Author | Zhou Zhiyu
Editor | Zhang Xiaoling
China's intelligent driving industry is undergoing a fierce reshuffle.
Recently, the former unicorn Haomo Zhixing announced a halt in operations, a news confirmed by Haomo Zhixing employees. Almost at the same time, Yuanrong Qixing announced in Guangzhou that it had captured 40% of the third-party assisted driving market, while Zhuoyu received the endorsement of the &#34;national team&#34; with a 3.6 billion yuan investment from FAW Group.
The stark contrast between cold and hot reflects the dramatic upheaval in the industry's power landscape. Car manufacturers are no longer blindly casting a wide net; BYD, Chery, and Geely are integrating internal intelligent driving resources or deeply binding only with a few leading players. The once flourishing supply chain is rapidly shrinking.
In the second half dominated by end-to-end large models, intelligent driving is no longer an embellishment but an expensive ticket to AGI. The market no longer believes in PPTs and grandiose statements; only players who master the entry to mass production data in the millions are qualified to board the Noah's Ark heading towards the future. As for those unable to cross the scale survival line, their names can only be ruthlessly swallowed in this torrent.
<h2>The Stalled Unicorn</h2>
Haomo Zhixing pressing the pause button may be the most cautionary tale in the intelligent driving industry for 2025.
Haomo Zhixing originated from the forward-looking department of Great Wall Motor's intelligent driving system, which became independent in 2019, with Zhang Kai, the deputy chief engineer of Great Wall Motor, as chairman, and Gu Weihao, the former chief architect of Baidu's intelligent automotive division, as CEO. This is a typical case of combining traditional automotive industry with internet thinking, and it was quite favored by capital in its early days.
In 2021, it secured two rounds of financing worth hundreds of millions within a year, with Meituan and Hillhouse Capital vying to get involved. At that time, Haomo had a dream start that all startups envied: not only continuous orders from Great Wall but also the joining of technical executives from Baidu.
However, the most recent financing for Haomo Zhixing can be traced back to the first half of 2024, which was a B round financing of over 100 million yuan. But in this financing, there was no sign of Great Wall Motor.
In the early days when Haomo Zhixing attracted capital attention, Great Wall Motor was its important backing, and its HPilot 1.0 intelligent driving system was first installed in the Wey brand Mocha model.
However, Haomo Zhixing soon fell out of favor with Lady Luck. One technical detail is that in the early stage of hardware selection, in pursuit of extreme cost-effectiveness, Haomo bet on the Qualcomm Snapdragon Ride chip instead of the industry mainstream NVIDIA Orin. In that era of &#34;high-precision maps + rule algorithms,&#34; this choice seemed wise, cheap, and sufficient. But as the industry rapidly shifted towards Transformer architecture, this foreshadowing turned into a cage that trapped Haomo Zhixing.
According to industry insiders, due to the high difficulty of adapting the chip platform to the new architecture, Haomo's engineers had to get bogged down in hand-writing operators and hardening adaptations. While competitors easily achieved high frame rates on the NVIDIA platform, Haomo struggled with basic operational efficiency
What is even more regrettable is the strategic arrogance and sluggishness. In July 2023, the trend in the intelligent driving market rapidly shifted from &#34;light maps&#34; to &#34;end-to-end large models.&#34; At this time, the technical executives of Haomo expressed a reserved attitude towards the implementation of large models in public, believing that the timing was still early.
Just a month after these remarks were made, Tesla launched the end-to-end FSD V12, completely rewriting the industry rules; ten months later, XPeng began mass production of vehicles equipped with end-to-end large models. This strategic misjudgment directly led to Haomo's cliff-like lag in technological generations.
Meanwhile, Haomo Zhixing was still promoting the implementation of the urban NOA no-map solution. In an internal letter last November, Zhang Kai and Gu Weihua set the focus for the next year's work on creating the most cost-effective urban no-map NOA solution.
Great Wall Motor's patience was ultimately exhausted by &#34;PPT delivery.&#34; Senior executives at Great Wall openly stated in internal meetings that they were only responsible for the market competitiveness of the complete vehicles, and would use whoever's technology was effective. Relevant executives admitted that although there would be pressure when making choices and it was difficult to decide, this route was beneficial for the healthy development of the market.
While Haomo Zhixing was slow to deliver, Great Wall Motor also brought in Yuanrong Qixing as a supplier of intelligent driving solutions, and just as Haomo Zhixing celebrated its 4th anniversary in November 2024, Great Wall Motor turned around and made a $100 million exclusive investment in Yuanrong Qixing.
Moreover, Haomo Zhixing's efforts to expand to other clients have not been satisfactory. It announced in 2023 that it had signed cooperation agreements with automakers such as Beijing Hyundai, but as of now, the number of models equipped with their solutions remains very limited.
