Driver assistance Tesla Fights for Its FSD in China with Four Competing Technologies

Related Vendors

FAULHABER Drive Systems

SIMTOS (Seoul International Machine Tool Show)

Tesla aims to bring its new version of the driver assistance system FSD to the streets in China later this year. Approval is now targeted for the third quarter of 2026, as Vaibhav Taneja, CFO of the U.S. automaker, told the Reuters news agency. However, this timeline is still not certain.

At the beginning of the year, CEO Elon Musk said he expected the green light from Beijing as early as February, reported the Chinese automotive portal CNEVPOST at the end of April. That did not happen. Now, Tesla is hoping for the third quarter of 2026.

At the same time, driver assistance technology among Tesla's competitors in China is evolving in different directions. Four distinct technological approaches can currently be observed in the Chinese market, as recently described in a detailed analysis by the Chinese automotive newspaper Zhongguo Qichebao.

Without Lidar and Radar

The first approach is Tesla's own method, a so-called "pure vision end-to-end model." The system forgoes lidar and radar, relying solely on cameras and passing the data directly to a neural network without traditional intermediate steps, which makes the driving decisions. The advantages of this approach include comparatively low costs, rapid data iteration across Tesla's global fleet, and driving behavior that imitates human actions quite well. The weaknesses are also well-known. In poor visibility conditions such as fog or heavy rain, camera systems quickly reach their limits. Additionally, the decisions of the network are hardly comprehensible for national regulatory authorities.

Multi-Sensor Fusion

The second approach, popular in China, is multi-sensor fusion combined with a "world model," a type of AI model that creates a digital representation of the external environment from sensor data. The most prominent Chinese representative of this approach is the technology corporation Huawei. Lidar, radar, and cameras simultaneously deliver signals, which the system then combines into a coherent image. Advantages include high safety even in poor visibility, strong predictive capabilities, and compatibility with the requirements of Chinese regulatory authorities, which value multiple independent sensor sources as additional redundancy. The downside of this solution is higher hardware costs and increased computational requirements.

Vision-Language-Action Architecture

The third approach is the Vision-Language-Action architecture, or VLA. A prominent proponent of this approach is the Chinese tech company XPeng. In VLA, language and perception are integrated into the decision-making process. Simply put, the system can verbally describe what it sees and make decisions based on this.

Advantages of this approach include relatively short response times and efficient traffic flow in complex situations. The central weakness is the typical black-box problem of modern AI systems. Those who need to certify the vehicle or reconstruct events in the event of damage face challenges.

Bird’s-Eye View

The fourth approach is the modular BEV method. BEV in this case stands for "Bird's Eye View," a bird's-eye perspective on traffic reconstructed from sensor data. Among experts, this approach is now considered "classic." Perception, planning, and control are clearly separated here. Each module can be individually tested, evaluated, and replaced if necessary. This ensures stability and traceability.

In complex urban traffic situations, such as during Shanghai rush hour with delivery scooters, pedestrians, and unpredictably parked delivery vans, the modular approach occasionally reaches its performance limits.

Parallel Existence of All Approaches

It is likely that several of these four approaches will coexist in the long term, according to Chinese market observers. However, end-to-end systems are expected to gain importance as the level of automation increases. At the same time, world models and VLA models are technologically converging more and more. Current regulations in the People's Republic clearly favor multi-sensor approaches. Whether a single manufacturer will dominate the mass market for autonomous driving will likely depend primarily on their ability to sufficiently reduce hardware costs.

Slow Recovery of Business

While the competition for the best architecture continues, Tesla's business in China and Europe is picking up again. Sales of the American-made electric cars produced in Shanghai rose by 36 percent in April 2026 compared to the previous year, according to data from the Chinese Passenger Car Association.

Deliveries from Tesla's factory in Shanghai, including exports to Europe and other markets, totaled 79,478 units. While this represents a 7.2 percent decline compared to the previous month of March, it is significantly above the level of the previous year. Tesla is thus stabilizing in its two most important markets outside the USA, after previously having to accept significant market share losses.

In Sweden, France, and Denmark, Tesla's sales continued to rise in April, according to various reports. Analysts write that the demand for battery electric vehicles has been noticeably boosted by higher oil prices resulting from the US-Iran conflict and the crisis in the Strait of Hormuz. Tesla is benefiting from this. The recovery follows a difficult period during which Tesla lost nearly half of its European market share by 2025.

Regulatory Issues

The approval of the next FSD version is facing challenges not only in China. In Europe, Tesla's system is also met with skepticism from regulatory authorities. Reuters cites unpublished emails from European regulatory insiders in the Netherlands, Sweden, Finland, Denmark, and Norway. According to these internal letters, authorities criticize that the Tesla system tends to exceed speed limits. Moreover, it is reportedly not reliable enough under winter road conditions, and safety features can be bypassed, according to Reuters.

Tesla is therefore still simultaneously grappling with regulatory issues in its two most important foreign markets, while Chinese competitors prepare their next steps with their own architectures, vast domestic data resources, and an overall supportive oversight in Beijing.

Tesla needs the newly requested update to its driver assistance system not only for market reasons. Ultimately, the sheer volume of data that can be collected on the road is the crucial competitive factor. The larger the vehicle fleet and the more extensive the data pool, the more powerful the system can become. Tesla currently still has advantages here. However, Chinese providers are catching up on their large domestic market month by month. 

Better Again

Tesla Reports More Sales and More Profit in the First Quarter