On June 17, at the "2026 9th Intelligent Driving & Overseas Expansion Conference" hosted by Gasgoo, Huawei intelligent driving expert Huang Ziliang pointed out that intelligent driving is rapidly becoming a standard feature of new energy vehicles, with L2 and L2+ level ADAS penetration rates continuing to climb. He emphasized that against this backdrop, the construction of fundamental computing power capabilities is becoming the core foundation of competition among automakers. The future industry landscape will be determined jointly by computing power reserves and data closed-loop capabilities.
ADAS Penetration: The Critical Transition from Optional to Standard
Data disclosed by Huang Ziliang at the conference shows that the intelligent driving penetration rate in China's NEV market is undergoing an accelerated climb.
L2-Level ADAS Penetration Trend:
- H1 2025: approximately 42%
- Q1 2026: approximately 58%
- May 2026: has surpassed 65%
L2+ Level ADAS Penetration Trend:
- H1 2025: approximately 18%
- Q1 2026: approximately 31%
- May 2026: has reached 39%
This means that over 60% of new NEVs already possess basic intelligent driving assistance capabilities, while nearly 40% of models are equipped with higher-level L2+ functions (such as highway NOA and automatic lane change). Intelligent driving is transitioning from a "premium option" to an "industry standard."
Computing Power: The Competition Variable Moving from Behind the Scenes to Center Stage
Huang Ziliang specifically noted that when intelligent driving becomes standard, the focus of competition undergoes a fundamental shift:
| Competition Phase | Core Capability | Representative Metrics |
|---|---|---|
| Phase One | Feature availability | Whether basic functions like AEB and ACC are present |
| Phase Two | Feature experience | Takeover rate, comfort, scenario coverage |
| Phase Three (Current) | Computing foundation | Chip computing power (TOPS), training data scale |
Huawei's investment in computing power layout is industry benchmark-level:
- Ascend AI chip clusters provide thousands of PetaFLOPS of cloud training computing power
- In-vehicle chips such as Kirin 990A and Kirin 9610A continue iterative upgrades
- The data closed-loop system processes hundreds of millions of kilometers of driving data daily
Other major players' computing power reserves are also rapidly catching up:
- XPeng: Self-developed Turing chip, targeting hundreds of TOPS
- NIO: Shenji chip, single-chip computing power exceeding 500 TOPS
- Li Auto: Partnership with NVIDIA, multi-Orin-X solution
- BYD: Partnership with Horizon Robotics, Journey series chips deployed
Impact on Overseas Markets
The computing power competition has profound implications for Chinese automakers' overseas expansion:
- Central Asia and Russia: Consumer awareness of intelligent driving is still in early stages, but high-computing platforms leave ample room for future OTA upgrades
- EU Market: Regulations have strict certification requirements for L2+ features, and computing power redundancy helps pass compliance reviews
- Southeast Asia Market: High price sensitivity requires finding a balance between computing power configuration and cost control
Feedback from overseas dealers sourcing Chinese NEVs through EX1000.COM shows that intelligent driving features have become one of the top three buyer concerns, after price and range.
The essence of the computing power competition is a three-in-one contest of data, algorithms, and hardware. In this arena, Chinese companies are building unique competitive advantages.
