⚛️量子位•Stalecollected in 51m
Re-evaluating BYD through the lens of intelligent driving
💡Understand how a global EV giant is scaling AI infrastructure to compete in the autonomous driving race.
⚡ 30-Second TL;DR
What Changed
BYD is accelerating R&D in end-to-end autonomous driving models
Why It Matters
BYD's pivot to intelligent driving signals a major shift in the Chinese EV landscape, forcing competitors to accelerate their own AI roadmaps. This could lead to a rapid commoditization of L2+ driving features.
What To Do Next
Monitor BYD's open-source contributions or technical whitepapers regarding their perception stack to understand their approach to end-to-end learning.
Who should care:Developers & AI Engineers
Key Points
- •BYD is accelerating R&D in end-to-end autonomous driving models
- •Shift from hardware-centric to software-defined vehicle architecture
- •Strategic investment in AI talent and compute infrastructure for training
🧠 Deep Insight
Web-grounded analysis with 45 cited sources.
🔑 Enhanced Key Takeaways
- •BYD has unveiled its self-developed Xuanji A3, China's first 4nm automotive-grade driving chip, designed to support Level 3 and Level 4 autonomous driving capabilities.
- •The company is implementing a 'smart driving equality' strategy, aiming to popularize high-speed Navigation on Autopilot (NOA) features to mainstream vehicle models priced between 100,000 and 150,000 yuan by 2025.
- •BYD has committed to investing over RMB 100 billion (approximately $14.75 billion) in intelligent driving research and development over the next three years.
- •To build user trust, BYD is offering full damage coverage for accidents that occur while its God's Eye Advanced Driver-Assistance Systems (ADAS), including urban NOA, are engaged.
- •BYD's intelligent driving strategy involves a significant shift from traditional rule-based methods to data-driven end-to-end (E2E) learning methods, aiming for higher performance ceilings.
📊 Competitor Analysis▸ Show
Competitor Analysis: Intelligent Driving Capabilities
| Feature/Company | BYD (God's Eye / DiPilot) | XPeng (XPILOT ASSIST) | NIO (NIO World Model / Aquila) | Huawei (ADS) |
|---|---|---|---|---|
| Core Chip | Xuanji A3 (4nm, L3/L4 support, up to 2100+ TOPS with 3 chips) | XBrain, XNet neural network, XPlanner | NX9031 (5nm, for ES9) | Ascend 610 (400 TOPS with 2 chips for ADS 1.0) |
| Sensor Suite | LiDAR (optional/standard on some models), HDR cameras, dual long-wave infrared cameras, radar, ultrasonic sensors | Multi-sensor fusion: cameras, LiDAR, radar | 33 high-performance sensing units: LiDAR, 8 MP cameras, 3 MP surround-view cameras, ADMS, millimeter-wave radars, ultrasonic sensors, GPS, IMU, V2X | ADS 2.0: 128 LiDARs, 11 HD cameras, 3 MMW radars, 12 ultrasonic radars |
| Key Features | High-speed NOA, Urban NOA, intelligent parking, DiLink AI Intelligent Cockpit with digital assistant, full damage coverage for ADAS accidents | Navigation Guided Pilot (NGP) for highways and cities, AI Chauffeur (customizable driving/parking), Valet Parking Assist (VPA), over 10 active/passive safety functions | Active safety (driver disability, rear-end collision prevention, obstacle recognition), highway/urban navigation assistance, intelligent parking (memory, autonomous garage navigation) | Navigation Cruise Assist (NCA) for highways and cities, General Obstacle Detection (GOD), Road Cognition & Reasoning (RCR), automatic/valet parking |
| Strategic Approach | 'Smart driving equality' (democratizing ADAS), vertical integration (chip development), end-to-end learning, cloud-based computation exploration | AI-powered in-car OS, end-to-end large models (XNet, XPlanner, XBrain), personalized experiences | Full-stack in-house development (perception, localization, control, platform software), AI-first strategy with SkyOS | Full-stack smart car solutions (hardware & software), partnerships with OEMs, focus on real-world data training |
| Partnerships | NVIDIA (DRIVE Orin, DRIVE Hyperion), DeepSeek, Hesai (LiDAR) | N/A | N/A | Audi, Toyota, BAIC Group, Chery, JAC Group, SAIC Motor, Dongfeng Motor, Seres Group |
🛠️ Technical Deep Dive
- Xuanji A3 Driving Chip: China's first self-developed 4nm automotive-grade System-on-Chip (SoC). It features a 16-core CPU with 420k DMIPS compute power, a Neural Processing Unit (NPU) exceeding 700 TOPS, and 273 GB/s memory bandwidth. It is designed to meet ASIL-D safety standards and supports Level 3 and Level 4 autonomous driving. A three-chip configuration can deliver over 2,100 TOPS of computing power.
