zhangtongshe.com
Entering the Market Late, but at the Perfect Time.
Recently, Shanghai Zaofu Intelligent Technology (hereinafter referred to as "Zaofu Technology") completed a new round of financing of approximately USD 100 million. This round was led by the Shanghai Guotou Xiandao Fund, with follow-on investments from Minjintou (Shanghai Minhang Jinrong Touzi) and China Creation Ventures. Following the financing, the company's valuation is approaching USD 3 billion.
Zaofu Technology takes over Hellobike's Robotaxi business, focusing on the R&D, application, and commercialization of Level 4 autonomous driving technologies. Previously, strategic shareholders such as Ant Group, Alibaba Group, and CATL (Contemporary Amperex Technology Co., Limited) have successively injected capital. The funds from this round will be primarily used for R&D investment, with a focus on deploying a 10,000-card cluster computing power platform, iterating end-to-end technology for AI (Artificial Intelligence) large models, and building a high-quality data supply chain system.
In the Robotaxi track, Zaofu Technology did not start early, but backed by Hellobike, Ant Group, and CATL, it possesses a latecomer advantage that is difficult for most startups to replicate.
A New Player "Born with a Golden Spoon"
In June 2025, Hellobike, Ant Group, and CATL joined forces with an initial investment of over CNY 3 billion to establish Zaofu Technology, which focuses on Level 4 autonomous driving, making a high-profile entry into the Robotaxi track.
Backed by the "Golden Triangle" formed by Hellobike, Ant Group, and CATL, Zaofu Technology was born with inherent advantages in scenarios, users, and technology, almost as if born with a golden spoon in its mouth.
This is not the first collaboration among the three. As early as 2019, Hellobike, Ant Group, and CATL co-established a joint venture to launch the XiaoHa battery swapping business. This cooperative experience laid the foundation for the three parties to join forces again in vehicle manufacturing in the future.
Founded in 2016, Hellobike started with shared bikes. Relying on a differentiated strategy targeting third- and fourth-tier cities, it survived the reshuffling of the shared bike war and has been steadily expanding its business.
As a leading enterprise in the shared mobility sector, Hellobike's advantage in entering the Robotaxi market lies in its massive user base and operational experience. The company has over 800 million registered users, with core business covering more than 400 cities nationwide, daily orders exceeding 10 million, and an operation and maintenance network across 360 cities. These resources can provide scenario support for Robotaxi.
Ant Group, on the other hand, provides support in AI and data capabilities. Through ecosystems such as Alipay and Amap, Ant Group has acquired massive data on users' travel habits, which can be used to optimize path planning and demand prediction for autonomous driving. In addition, its blockchain technology and payment system can also provide underlying support for Robotaxi settlement and data security.
The addition of CATL fills the gap in the hardware platform. At the end of 2024, CATL officially released the Pan Shi Chassis, which adopts a three-dimensional bionic tortoise shell structure and can achieve a safety level of no fire or explosion in a 120km/h frontal collision while energized. It is reported that the Pan Shi Chassis can provide a standardized hardware platform and adaptation interfaces for intelligent driving systems, greatly simplifying the entire process from Robotaxi design and R&D to mass production and manufacturing.
Hellobike grew out of real travel scenarios and naturally understands user needs and operational efficiency; Ant Group provides the data and payment ecosystem; CATL provides the hardware platform. Each of the three parties has its own strengths and complements each other, which constitutes the underlying logic for Zaofu Technology's entry into the Robotaxi track.
Building the First Vehicle in Three Months
Because it is "standing on the shoulders of giants," Hellobike's Robotaxi business has advanced at a very fast pace. In just three months, the first pre-installed mass-produced Robotaxi model, "HR1" (Hello Robot1), was officially unveiled.
At the same time, Hellobike announced strategic cooperation with core industry chain partners such as Dongfeng Venucia, Alibaba Cloud, Horizon Robotics, and Hesai Technology, working together to promote the full-chain synergy of Robotaxi from technology R&D and vehicle manufacturing to operational implementation.
