๐ญ๐ฐSCMP TechnologyโขStalecollected in 1m
Chinese EV Makers Pivot from Price Wars to AI
๐กUnderstand how Chinese EV giants are pivoting to AI to survive market saturation and regulatory shifts.
โก 30-Second TL;DR
What Changed
Chinese EV market competition is moving away from price-cutting strategies.
Why It Matters
This shift signals a major R&D investment cycle in computer vision and sensor fusion for the Chinese automotive sector. AI practitioners should anticipate increased demand for edge-computing talent and autonomous driving software stacks in the region.
What To Do Next
Analyze the sensor fusion architecture requirements for L3 autonomous systems to identify gaps in current open-source perception models.
Who should care:Developers & AI Engineers
Key Points
- โขChinese EV market competition is moving away from price-cutting strategies.
- โขCarmakers are prioritizing AI-driven autonomous driving features to differentiate products.
- โขThe industry is targeting the deployment of Level 3 (L3) conditionally autonomous driving systems.
- โขStrategic shift is a response to cooling demand and stricter regulatory environments.
๐ง Deep Insight
Web-grounded analysis with 19 cited sources.
๐ Enhanced Key Takeaways
- โขChina's Ministry of Industry and Information Technology (MIIT) issued the first Level 3 (L3) autonomous driving certifications in December 2025 to Changan Auto and BAIC Motor's Arcfox, allowing them to operate on public roads in specific areas.
- โขNew draft safety standards from February 2026 significantly raise the baseline for L3 systems, requiring them to execute minimal risk maneuvers independently if the driver fails to respond to takeover requests, effectively pushing L3 capabilities closer to Level 4 (L4).
- โขBYD has unveiled its Xuanji A3, China's first automotive-grade 4nm self-driving chip, which is already in mass production and supports L3 and L4 autonomous driving functions.
- โขBYD has committed to assuming full financial liability for traffic accidents caused by its urban navigate-on-autopilot system (God's Eye) in China, a strategic move to build consumer trust and differentiate its technology.
- โขThe end of the EV price wars in China, partly due to government intervention barring sales below manufacturing cost, has accelerated the shift towards technological differentiation, particularly in AI and autonomous driving features, to stimulate demand through product strength.
๐ Competitor Analysisโธ Show
Chinese EV Makers and AI/Autonomous Driving Comparison
| Feature / Company | BYD (Xuanji A3 / God's Eye) | XPeng (XNGP) | Li Auto (AD Max) | NIO (NOP+) | Huawei (ADS) | Tesla (FSD) |
|---|---|---|---|---|---|---|
| Approach | Vertically integrated, in-house chip, full liability for ADAS accidents. | Full in-house software/hardware, aggressive urban scenario rollout. | In-house development, focus on practical application & safety, ADAS as standard. | Full in-house R&D, tightly coupled with vehicle, battery swap integration. | Full-stack hardware & software solution provider (Huawei Inside). | Vision-only approach using cameras and neural networks. |
| Key Hardware/Chips | Xuanji A3 (4nm, 700 TOPS single, 2100 TOPS triple), LiDAR option across all models. | Dual NVIDIA Orin-X (508 TOPS), two LiDAR units. | Dual NVIDIA DRIVE Orin SoCs (508 TOPS), camera-first with LiDAR backup. | Adam supercomputer (quad NVIDIA DRIVE Orin SoCs, 1016 TOPS), LiDAR. | Ascend 610 chips (400 TOPS total for ADS 1.0), proprietary high-resolution LiDAR. | Proprietary chips, vision-only sensors. |
| L3/L4 Support | Supports L3/L4. | Targeting L4 by 2026, L3 test license obtained. | L3 capabilities. | Approved for L3 autonomous driving. | ADS 3.0 supports city NCA from parking to parking. | Requires active human intervention, marketed differently in China. |
| Deployment Status | God's Eye expanding to mass-market models, 3.15M vehicles with ADAS hardware. | XNGP available in dozens of cities for point-to-point assisted driving. | NOA rolling out across L-series SUVs. | NOP+ works in 726 cities. | ADS 2.0 realized city NCA across China by end of 2023. | Launched FSD in China with limitations. |
| Liability/Safety | Full damage coverage for urban NOA accidents. | - | - | - | - | - |
| Cost (ADAS option) | 12,000 yuan ($1,770) for LiDAR-equipped God's Eye B. | - | - | - | ADS SE: 5,000 yuan upfront or 100 yuan/month. | - |
๐ ๏ธ Technical Deep Dive
- BYD Xuanji A3 Chip: China's first automotive-grade 4nm self-driving System-on-a-Chip (SoC). A single Xuanji A3 delivers 700 TOPS (Tera Operations Per Second) of computing power, with a cluster of three chips reaching 2,100 TOPS, sufficient for Level 3 and Level 4 autonomous driving functions. It boasts the lowest power consumption per unit of compute in its class, drawing approximately 20% less than comparable semiconductors. The chip is the centerpiece of a new central computing platform that unifies smart cockpit, driver-assistance, and electric propulsion domains.