- God's Eye Platform Upgrades: The latest version introduces XUANJI Architecture 2.0, an industry-first satellite sensor architecture, an upgraded physical AI large model, and a self-evolving data flywheel based on real-world driving scenarios.
- Sensor Configuration: The God's Eye LiDAR Version can be optionally equipped across BYD's entire vehicle lineup, featuring over 1,000-line LiDAR, HDR cameras, and dual long-wave infrared cameras.
- DiLink AI Intelligent Cockpit: The latest generation includes a new AI-powered digital assistant capable of proactive task execution and advanced reasoning, designed for a personalized and continuously evolving in-car experience.
- Compute Infrastructure: BYD utilizes NVIDIA DRIVE Orin (254 TOPS) and NVIDIA DRIVE Hyperion computing architecture in its NEVs for automated driving and parking. The company also plans to use NVIDIA's AI infrastructure for cloud-based AI development and training.
- Software Architecture: BYD is transitioning from traditional rule-based methods to data-driven end-to-end (E2E) learning methods for high-level intelligent driving. The e-Platform 3.0 features an integrated, domain-controlled electronic and electrical architecture compatible with BYD OS, enabling continuous over-the-air (OTA) updates.
- Cloud Computing: BYD is actively testing the shift of computation for real-time applications to the cloud to keep pace with rapidly developing data volumes and software applications.
- e4 Platform: This platform features four independent motors and wheel-level torque control, allowing for decentralized power delivery and reframing vehicle dynamics as a software problem.
🔮 Future ImplicationsAI analysis grounded in cited sources
BYD will significantly disrupt the premium intelligent driving market by offering advanced features at lower price points.
Its 'smart driving equality' strategy and cost-effective self-developed chips like Xuanji A3 aim to democratize high-level ADAS, making it accessible in mainstream models.
BYD's vertical integration in chip development and software will enhance its competitive advantage and control over the intelligent driving stack.
Developing its own 4nm chip (Xuanji A3) reduces reliance on external suppliers and allows for deeper optimization and cost control, similar to its successful battery strategy.
The company's commitment to full damage coverage for ADAS-related accidents will accelerate consumer adoption and trust in its autonomous driving systems.
This policy directly addresses a major consumer concern regarding liability in assisted driving, potentially lowering the psychological barrier to using such features.
⏳ Timeline
2018
Launch of DiLink 1.0, BYD's intelligent cockpit and connected-car ecosystem.
2022-01
BYD partnered with Nuro to manufacture all-electric autonomous delivery vehicles.
2022-03
BYD announced collaboration with NVIDIA to use DRIVE Hyperion computing architecture and DRIVE Orin SoC in NEVs starting H1 2023.
2023-05
BYD established a new intelligent driving division in Shanghai and embarked on a major hiring spree, recruiting 4,000-5,000 software engineers.
2024-01
BYD unveiled its Integrated Vehicle Intelligence strategy and XUANJI Architecture at Dream Day 2024.
2025-02
BYD partnered with DeepSeek for smart driver assistance tech and announced the rollout of its 'DiPilot' system across its lineup.
2026-05
BYD unveiled the Xuanji A3, China's first self-developed 4nm automotive-grade Driving SoC, supporting L3 and L4 autonomous driving.
📎 Sources (45)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- electriccarsreport.com
- bydukmedia.com
- thedriven.io
- movemnt.net
- marketscreener.com
- japantimes.co.jp
- carscoops.com
- reddit.com
- eeworld.com.cn
- sundayworld.co.za
- thelec.net
- xpengcars.co.uk
- cleantechnica.com
- medium.com
- hesaitech.com
- youtube.com
- auto-007.com
- nio.com
- nio.ae
- wikipedia.org
- electronicsforyou.biz
- xpeng.com
- xpeng.com
- youtube.com
- carnewschina.com
- huawei.com
- table.media
- 1strespondernews.com
- iot-analytics.com
- just-auto.com
- byd.com
- nvidia.com
- bydglobal.com
- businessinsider.com
- nvidia.com
- byd.com
- youtube.com
- bydcarchina.com
- byd.com
- investing.com
- indiatimes.com
- clickpetroleoegas.com.br
- thestandard.com.hk
- news18a.com
- byd.com
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 量子位 ↗