China smart driving technology
The AI Chip Race: Li Auto’s Mach M100 Unveiled
China’s electric vehicle industry is undergoing a fundamental transformation from hardware-focused competition to software-defined intelligence. At the center of this revolution is Li Auto’s unveiling of the Mach M100 chip—a self-developed AI processor claiming 2,560 TOPS of total computing power for the production vehicle market.
The chip debuts in the new Li L9 Livis, marking Li Auto’s transition from an extended-range vehicle specialist to a full-stack AI company.
Mach M100 Technical Specifications
| Specification | Value |
|---|---|
| Process Node | 5nm automotive-grade |
| Architecture | Dataflow edge inference |
| Single-Chip TOPS | 1,280 |
| Dual-Chip TOPS | 2,560 |
| Latency Reduction | 40% improvement |
| Sensor-to-Action Delay | 200ms target |
Li Auto CEO Li Xiang described the Mach M100 as “the world’s most powerful computing solution for production vehicles”—a claim that, if validated, would surpass existing automotive AI processors from Mobileye, Nvidia, and Qualcomm.
XPeng VLA 2.0: The “DeepSeek Moment” for Autonomous Driving
Breaking Down VLA Technology
XPeng’s Vision-Language-Action 2.0 represents a paradigm shift in autonomous driving. Unlike traditional modular systems (perception → planning → control), VLA employs an end-to-end neural network approach:
- Vision: Raw camera inputs processed directly
- Language: Semantic understanding of driving context
- Action: Direct output of driving commands
CEO He Xiaopeng declared VLA 2.0 the “self-driving DeepSeek moment”—referencing the AI breakthrough that stunned the technology world. The key innovation: VLA 2.0 eliminates the “language translation” step, achieving direct vision-to-action generation for the first time.
XPeng’s Autonomous Timeline Revision
At the 2026 Xuanyuan Automotive Blue Book Forum, He dramatically revised XPeng’s autonomous driving timeline:
- Previous L5 doubt: “Until last year, I doubted L5 would be achievable in my lifetime”
- New timeline: L4 software capability by 2028
- L5 prototype: Expected by 2030
- Reason for acceleration: Data-as-fuel flywheel effect accelerating AI evolution by 6x
Geely’s Intelligent Driving: 110 Million Kilometers of Evidence
Qianli Haohan System Performance
Geely’s Qianli Haohan (Thousand-Mile Hero) intelligent assisted-driving system has accumulated compelling real-world data:
- Assisted-driving mileage: Over 110 million kilometers
- Equivalent to: ~2,750 circumnavigations of Earth
- Accident prevention: 225,000 potential accidents avoided through active collision-avoidance
- 2026 deployment: Level-3 highway autonomous driving and Level-4 low-speed functions where regulations permit
Geely Automobile Research Institute President Li Chuanhai emphasized: “Safety is the core of any intelligent driving system.”
Beijing Auto Show 2026: Intelligence Takes Center Stage
Key Autonomous Driving Exhibits
The 2026 Beijing Auto Show (Auto China) showcased China’s intelligent driving leadership:
BYD Flash Charging + Intelligent Driving
- World’s first vehicle series fully equipped with flash charging
- Fangchengbao Titanium 7: 755 km range with second-generation blade batteries
- Advanced driver assistance in all new models
Chassis-by-Wire: The New Frontier
- Li Auto L9 Livis: 800V architecture with steer-by-wire, four-wheel steering, electric mechanical braking
- Onvo L80: 900V platform with battery swapping capability
- Luxeed V9: Advanced driver assistance under Huawei’s HIMA ecosystem
Highway Autonomy Milestones
Regulatory momentum accelerated in late 2025, with China approving higher-level autonomous driving functionality on designated highways and urban routes.
Multi-Sensor Fusion vs. Pure Vision: China’s Distinctive Approach
The European-American Divide
While Tesla’s FSD relies on pure vision, Chinese manufacturers predominantly adopt multi-sensor fusion + vehicle-infrastructure cooperation:
è¡¨æ ¼
| Approach | Advantages | China Advantage |
|---|---|---|
| Pure Vision | Lower hardware cost, simpler design | Requires extensive training data |
| Multi-Sensor Fusion | Better in adverse conditions, redundancy | Complementary to 5G network coverage |
| V2X Cooperation | Enhanced situational awareness | Leading global 5G infrastructure |
This hybrid approach, leveraging China’s 5G network coverage and roadside infrastructure, demonstrates stronger scene adaptability than pure vision systems.
European Luxury Brands Adopt Chinese Smart Driving
The 2026 Surprise: BBA Goes Chinese
A remarkable development in 2026: European luxury manufacturers (BMW, Benz, Audi—collectively “BBA”) began adopting Chinese intelligent driving solutions for their China-market vehicles.
This represents a stunning reversal of the traditional technology transfer direction—from China adopting European technology to Europe adopting Chinese innovations.
Regional Regulatory Support
Robotaxi Deployment
Multiple Chinese cities have approved Robotaxi commercial operations:
- Beijing: Extensive autonomous ride-hailing zones
- Shanghai: Growing autonomous taxi fleet
- Shenzhen: Aggressive Robotaxi expansion
- Wuhan: Large-scale autonomous bus deployment
XPeng, Baidu Apollo, and Huawei-backed automakers are positioning for the autonomous mobility-as-a-service market.
FAQ
What is Li Auto’s Mach M100 chip computing power?
Li Auto’s Mach M100 delivers 1,280 TOPS per chip, with dual-chip configuration providing 2,560 TOPS total—claimed to be the world’s most powerful automotive AI processor.
When will XPeng achieve L4 autonomous driving?
XPeng CEO He Xiaopeng projects L4 software capability by 2028, with L5 prototype by 2030, accelerated by the data-as-fuel flywheel effect.
How does XPeng VLA 2.0 differ from Tesla FSD?
VLA 2.0 uses an end-to-end neural network with vision-language-action integration, eliminating intermediate language translation steps for direct vision-to-action generation.
What driving data has Geely accumulated?
Geely’s Qianli Haohan system has accumulated over 110 million kilometers of assisted-driving mileage, preventing 225,000 potential accidents.
Why are European brands adopting Chinese smart driving technology?
Chinese multi-sensor fusion and V2X cooperative driving approaches offer advantages in complex urban environments, and China’s 5G infrastructure enables capabilities difficult to replicate elsewhere.
About the Author
