Skip to main content

元戎启行 (DeepRoute.ai)

Category: AI in Automotive / Autonomous Driving

Chinese L4 autonomous driving technology company developing full-stack self-driving solutions including perception, decision-making, and control systems, with a focus on VLA (Vision-Language-Action) foundation models. 元戎启行 (DeepRoute.ai) was founded in 2019. The company is led by Zhou Guang (周光). Based in Shenzhen, China. Team size: 101-500. Total funding raised: $500.0M. Latest round: Undisclosed. Key investors include Alibaba Group, Great Wall Motor, Fosun RZ Capital, Yunqi Partners, Vision Plus Capital, GSR Capital, Ventech China, Pine Venture Capital, Dongfeng Asset Management.

Founded
2019
Headquarters
Shenzhen, China
Team size
101-500
Total funding
$500.0M

Value proposition

Full-stack L4 autonomous driving solutions that are high-performance, cost-effective (sub-$10K L4 solution), and map-free, powered by end-to-end AI models and VLA (Vision-Language-Action) foundation models for mass production.

Products and solutions

DeepRoute IO (end-to-end intelligent driving platform), DeepRoute IO 2.0 (VLA model-based platform, launched Aug 2025), DeepRoute-Driver 3.0 (high-precision map-free solution, launched Mar 2023), DeepRoute-Driver 2.0 (L4 solution under $10K, Dec 2021), D-PRO (urban NOA solution), D-AIR (cost-effective ADAS solution), Robotaxi service "元启行", Autonomous light truck "元启运"

Unique value

First Chinese autonomous driving company to deploy a 40B-parameter VLA foundation model for ADAS, achieving 10x R&D efficiency improvement; one of the first in China to access NVIDIA's Thor chip; map-free end-to-end solution at mass-production cost points.

Target customer

Automotive OEMs (Great Wall Motor, Geely, Dongfeng, Smart); ride-hailing platforms (Cao Cao Mobility); logistics companies (Deppon Express)

Industries served

Automotive (passenger vehicles ADAS/NOA), Robotaxi, Autonomous logistics/light truck delivery

Technology advantage

40B parameter VLA (Vision-Language-Action) foundation model with chain-of-thought reasoning; map-free end-to-end architecture; self-developed AI inference engine (6x faster than mainstream frameworks); NVIDIA DRIVE AGX Thor chip integration; data flywheel from 300K+ production vehicles; 10x R&D efficiency via foundation model-based data characterization, quality assessment, and virtual evaluation.

How they differentiate

Pioneered map-free (no HD map) autonomous driving in China; transitioned from modular to end-to-end to VLA foundation model architecture; 40B parameter VLA base model with chain-of-thought reasoning for interpretability; deep cooperation with NVIDIA (first batch for Thor chip); backed by major OEMs (Great Wall Motor, Geely) as strategic investors; achieved rapid mass production scale (300K+ vehicles by Apr 2026).

Main competitors

Huawei (Yinwang/ADS), Momenta, WeRide, Pony.ai, Mobileye

Key partnerships

NVIDIA (deep cooperation, first batch for Thor chip), Great Wall Motor (strategic investor & OEM partner), Geely (OEM partner, Smart brand), Dongfeng Motor (OEM partner), Cao Cao Mobility (Robotaxi), Deppon Express (autonomous trucking), Qualcomm (Snapdragon Ride Platform), Microsoft (Azure cloud platform), Volvo Cars (via Lynk & Co 900 Thor platform), RoboSense (LiDAR partnership)

Notable customers

Great Wall Motor (Wey brand - Lanshan, Gaoshan, Tank 400/500/700), Geely (Smart #5, Galaxy M9), Dongfeng Motor, Lincoln, Ford

Major milestones

2019: Founded in Shenzhen, 2019: Partnered with Dongfeng for live-streamed autonomous driving demo, 2020: Partnered with Cao Cao Mobility for Robotaxi testing, 2021: $300M Series B led by Alibaba, 2021: Unveiled sub-$10K L4 solution, 2023: Launched map-free DeepRoute-Driver 3.0, 2024: $100M Series C1 from Great Wall Motor, 2024: First batch in China to access NVIDIA Thor chip, 2025: Launched VLA model and DeepRoute IO 2.0 platform, 2025: Hired DeepSeek core researcher Ruan Chong as Chief Scientist, 2026: 300K+ vehicles equipped, 2026: Secretly filed for Hong Kong IPO

Growth metrics

300,000+ vehicles equipped with ADAS system as of April 2026; targeting 1.3M by end of 2026; 6 funding rounds completed; cumulative funding over $500M

Market positioning

Top 3 ADAS supplier in China alongside Huawei and Momenta; widely viewed as having secured the last major position in China's top-tier assisted driving race; rapidly scaling from 300K to 1.3M vehicles targeted by end of 2026.

Geographic focus

China (primary); expanding globally with overseas OEM partnerships

Patents and IP

Self-developed AI inference engine (6x faster than mainstream frameworks); proprietary VLA foundation model architecture; map-free autonomous driving technology; end-to-end neural network driving system

About Zhou Guang (周光)

Ex-Roadstar.ai Co-founder & Chief Scientist; Ex-Baidu USA autonomous driving team (sensor calibration, time synchronization, perception deep learning algorithms); PhD in AI from University of Texas at Dallas; BS from Tsinghua University Fundamental Science Class

Official website: