Representative End User Clients
Representative Automation Clients
Representative Software Clients
On July 8, 2026, Momenta was listed on the Hong Kong Stock Exchange, becoming the first “Physical AI stock” in Hong Kong. Its market capitalization exceeded HK$700 billion on its first trading day. The core narrative behind Momenta’s IPO points to the significant growth potential of Physical AI.
The underlying foundation of Physical AI is the world model. It not only processes information but also understands physical principles such as space, motion, causality, and interaction. The year 2026 has been described as the “first year of Physical AI,” and autonomous driving is the only market segment to have successfully closed both the data and commercial loops, with Momenta emerging as a leading player.
The R7 World Model: A Unified Foundation
In April 2026, Momenta took the lead in mass-producing and launching the R7 World Model, becoming the first company in the industry to bring a world model into mass production.
The R7 World Model uses a three-tier architecture:
World Model Pretraining: Compresses physical common sense and causal relationships derived from massive volumes of driving data into the base model.
World Model Simulation: Conducts closed-loop testing for extremely rare, long-tail scenarios.
Reinforcement Learning: Enables the model to iterate and evolve through exploration, trial, and error.
Its technical foundation is based on more than 120 billion kilometers of real-world and simulated driving mileage accumulated by Momenta. From this data, the company has extracted more than 100 million segments of high-value “golden data,” all sourced from its mass-production fleet.
As of July 2026, Momenta’s solutions had been deployed in more than 1 million vehicles. The company had delivered more than 100 mass-produced vehicle models and secured designations for more than 210 additional models.
The R7 World Model could have a significant impact on the robotics sector. Traditional robots generally follow a specialized-model approach, under which the development of almost every new robot category must begin from scratch.
Momenta provides a unified Physical AI foundation model capable of supporting a variety of mobile entities. Whether they are cars, vans, or household robots, these systems operate in the same physical world and must understand the same principles of gravity, friction, spatial relationships, and causal interactions.
The robotics industry is therefore progressing from the fragmented era of “one model for one robot” toward a generalized era of “one model for all scenarios.”
“One Flywheel, Two Legs”
Since its inception, Momenta has followed its “One Flywheel, Two Legs” strategy.
One Flywheel refers to the closed loop of data-driven algorithm development. As more mass-produced vehicles are deployed, the volume and diversity of available data increase. This enables algorithms to improve more quickly, attracting additional automakers and creating a virtuous cycle.
Two Legs refers to the company’s L2 mass-produced assisted driving and L4 fully autonomous driving businesses. The mass-production business generates data and revenue, while the fully autonomous driving business feeds advanced technologies back into mass-production solutions. Both businesses share the same underlying technical architecture.
This strategy has produced notable results. From 2023 to 2025, revenue increased from 743 million yuan to 2.413 billion yuan, representing a compound annual growth rate of more than 80 percent. Gross margin increased from 17.5 percent to 71.6 percent, while technology licensing revenue grew 42-fold over three years. Cash reserves exceeded 10 billion yuan.
Momenta holds a 64.5 percent share of the global market for independent urban navigation-on-autopilot solution providers. Its customer base includes all major Chinese passenger vehicle manufacturers, and the company works with nine of the world’s top 10 automakers. Its solutions have been deployed in more than 10 countries and regions across Asia and Europe.
This data flywheel also has important implications for robot development. The robotics industry has long faced challenges related to the difficulty and cost of acquiring high-quality data.
Momenta’s understanding of the physical world, developed through autonomous driving applications, offers cross-scenario generalization capabilities. These capabilities could be transferred efficiently to robotics, strengthening environmental perception and dynamic decision-making while reducing the time required to move from research and development to commercial deployment.
Full-Spectrum Expansion
Based on the R7 World Model, Momenta is developing a portfolio that supports multiple types of mobile entities and operating scenarios.
Commercial Deployment at Scale
Momenta’s current commercial deployments include:
Passenger Vehicles: Mass-produced assisted driving solutions deployed in more than 1 million vehicles.
Robotaxis: Partnerships with Mercedes-Benz, Uber, Grab, and other companies, with plans to advance overseas commercialization in Abu Dhabi, Munich, and other cities during 2026.
Robovans: Unmanned logistics vehicles that have already entered commercial deployment.
Planned Expansion
Momenta plans to expand into robotruck applications in 2027 and eventually enter the household service robot market.
At the company’s listing ceremony, CEO Xudong Cao said: "Over the past decade, we have enabled AI to master driving and brought a dedicated driver to every household. In the next decade, we will bring dedicated robot service scenarios such as housekeepers, doctors, and teachers to every household, creating the ‘GPT moment’ for Physical AI."
Momenta positions itself as a platform-level company, using a single large model to support multiple scenarios, reduce marginal costs, and build a collaborative business ecosystem.
This approach also presents a potential commercialization model for the robotics industry: developing and refining technologies through mass-produced assisted driving before applying mature capabilities to emerging markets.
From Autonomous Driving to Embodied Intelligence
The first phase of Physical AI is autonomous driving, while the next phase is embodied intelligence. Rather than developing robotic capabilities from the ground up, Momenta is transferring technologies proven in autonomous driving, including 3D spatial understanding, precise positioning, and multisensor fusion.
These are the same fundamental challenges that autonomous driving systems must address every day.
With the R7 World Model, Momenta has established a strong position in the emerging Physical AI market. Its portfolio extends from passenger vehicles and robotaxis to robovans, robotrucks, and household service robots.
Through a unified Physical AI foundation model, the company aims to support mobile scenarios ranging from transporting goods and people to delivering services within homes and other environments.
As CEO Xudong Cao has stated, autonomous driving is the prologue to Physical AI. Momenta’s full-spectrum Physical AI strategy is extending that prologue while creating a potential pathway for the wider robotics industry toward more generalized artificial intelligence.
Related ARC Advisory Group Content
ARC Advisory Group has published several related articles examining the technologies, platforms, and market developments shaping Physical AI, world models, autonomous systems, and embodied intelligence:
NVIDIA Announces Blueprint for Open Physical AI Data Factory
Forging a New Language—ARC’s Unified Taxonomy for the AI Robotics Era
Gemini Robotics-ER 1.6 Advances Embodied Reasoning for Industrial Robotics
Navigating the Ecosystem of Industrial Humanoids: Early Movers and Key Partnerships
Together, these articles provide additional context on how Physical AI is evolving from autonomous mobility toward more generalized robotic systems capable of understanding and interacting with the physical world.