Source钛媒体•Stalecollected in 15m
Tesla FSD Enters China, Challenging Local Competitors
See how Tesla's FSD entry is disrupting the competitive landscape for autonomous driving in China.
30-Second TL;DR
What Changed
Tesla FSD is officially entering the Chinese market
Why It Matters
The entry of Tesla FSD will likely accelerate the adoption of end-to-end autonomous driving technology in China, forcing local players to innovate faster.
What To Do Next
Benchmark your autonomous driving stack against Tesla's FSD V12 end-to-end approach to identify potential performance gaps.
Who should care:Developers & AI Engineers
Key Points
- •Tesla FSD is officially entering the Chinese market
- •Local autonomous driving firms face a shrinking window of opportunity
- •Competition is intensifying between global leaders and local startups
Deep Insight
Background and context from public sources — not the original article. 24 sources cited.
Enhanced Key Takeaways
- •Tesla has established a dedicated data center in Shanghai and partnered with Baidu for mapping services to comply with China's stringent data localization and regulatory requirements for autonomous driving technology.
- •While Tesla's FSD Supervised is now officially available in China, full regulatory approval for a nationwide rollout to all eligible vehicles is still pending, with the company targeting Q3 2026 for broader clearance.
- •Unlike other global markets where Tesla transitioned to a subscription-only model for FSD in February 2026, China currently maintains a one-time purchase option for 64,000 yuan (approximately $9,400 USD), with a monthly subscription service yet to be introduced.
- •Chinese domestic competitors like Changan Auto and BAIC Motor have already received Level 3 autonomous driving certifications as of December 2025, a capability that Tesla's FSD (Supervised) (a Level 2 system) does not yet offer.
- •Local players like DeepRoute.ai have achieved significant deployment scale, with over 300,000 cars using its assisted-driving system on Chinese roads by April 2026 and a target of 1 million by year-end 2026, while Momenta has secured partnerships with major OEMs including SAIC, BYD, Mercedes-Benz, and BMW.
Competitor Analysis
Autonomy Level
- Tesla FSD (Supervised)
- Level 2 ADAS (requires constant driver attention)
- Momenta
- ADAS, Robotaxi (L3 testing with GM in Shanghai)
- DeepRoute.ai
- ADAS, Robotaxi, Robotrucks (L4 solution available)
Perception System
- Tesla FSD (Supervised)
- Vision-only (cameras)
- Momenta
- Deep learning software, partnerships with OEMs for various sensor suites
- DeepRoute.ai
- Full-stack, supports both LiDAR-based and vision-only configurations
Key Technologies
- Tesla FSD (Supervised)
- End-to-end AI, Occupancy Networks, World Model, Video-based neural networks
- Momenta
- Deep learning software, intelligent driving technologies
- DeepRoute.ai
- Vision-Language-Action (VLA) model, chain-of-thought reasoning, large language model integration
Pricing (China)
- Tesla FSD (Supervised)
- 64,000 yuan (one-time purchase); no monthly subscription yet
- Momenta
- B2B model (software licensing to OEMs); no direct consumer pricing found
- DeepRoute.ai
- Targets vehicles priced 150,000-300,000 yuan for assisted-driving system
Deployment Scale (China)
- Tesla FSD (Supervised)
- Limited, supervised deployment; full rollout pending Q3 2026 approval
- Momenta
- Partnerships with SAIC, BYD, Mercedes-Benz, Toyota, GM, BMW for ADAS and robotaxi operations
- DeepRoute.ai
- Over 300,000 cars with assisted-driving system by April 2026; targeting 1 million by end of 2026
Data Handling
- Tesla FSD (Supervised)
- Local data center in Shanghai, local AI training center for closed-loop data processing within China
- Momenta
- Not explicitly detailed, but operates in China with local partnerships
- DeepRoute.ai
- Not explicitly detailed, but operates in China with local partnerships
| Feature/Category | Tesla FSD (Supervised) | Momenta | DeepRoute.ai |
|---|---|---|---|
| Autonomy Level | Level 2 ADAS (requires constant driver attention) | ADAS, Robotaxi (L3 testing with GM in Shanghai) | ADAS, Robotaxi, Robotrucks (L4 solution available) |
| Perception System | Vision-only (cameras) | Deep learning software, partnerships with OEMs for various sensor suites | Full-stack, supports both LiDAR-based and vision-only configurations |
| Key Technologies | End-to-end AI, Occupancy Networks, World Model, Video-based neural networks | Deep learning software, intelligent driving technologies | Vision-Language-Action (VLA) model, chain-of-thought reasoning, large language model integration |
| Pricing (China) | 64,000 yuan (one-time purchase); no monthly subscription yet | B2B model (software licensing to OEMs); no direct consumer pricing found | Targets vehicles priced 150,000-300,000 yuan for assisted-driving system |
| Deployment Scale (China) | Limited, supervised deployment; full rollout pending Q3 2026 approval | Partnerships with SAIC, BYD, Mercedes-Benz, Toyota, GM, BMW for ADAS and robotaxi operations | Over 300,000 cars with assisted-driving system by April 2026; targeting 1 million by end of 2026 |
| Data Handling | Local data center in Shanghai, local AI training center for closed-loop data processing within China | Not explicitly detailed, but operates in China with local partnerships | Not explicitly detailed, but operates in China with local partnerships |
Technical Deep Dive
- Tesla FSD:
- Vision-Only Philosophy: Relies exclusively on a suite of cameras for perception, eschewing LiDAR and radar, based on the premise that humans drive using vision.
- Neural Network Stack: Utilizes a pipeline of interconnected neural networks, transitioning towards video-based neural networks to analyze temporal continuity and understand motion across frames.