HR1 is jointly built by Hellobike and Dongfeng Venucia, with the core Level 4 autonomous driving software and hardware systems mainly provided by Horizon Robotics. In July this year, this vehicle was successfully mass-produced and rolled off the production line, exceeding a scale of 100 units, indicating that the technical solution now has a reliable physical carrier.
Unlike the traditional modular technical architecture of Robotaxi, Hellobike's Robotaxi adopts an end-to-end integrated architecture. Perception, prediction, planning, and control no longer operate in isolation; instead, multi-sensor inputs are directly mapped into vehicle control commands, thereby improving both system response speed and decision-making consistency.
Behind this end-to-end Level 4 system are three pillars: "big data, massive computing power, and large models."
At the data level, through an automated and compliant data production line, Hellobike plans to build a high-quality autonomous driving dataset of no less than 10 million clips (training segments) by 2026.
At the computing power level, relying on deep cooperation with Alibaba Cloud, Hellobike will build a 10,000-card computing power cluster to train and iterate autonomous driving models, and will deeply cooperate with Alibaba Cloud's Qwen team to develop the "DaoYu" large model with 10 billion parameters.
In terms of large models, Hellobike did not follow the commonly used autoregressive models in previous autonomous driving, but directly built an end-to-end driving large model based on diffusion models to process multimodal outputs of driving decisions, achieving comprehensive learning of different driving strategies in complex scenarios.
According to the plan, Hellobike's Robotaxi will achieve business scale within two years, reaching a scale of 10,000 vehicles in 2026, and deploying over 50,000 vehicles in 2027.
Currently, the core technical team has been basically established, with an average age of 27, driving product R&D with an AI-native concept. From talent and computing power to data, R&D investment is being transformed into the underlying capability for the continuous iteration of Robotaxi.
Entering the Market Late, but at the Perfect Time
Robotaxi is regarded as the ultimate answer to the commercialization of autonomous driving and has almost always been in the spotlight of capital in recent years.
Elon Musk once predicted that this is a "trillion-dollar opportunity." According to research report forecasts, China's Robotaxi industry will reach a market turning point in 2027, with market sizes of CNY 83.1 billion and CNY 709.6 billion in 2030 and 2035, respectively.
The turning point has not yet arrived, but the conditions for an explosion are gradually in place.
At the technical level, perception large models and end-to-end algorithms are continuously iterating, and the processing capability for long-tail scenarios is advancing. NVIDIA released the open-source Alpamayo 2 Super autonomous driving VLA model in 2026. Intelligent driving systems are evolving from "see-react" to "understand-decide," and more and more intelligent driving enterprises are accelerating their layout in the Level 4 Robotaxi track.
At the policy level, unified national Level 4 testing and operational standards are being established, and autonomous driving road rights are gradually expanding from first-tier cities to second- and third-tier cities. Shanghai has proposed to achieve over 6 million passenger trips via Level 4 autonomous driving by 2027, and multiple regions in Guangdong have implemented detailed rules for Robotaxi operations. Countries and regions such as North America and the Middle East are also introducing special regulations, supported by operational subsidies and priority road rights.
At the user level, the urbanization process is accelerating, and urban commuting pressure is prominent. Robotaxi can make up for the capacity gap of traditional ride-hailing services, and after years of pilot programs, users' acceptance of autonomous driving is steadily improving.
Some players are already running ahead in the track. Pony.ai, WeRide, and Apollo Go have deployed fleets in multiple cities, and hardware costs are also declining. In some pilot scenarios, signals of positive unit economics for single vehicles have begun to emerge. However, losses have not disappeared accordingly, and leading players have not yet fully solved the problem of scaled replication.
This means that the industry has passed the technology validation stage and is moving towards scaled commercialization, but the profitability model is still being explored.
Robotaxi is an asset-heavy business. The ultimate competition is not just about algorithms, but the systemic synergy of data, computing power, vehicle engineering, operational networks, and urban implementation. Hellobike understands users and operations, and can leverage the ecosystems of Alibaba, Ant Group, and CATL for synergy—this is the core competitiveness of Zaofu Technology.
Text | Zhouyu Editor | Yida