- Redundant Safety Architecture: Chinese L3 systems are increasingly mandated to include extensive redundancies across critical systems, including perception, decision-making, control, power supply, braking, and steering, to ensure safe operation even if a driver fails to take control.
- Autonomous Driving Data Recording System (DSSAD): Mandatory for autonomous vehicles in China since January 2026, this 'black box' system records critical operational data to reconstruct scenarios after traffic accidents.
- Sensor Fusion: Chinese EV makers like Li Auto, XPeng, and NIO utilize a combination of high-definition cameras, LiDAR, millimeter-wave radar, and ultrasonic sensors for their advanced driver-assistance systems.
- NVIDIA DRIVE Orin: Many Chinese EV manufacturers, including Li Auto and NIO, leverage NVIDIA's DRIVE Orin SoCs. Li Auto's AD Max uses dual Orin chips for 508 TOPS, while NIO's Adam supercomputer employs four Orin SoCs for a total of 1,016 TOPS. NVIDIA's next-gen Drive Thor is targeting 2,000 TOPS.
- Huawei ADS: Huawei's ADS 1.0 system utilized two Ascend 610 chips, providing 400 TOPS of compute power, and incorporated transformer-based BEV (Bird's Eye View) architecture. Later versions (ADS 2.0 and 3.0) focused on reducing reliance on HD maps and improving perception and planning with neural networks.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Consumer trust and adoption of L3/L4 autonomous vehicles in China will accelerate significantly.
BYD's commitment to full liability coverage for accidents caused by its ADAS, coupled with stricter regulatory safety standards and mandatory data recording, is likely to reduce consumer apprehension and foster greater confidence in autonomous driving technology.
Competition among Chinese EV makers will intensify in AI and chip development, rather than solely on price.
The strategic shift away from price wars, exemplified by BYD's in-house 4nm chip and other companies' focus on advanced ADAS features, will compel manufacturers to invest heavily in AI R&D and semiconductor technology to differentiate their products.
China will solidify its position as a global leader in autonomous driving technology and deployment.
Proactive government policies, rapid technological advancements including in-house chip development, and extensive real-world data collection from large EV fleets are creating a robust ecosystem that positions China to lead the next generation of intelligent transportation systems.
โณ Timeline
2001
China launches '863 EV Project' to develop new energy vehicles.
2022
China's national EV purchase subsidies officially end, contributing to the onset of price wars.
2023
EV price war intensifies in China, with BYD initiating significant price cuts across its lineup.
2024-09
China implements recommended national standard 'General Technical Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles'.
2025-07
BYD pledges to cover full losses for users involved in accidents while using its smart parking feature in China.
2025-12
China's MIIT grants first L3 autonomous driving certifications to Changan Auto and BAIC Motor's Arcfox for public road operations.
2026-01
China's mandatory national standard on Autonomous Driving Data Recording Systems (DSSAD) takes effect.
2026-02
China's MIIT drafts new mandatory safety standards for L3 autonomous driving, pushing capabilities closer to L4.
2026-05
BYD unveils Xuanji A3, China's first 4nm self-driving chip, and announces full financial liability for urban NOA accidents.
๐ Sources (19)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: SCMP Technology โ