductor for autonomous driving, late Thursday at its technology day in Shenzhen, China.

 

The chip has already entered mass production and is designed to support level 3 and level 4 autonomous driving, the company said, referring to cars that can handle all aspects of driving under specific conditions and cars capable of driving in specific areas without human intervention, respectively.

 

The launch marks BYD's latest effort to showcase its edge in technology as growth in China's EV market slows and pricing pressure weighs on profitability across the sector.

 

BYD also said it aims to invest more than 100 billion yuan, equivalent to $14.75 billion, in intelligent technology research and development during the next three years.

 

The total hardware cost of the new chip is roughly one-third of Nvidia's Thor-based solutions [1], Citi analysts wrote, adding that it would ease the cost pressure of high-end intelligent-driving hardware.

 

As China's EV sales champion has faced many months of slowing demand in its home market and weaker profit in the first quarter, it has sharpened its focus on technology and overseas expansion.

 

BYD is working on making its advanced autonomous-driving technology more accessible for more models. The company said last year that it would deploy its "God's Eye" driver-assistance system in its mass-market models.

 

At the event, BYD said buyers of cars equipped with certain versions of the "God's Eye" system would receive one year of full compensation for any losses arising from accidents during assisted-driving operations, including vehicle damage and third-party liability.

 

Citi viewed the move as significant because it effectively shifts part of the liability burden to the automaker from drivers under China's current level 2 assisted driving regulatory framework.

 

BYD's latest strategy could reshape competition across China's intelligent-driving supply chain and pressure rivals from EV startups to chip makers and autonomous-driving technology providers, Citi said.

 

Chinese EV makers including NIO, Li Auto and XPeng also are developing their own autonomous-driving chips.

 

BYD's shares closed down 2.1% in Hong Kong on Friday, underperforming the benchmark stock index's 0.7% gain.” [2]

 

1. Nvidia's Thor-based architectures—built on the Blackwell platform—are centralized processors delivering up to 2,070 FP4 TFLOPS of AI compute (Trillion Floating-Point Operations Per Second). They power real-time AI reasoning for the automotive sector and physical AI/robotics. General "Robot Brain" Hardware: Nvidia prices the Jetson AGX Thor T5000 developer kits for autonomous testing at $3,499 each. For bulk production module orders exceeding 1,000 units, the price drops to $2,999 per module

Core solutions and developer kits include:

 

1. Automotive & Autonomous Vehicles

•     DRIVE AGX Thor: A centralized in-vehicle supercomputer consolidating traditional electronic control units (ECUs) into a single system. It unifies automated driving, parking, and infotainment.

 

•           Performance: It uses up to 14 Arm Neoverse V3AE cores to deliver more than 1,000 INT8 TOPS (2,000 FP4 FLOPs) while remaining compliant with ASIL-D functional safety standards.

•           Adoption: Many automakers use this technology to power their next-generation software-defined vehicles (SDVs).

2. Physical AI & Robotics

•           Jetson AGX Thor: Modules like the T5000 and T4000 are designed for edge operation and complex physical AI.

•           Capabilities: The developer kit provides 128 GB of LPDDR5X memory and uses Multi-Instance GPU (MIG) for secure hardware partitioning.

•           Real-World Applications: Deployed in humanoid robots (such as Boston Dynamics Atlas and Agility Robotics Digit), medical robots, and autonomous industrial platforms.

3. Industrial & Medical Edge

•           IGX Thor Platform: An industrial-grade edge platform (modules like the IGX T5000 SoM and IGX T7000).

•           Use Case: Brings server-level AI compute to mission-critical environments, such as real-time surgical assistance systems and AI-assisted medical diagnostics

 

2. EXCHANGE --- China's BYD Launches Chip for Autonomous-Driving Technology. Huang, Jiahui.  Wall Street Journal, Eastern edition; New York, N.Y.. 30 May 2026: B9.