This makes it clear that Haomo Zhixing's stagnation is not a coincidence.
<h2>Competing for Entry Tickets</h2>
In the intelligent driving battlefield of 2025, the dimensions of competition have been forcibly elevated. In the past, the competition was about whose demo was cooler; now, it is about who can cross the &#34;one million vehicles&#34; physical threshold.
&#34;One in every three smart cars is equipped with Horizon,&#34; said Lu Peng, Vice President of Horizon, at a recent communication meeting, showcasing their strength. This unicorn, which has just been listed on the Hong Kong stock market, is attempting to build an alliance similar to &#34;Wintel&#34; in the PC era.
Why is there such a focus on one million vehicles? Yuanrong Qixing CEO Zhou Guang revealed the secret: &#34;Assisted driving in pure highway scenarios can hardly form an effective data closed loop, and there are issues with the business model.&#34; In other words, without the data generated from a million mass-produced vehicles running day and night on the roads, any sophisticated code will run dry.
In the era of end-to-end large models, data is no longer just a record but the fuel for algorithm evolution. Only when the fleet size reaches the million level and daily activity is sufficiently high can enough Corner Cases (long-tail scenarios) be covered, allowing the model to produce &#34;intelligent emergence.&#34; Players that do not reach this scale, no matter how advanced their technology, will still collect data that is singular, ordinary, and without value
This also explains why Yuanrong Qixing was able to emerge victorious from the anxious battle among suppliers, securing orders for core models such as Great Wall's Wey Blue Mountain and Tank 500.
Zhou Guang shared a detail: to win over a major client, the team once waited outside the client's hotel at four in the morning, just to secure a test drive opportunity. This kind of &#34;wolf-like&#34; behavior is driven by an extreme thirst for scale. The results rewarded this thirst—monthly sales of the Wey Gaoshan MPV equipped with the Yuanrong solution skyrocketed from 300 units at the beginning of the year to nearly 10,000 units in October.
&#34;Clients are willing to entrust us with their core blockbuster models.&#34; Zhou Guang's statement reveals the new role of intelligent driving suppliers: they are no longer behind-the-scenes component providers but strategic partners who determine the life and death of models.
The same logic applies to the marriage between FAW Group and Zhuoyu (formerly known as DJI Automotive). FAW Group's willingness to invest 3.6 billion yuan in Zhuoyu is far from a simple financial investment. For the &#34;first son of the republic,&#34; it needs a &#34;super Tier 1&#34; that can connect a wide ecosystem and has massive data throughput capabilities.
Zhuoyu's commitment to maintaining independent operations is precisely the core value of this transaction.
If Zhuoyu becomes a &#34;mole&#34; of FAW, it loses the possibility of accessing data from other brands, thus losing its value as a platform. What FAW values is Zhuoyu's cost-effectiveness and openness, a capability that is a lifeline for all automakers in the current fierce price war.
Currently, the power structure of the intelligent driving industry is being reshaped. Players with &#34;million-level&#34; data access are gaining absolute pricing power over the industry chain. The &#34;chip + algorithm&#34; ecosystem built by Horizon, the blockbuster models secured by Yuanrong Qixing, and the &#34;national team&#34; endorsement obtained by Zhuoyu are essentially all about seizing the high ground of data.
In this scale war, there is no middle ground. You either grow strong and become a platform, or remain small and beautiful, then slowly wither away. For those still reveling in a few thousand units of shipment, the countdown to exit the table has already begun.
<h2>The Final Game</h2>
The industry seems to have reached a consensus: intelligent driving is no longer an embellishment but a ticket to the next era. The destination of this ticket is AGI (Artificial General Intelligence).
In the narratives of Zhou Guang and Lv Peng, what they are doing has long surpassed merely helping automakers build good cars. They are training a universal &#34;world model&#34; through the largest intelligent terminal, the automobile.
Horizon repeatedly emphasizes the &#34;intuitive system.&#34; Lv Peng believes that a true experienced driver relies 95% on intuition rather than calculation. The &#34;one-stage end-to-end&#34; architecture built by Horizon aims to give cars this human-like intuition, no longer rigidly executing the code of &#34;stop at red lights and go at green lights,&#34; but instead smoothly perceiving the environment and interacting like a living organism.
Yuanrong Qixing has introduced the VLA (Vision-Language-Action) model. Zhou Guang described a scenario: when a vehicle encounters complex tidal lanes or special road conditions directed by traffic police, traditional rule-based algorithms may fail due to an inability to understand road signs or gestures. However, vehicles equipped with the VLA model possess COT (Chain of Thought) capabilities—they can reason like humans: &#34;Although it is a red light now, I see the traffic police waving, and the road sign indicates that passage is allowed during this period,