- Occupancy Networks: A key breakthrough that predicts which 3D space is occupied versus free, even with partial occlusions, improving depth estimation and prediction of hidden obstacles.
- World Model: Converts perception data into a vector space representation, where elements like lanes become splines and vehicles are dynamic objects with predicted trajectories.
- End-to-End AI: FSD V12.0, released in August 2023, marked a significant shift by replacing approximately 300,000 lines of C++ code with neural network decision-making, directly processing raw input data to output control commands.
- Training Infrastructure: Historically used NVIDIA GPUs, but now leverages its custom-built Dojo supercomputer, optimized for video training workloads and neural network scaling.
- Hardware Requirements: Requires Hardware 3 (HW3) or newer, with newer vehicles shipping with Hardware 4 (HW4). Older HW3 vehicles may require a hardware retrofit for full FSD capability.
- DeepRoute.ai:
- VLA Model (Vision-Language-Action): A core component of its DeepRoute IO 2.0 platform, integrating visual perception, semantic understanding, and action decision-making into a unified framework. It employs chain-of-thought reasoning and large language model integration for enhanced interpretability and generalization across diverse driving cultures.
- DeepRoute IO 2.0 Platform: Features a flexible "multi-modal, multi-chip, multi-vehicle" architecture, supporting both LiDAR-based and vision-only sensor configurations, adaptable to various vehicle types.
- DeepRoute-Driver 2.0: An L4-level autonomous driving solution that includes five solid-state LiDAR sensors, eight cameras, a proprietary computing system, and optional millimeter-wave radar, designed for a production cost under US$10,000.
Future ImplicationsAI analysis grounded in cited sources
Tesla's FSD entry will intensify price wars and accelerate ADAS feature adoption in China.
With Tesla's established brand and technology now directly competing, local players will be pressured to innovate faster and potentially lower prices to maintain market share, driving broader adoption of advanced ADAS features across vehicle segments.
China's regulatory framework for foreign AI systems in vehicles will become a global benchmark.
The precedent set by Tesla's compliance with strict data localization and operational requirements for FSD deployment marks a significant opening for foreign AI, potentially influencing how other countries regulate advanced autonomous driving technologies from international companies.
The distinction between Level 2 (supervised FSD) and Level 3 (conditional autonomy) will become a key competitive battleground.
As Chinese domestic brands already offer Level 3 certifications, Tesla's Level 2 FSD Supervised will need to demonstrate superior performance and safety to justify its premium, pushing both global and local players to accelerate their L3 and L4 development.
Timeline
2019
Tesla's Shanghai Gigafactory opens, enabling localized production of Model 3/Y with Autopilot hardware.
2020-2023
Tesla establishes a data center in Shanghai and partners with Baidu for mapping services to comply with Chinese data laws.
2023-08
Elon Musk live-streams the release of Tesla's FSD V12.0, its first end-to-end AI autonomous driving system.
2024-04
China Association of Automobile Manufacturers (CAAM) confirms Tesla satisfied four mandatory data security requirements.
2025-02
Tesla pushes 'City Autopilot' to some Chinese owners, but suspends rollout due to approval issues.
2026-02
Tesla's Shanghai Lingang AI training center officially begins operations, ensuring local data processing and model training.
2026-05-21
Tesla officially confirms FSD (Supervised) support for the Chinese market.
- 2019Tesla's Shanghai Gigafactory opens, enabling localized production of Model 3/Y with Autopilot hardware.
- 2020-2023Tesla establishes a data center in Shanghai and partners with Baidu for mapping services to comply with Chinese data laws.
- 2023-08Elon Musk live-streams the release of Tesla's FSD V12.0, its first end-to-end AI autonomous driving system.
- 2024-04China Association of Automobile Manufacturers (CAAM) confirms Tesla satisfied four mandatory data security requirements.
- 2025-02Tesla pushes 'City Autopilot' to some Chinese owners, but suspends rollout due to approval issues.
- 2026-02Tesla's Shanghai Lingang AI training center officially begins operations, ensuring local data processing and model training.
- 2026-05-21Tesla officially confirms FSD (Supervised) support for the Chinese market.
Sources (24)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
1notanfsdtracker.comvertexaisearch.cloud.google.com2basenor.comvertexaisearch.cloud.google.com3teslaacessories.comvertexaisearch.cloud.google.com4globalchinaev.comvertexaisearch.cloud.google.com5tmgm.comvertexaisearch.cloud.google.com6carnewschina.comvertexaisearch.cloud.google.com7cnevpost.comvertexaisearch.cloud.google.com8thenextweb.comvertexaisearch.cloud.google.com9wikipedia.orgvertexaisearch.cloud.google.com10counterpointresearch.comvertexaisearch.cloud.google.com11benzinga.comvertexaisearch.cloud.google.com12prnewswire.comvertexaisearch.cloud.google.com13momenta.aivertexaisearch.cloud.google.com14bmwgroup.comvertexaisearch.cloud.google.com15youtube.comvertexaisearch.cloud.google.com16wikipedia.orgvertexaisearch.cloud.google.com17tracxn.comvertexaisearch.cloud.google.com18medium.comvertexaisearch.cloud.google.com19yeslak.comvertexaisearch.cloud.google.com20wikipedia.orgvertexaisearch.cloud.google.com21gasgoo.comvertexaisearch.cloud.google.com22eeworld.com.cnvertexaisearch.cloud.google.com23teslaacessories.comvertexaisearch.cloud.google.com24gasgoo.comvertexaisearch.cloud.google.com
Weekly AI Recap
Read this week's curated digest of top AI events →
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 钛媒体 ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.