Auto Show Preview: L3 Adoption and Domestic Chip Dominance [Analysis]

After reading this Beijing Auto Show preview, what impressed me most were two main themes: Technology Democratization and Ecosystem Reconstruction. The industry has finally moved from "flexing muscles" to "calculating books," and 2026 really seems like a tipping point.

First, regarding L3 scaling. The article mentions L3 legal highway sections opened in 23 cities nationwide, along with a "10-second legal takeover" mechanism, which is far more important than mere technical parameters. Previously, L3 was stuck on regulatory liability definitions; now, with clear automaker primary responsibility + black boxes + specialized insurance, it gives consumers peace of mind. More critically, standard L3 on 200k-300k RMB vehicles and optional L3 on 150k RMB vehicles (cost 10k-20k) directly shatters the old notion that "advanced smart driving = luxury exclusive." Price drops stem from sensor and chip cost dilution and algorithm efficiency improvements, not feature cuts. Real vehicle demos of the AITO M9 and Voyah Taishan Ultra will verify this.

Another major highlight is the battle for pricing power among domestic smart driving chips. Black Sesame's Huashan A2000 single chip exceeds 1000 TOPS, Horizon Journey 6P reaches 560 TOPS—unimaginable two years ago. But more important than raw compute stacking is that LiDAR unit prices dropped from 3000-5000 RMB to the thousand-RMB level, and chip costs compressed by 30%. This directly changes the vehicle BOM structure—previously, advanced smart driving hardware costs were too high to penetrate the mainstream market. Now, domestic chips + deep Tier1 partnerships (Horizon + ZF, Black Sesame + GAC Intelligent Control) allow software-hardware decoupling, giving automakers flexible selection space and breaking NVIDIA's monopoly.

In terms of technology route comparison, Huawei ADS 5.0's luxurious fusion approach (4x 896-line LiDARs + 13 cameras) versus Xpeng XNGP 5.0's lightweight pure vision + dual Orin-X (508 TOPS) approach is essentially a trade-off between cost and scenario coverage. Huawei achieves a 99.5% success rate in unprotected left turns but has high sensor costs; Xpeng's VLM model + lighter sensors is more aggressive, suitable for volume sales. This differentiation benefits consumers—different budgets offer different accessible experiences.

As for autonomous buses and subscription models, I think they are inevitable for closing the commercial loop. Mushroom Car Union's MOGOBUS entering Singapore's public transport backbone proves the replicability of Chinese solutions in developed markets. Subscription models (Tesla FSD monthly rent 680 RMB) convert one-time high barriers into continuous service revenue, key to the industry's profitability inflection point.

In short, this auto show is no longer a carnival of concept cars but a real exam for mass production implementation. I recommend paying attention to actual takeover experiences of L3 models, measured performance of domestic chips, and the pricing battle between Huawei and Xpeng in the 200k RMB market segment.

https://www.36kr.com/p/3768944509186569