Following a series of negotiations, Tesla's Supervised FSD (Full Self-Driving) appears to be on the verge of entering the Chinese market. In the landscape of urban assisted driving in Chinese cities, will FSD replicate its dominance as seen in the United States? Can it carve out a unique niche? Tesla's strengths lie in its end-to-end strategy, global fleet data, and adept handling of complex scenarios.

The pressure is palpable: Domestic mainstream assisted driving companies have spent years honing their skills in urban NOA (Navigate on Autopilot), highway NOA, parking, memory driving, LiDAR perception, and localized scenarios.

If Supervised FSD merely reaches a "finally usable" state in China, it will represent a catch-up move for Tesla owners rather than a revolutionary change.

Should FSD deliver a consistently reliable experience at complex Chinese intersections, amid mixed traffic with non-motorized vehicles, construction detours, aggressive lane changes, and unprotected left turns, we anticipate that domestic assisted driving companies will invest heavily in benchmarking and testing.

Two contrasting viewpoints emerge: one hails the arrival of the "catfish" (a disruptive force), while the other cautions, "Success in North America doesn't guarantee success in China."

FSD's entry into China transcends a mere software update. It encompasses the capabilities of Supervised FSD itself, local data and regulatory constraints in the Chinese market, and the gap between domestic assisted driving companies (empowered by LiDAR technology) and Tesla's FSD. Only through competition will the true potential of FSD in China be revealed.

01

The Advent of Supervised FSD

FSD now comes in supervised and unsupervised versions. The supervised version caters to drivers in vehicles, while the unsupervised version is designed for L4 Robotaxi applications. According to UN Regulation R-171, Supervised FSD falls under the L2 driving assistance category.

Thus, the competition still pits domestic companies (Li Auto, Xpeng, NIO) and assisted driving suppliers (Huawei, Momenta, Horizon Robotics, Yuan Rong, Zhuoyu) against each other.

Previously, FSD operated with limited capabilities. This time, it seems poised to unleash its full potential. Can it cover a broader range of road scenarios, reduce driving burdens, and provide stable, predictable actions in complex situations?

Chinese Tesla users have already paid 64,000 RMB for the FSD software package, yet the experience has long been incomplete. This time, whether it's urban NOA or highway NOA, the experience is finally comprehensive.

FSD truly garners attention by training its neural networks with vast amounts of real-world road data, enabling the system to learn judgment and actions from diverse driving scenarios.

In early May 2026, Tesla's official website announced that the global cumulative mileage of the Supervised FSD fleet had surpassed 10 billion miles (approximately 16.1 billion kilometers). From 1 billion miles in June 2024 to 10 billion miles in May 2026, it achieved a tenfold increase in less than two years.

Tesla's underlying advantage with FSD lies not in single-vehicle intelligence but in fleet size and data closed-loop. More vehicles mean more scenarios; more scenarios enhance the model's ability to cover long-tail problems.

However, this also poses the first challenge for FSD's entry into China: Chinese roads are not mere replicas of North American roads. Chinese cities boast more electric vehicles, tricycles, food delivery riders, temporarily parked vehicles, narrow road encounters, complex ramps, non-standard construction, and high-density aggressive lane changes.

These scenarios cannot be naturally transferred by simply "performing well in North America." Currently, Tesla faces regulatory obstacles regarding data export in China. Road data collected in China must be stored domestically and not transmitted abroad.

For FSD to achieve more complete localization in China, it must navigate the relationships between data storage, processing, model training, and regulatory approval. In the initial phase of Supervised FSD's entry into China, the focus is on refining the basic experience on Chinese roads and swiftly achieving "full potential" through Chinese data.

02

How Has Domestic Assisted Driving Evolved?

FSD's entry into China is dubbed the "catfish effect" because Tesla has long been a benchmark in intelligent driving discussions. However, the Chinese market is no longer the same as it was a few years ago, awaiting FSD's guidance.

In recent years, domestic mainstream assisted driving companies and automakers have rapidly adopted end-to-end, VLA (Vision-Language-Action), and world model approaches, advancing competition to the urban NOA stage.

Unlike Tesla, many domestic solutions have opted for LiDAR for perception hardware.

LiDAR is not a cure-all; it cannot replace algorithms or automatically solve all long-tail problems.

Nevertheless, LiDAR does elevate the baseline for many assisted driving experiences. In 2026, stronger and more LiDAR combinations can directly provide the three-dimensional shape and distance of obstacles ahead, aiding in recognizing abnormal obstacles, stationary objects, cones, construction zones, and low obstacles.

After upgrading to LiDAR, mainstream domestic assisted driving solutions have become more stable within conservative boundaries in complex urban environments.

For instance, when cones, road occupancy construction, temporarily parked vehicles, or crossing electric vehicles appear, the system can more easily establish spatial awareness. In scenarios like unprotected left turns, narrow road encounters, and intersection negotiations, LiDAR provides a layer of more certain distance and obstacle information.

Of course, if the LiDAR algorithm is subpar, LiDAR becomes mere decoration, enhancing perception redundancy but still relying on systematic capabilities for final performance. This is the core route difference between FSD and domestic solutions. Domestic solutions often adopt multi-sensor fusion, adding LiDAR alongside cameras to reassure consumers at the delivery level.

The benefits of FSD's route will become apparent after its entry into China. Can it achieve equal or superior stability, comfort, and disengagement rates with fewer sensors in front of users already educated by domestic automakers? Can it convince domestic users? The answer remains to be seen.

FSD's current focus is on learning Chinese regulations and road conditions. Chinese urban arterials feature highly specific traffic lights, various tidal and bus lanes, non-motorized vehicles, temporarily parked vehicles, food delivery riders, and dense aggressive lane changes.

Complex intersections truly test urban assisted driving. Chinese urban intersections present numerous challenges: unprotected left turns, U-turns, pedestrians and electric vehicles jostling, unclear lane markings, complex diversion lines, and non-standard traffic light positions.

When to yield, when to proceed, when not to hesitate, and when to be conservative—we await FSD's upper performance limits in Chinese cities.

Of course, after FSD's entry into China, its success will largely hinge on long-tail abnormal scenarios, such as construction detours, temporary cones, reverse-flow electric vehicles, tricycle crossings, nighttime low light, rainy day puddles, street vendor occupancy, and parking lot exit mixes. Encountering these scenarios will determine user trust. Tesla's advantage lies in global fleet data and model iteration, requiring it to undergo a process of Chinese localization examination again.

Summary

Finally, regarding pricing, Supervised FSD currently only supports a subscription model. The one-time purchase option was globally discontinued in February 2026. The subscription price has not been finalized yet, but China will also adopt the subscription model, turning FSD into a competition of experience and pricing.