&#34;So I want to follow the car in front.&#34;
This kind of logical reasoning ability, a display of &#34;high intelligence,&#34; is precisely the core capability required by general robots.
Today's smart cars are essentially robots with four wheels. Every end-to-end large model trained by autonomous driving companies and every strategy accumulated for complex road conditions will ultimately be reused in humanoid robots, logistics robots, and other broader physical worlds.
The RoadAGI plan of Yuanrong Qixing is a testament to this vision. They aim to enable the autonomous driving system not only to drive on the road but also to command robots to deliver takeout to the doorstep. This ambition for the &#34;last 100 meters&#34; reveals the ultimate goal of autonomous driving companies: to become the brain of the physical world.
In this grand narrative, Robotaxi is no longer an unattainable research project.
Yuanrong Qixing and Horizon have firmly chosen the &#34;Tesla path&#34;—that is, to achieve L4 level autonomous driving through large model upgrades using data and hardware from mass-produced passenger cars. When Yuanrong Qixing announced the direct use of mass-produced models like Great Wall's Blue Mountain (costing only 200,000 to 300,000 yuan) for Robotaxi, it meant that the unit economic model (UE) for Robotaxi had finally been validated.
This marks the official transformation of Robotaxi from a research project to a consumer-level service. In this new paradigm, the boundary between assisted driving and unmanned driving has been completely broken. The millions of private cars on the road are not only users' travel tools but also stealthy data collectors, continuously providing Corner Cases for the L4 model. This &#34;laying eggs along the way&#34; model completely shatters the barriers of heavy asset operations like Waymo.
Future mobility services will no longer be dominated by taxi companies with thousands of modified vehicles but will be defined by autonomous driving platforms that possess the data access of millions of mass-produced vehicles.
This also explains why the capital market was still willing to invest in leading autonomous driving companies during the winter. They are not betting on the automotive parts business but on a ticket to the era of &#34;embodied intelligence.&#34;
For the Chinese autonomous driving industry, 2025 will be a winter of bubble bursting and a warm spring of value return.
Momenta CEO Cao Xudong once predicted: &#34;Next year, the landscape of assisted driving in Chinese cities will be set, and in the end, there may only be two or three companies left.&#34; This is not alarmism. The current market share is rapidly concentrating towards leading players like Huawei, Momenta, and Horizon. Those in the middle tier, with severe technological homogenization and lacking the ability to generate profits, are being ruthlessly eliminated.
For car manufacturers, autonomous driving has transformed from an &#34;optional feature&#34; into a &#34;core asset.&#34; In the final circle, no one is willing to completely hand over their soul to others. This means that the window of opportunity for third-party suppliers is closing. Unless they can provide capabilities that car manufacturers cannot surpass in the short term, like Huawei and Horizon, or offer extreme cost-effectiveness like Zhuoyu.
The future landscape is gradually taking shape: one type is self-research car manufacturers that possess millions of sales, relying on their closed-loop data iteration algorithms; the other type is super Tier 1 suppliers serving other car manufacturers, gathering data through alliances to compete for tickets to the final circle
When one million vehicles become the line of life and death, for those remaining at the table, such as Horizon and Yuanrong Qihang, the real game has just begun.
Risk Warning and Disclaimer
The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk

TSLA

601633

02333

002594

BYDDF

BYDDY

01211

XPEV

09868

NVDA

The Chinese intelligent driving industry is undergoing intense reshuffling, with the former unicorn Haomo Zhixing halting operations, while Yuanrong Qixing and Zhuoyu are making progress in the market. Car companies are no longer blindly expanding, but are integrating resources or collaborating with a few leading players. Intelligent driving has become the key to AGI, and the market is placing more emphasis on actual data. The halt of Haomo Zhixing serves as a warning for the industry, as it now faces challenges after once being favored by capital

- The Chinese intelligent driving industry is undergoing significant upheaval, highlighted by the suspension of the once-prominent unicorn, Haomo Zhixing, while competitors like Yuanrong Qixing capture substantial market shares.  
- Major automakers are consolidating their resources and forming strategic partnerships, as the industry shifts from a focus on flashy demonstrations to the necessity of large-scale data collection for effective autonomous driving solutions.  
- As the market consolidates, only those with access to extensive data will thrive, marking a critical transition towards advanced artificial intelligence in the automotive sector.

China's Intelligent Driving Launches a Brutal Breakthrough Battle