What lessons do the first batch of fallen intelligent driving unicorns leave for the survivors?

In late November this year, Haomo.AI, the first Chinese company to mass - produce autonomous driving technology, suddenly announced a full - staff suspension of work.

This smart driving unicorn, which was established six years ago and once had a valuation hyped up to $1 billion, ultimately failed to survive this winter. In the eyes of many, Haomo.AI fell in the industry's offensive and defensive battle due to falling behind in technology and losing orders. However, if we only understand Haomo.AI's "shutdown" as the failure of a single company, we actually overlook the real structural problems in the smart driving industry - almost all the premises that supported the rapid progress of China's smart driving industry in the past five years have "become invalid" at the same time.

Haomo.AI is just the first batch of fallen unicorns, but it won't be the last.

Smart Driving: From the "Trial - and - Error Period" to the "Liquidation Period"

Facing Haomo.AI's "shutdown", most people's first reaction from the outside world was "Why did it come so suddenly?" However, within the industry, those who truly understand the business logic of autonomous driving are not that surprised. To understand the inevitability of Haomo.AI's "shutdown", we need to start from the cyclical changes in the industry itself. In the past five years, China's smart driving industry has been operating in a "vacuum" environment. Capital was willing to subsidize the future, car manufacturers were willing to bet on technology, and users were willing to try out new technologies.

These three foundations are like three beams supporting the entire track. Although each alone is not strong enough, they interact and balance each other, enabling the entire industry to maintain rapid development. However, in 2024, these three beams "broke" simultaneously.

The first to fall was the narrative logic of "software subscription monetization". In the past, the industry believed that the profits of smart driving would come from the monthly subscription fees for high - level advanced driver - assistance systems, which was a high - margin model of "earning once from selling hardware and earning for a lifetime from selling software". However, as urban NOA has changed from an optional feature to a standard one, and then to "free for life", users' willingness to pay for smart driving functions has dropped to an all - time "low".

Even Tesla, which is known for its smart driving technology, is facing the same dilemma. Tesla charges a one - time subscription fee of $8000 for its high - level smart driving function "FSD" software in the US market, and the optional price of FSD in the Chinese market is 64,000 yuan. At Tesla's Q3 2025 earnings conference, Chief Financial Officer Vaibhav Taneja admitted that only a very small number of car owners actually use the FSD service.

According to his description, Tesla's FSD - related revenue in the second quarter of this year decreased compared with the same period in 2024. Taneja said: "The total number of customers who currently pay to use FSD is still very small, accounting for only about 12% of our existing fleet."

Domestic brands have pushed the Chinese auto market into the era of "free smart driving". The overall cost of DJI's high - speed NOA package has also dropped to the level of 5000 yuan. Coupled with the fact that Hesai and RoboSense have brought lidar into the "thousand - yuan era", and NVIDIA's Thor has reduced the unit cost of computing power even more than in the Orin era, the industry's price system has been completely rewritten. This has also led urban NOA into the "bargain - basement" era. Currently, the cheapest urban NOA model in the domestic market is the 100,000 - yuan - level SUV Baojun Yunhai.

Meanwhile, BYD has introduced high - level smart driving to its 70,000 - yuan - level Seagull model, which is equipped with functions such as remote vehicle exit, automatic parking, automatic on - and off - ramp driving in urban areas, and active overtaking/lanes changing. This proves that through large - scale production and algorithm optimization, car manufacturers can control the cost of high - level smart driving systems at a very low level.

When standard features become the mainstream, the industry has completely "disenchanted" the prospects of the subscription business model. For any autonomous driving Tier 1 supplier, losing the "subscription system" means that half of the future cash - flow model is directly cut off. Smart driving suppliers have changed from "earning money from software" to "relying on payments from car manufacturers". Against this background, players whose business models do not match their cost models are forced to face market "liquidation".

Secondly, capital confidence is also declining. The primary market has shifted from "betting boldly on the future" to "profit - oriented". When new concepts such as large models and embodied intelligence emerged, autonomous driving is no longer a super - hot topic.

Haomo.AI's downfall is not an isolated case. In recent years, at least seven autonomous driving companies that have achieved business implementation and have a considerable scale, such as Zhongzhixing, Qingyan Vision, Zongmu Technology, and Dazhuo Intelligence, have gone bankrupt, been liquidated, or undergone in - depth restructuring. According to Sina Finance statistics, among the ten smart driving companies listed on the secondary market this year, eight are still in the red.

Finally and most importantly, there has been a change in the attitude of car manufacturers. The increasingly fierce price war in recent years has forced all car manufacturers to re - allocate their resources back to the basic profit model. Under stronger regulatory requirements, more limited cost budgets, and shorter market - entry cycles, car manufacturers have begun to streamline their supply chains, reduce the scope of trial - and - error, and put forward clearer performance - to - cost ratio requirements for smart driving suppliers.

When these three underlying logics change simultaneously, the entire industry naturally enters the "liquidation period". Haomo.AI's shutdown is the most direct manifestation of this industry "correction". It marks the end of an old era that relied on capital infusion and technology stories, and the arrival of a new cycle that focuses on the essence of business and cost - efficiency.

Haomo.AI: Strong in the "Strengths of the Old Era", Trapped in the "Weaknesses of the New Era"

Focusing on Haomo.AI, as an enterprise, it was born in a cycle of "encouraging trial - and - error, undefined scale, and loose capital", but it fell in a cycle with extremely low tolerance for errors and extremely fast iteration. However, its problem is not just falling behind in technology, but that its corporate capabilities, cost structure, and organizational structure have not changed in line with the industry's changes.

In its early days, Haomo.AI's advantages were very obvious. As an autonomous driving enterprise with relatively early large - scale mass - production experience in China, Haomo.AI became the third company in the world, after Tesla and XPeng, to launch a full - stack self - developed navigation assistance system in 2021. This also attracted a large amount of hot capital, and the number of employees in the enterprise once expanded from hundreds to 1500.

According to the industry pace at that time, Haomo.AI was one of the autonomous driving suppliers that could easily "get on track". However, this advantage belongs to the past era. When the industry has entered the "race among hundreds of cities" and Huawei ADS, XPeng XNGP, and Li Auto NOA are being rolled out across the country, Haomo.AI's urban NOA is still in the state of "small - scale city pilot projects".

The essence of the autonomous driving industry is to exchange scale for data, data for algorithms, and algorithms for user experience. However, Haomo.AI dropped the ball during the scale - expansion period. In 2023, Haomo.AI tried to launch an urban NOA solution priced below 3000 yuan, but this seemingly attractive product has not been mass - produced to date. Missing out on mass - production means missing out on data, which ultimately led to Haomo.AI's algorithm iteration always lagging behind its competitors.

A deeper problem lies in the problems with Haomo.AI's own organizational structure and governance structure. Since 2023, the company's core leaders, such as the vice - president of technology, the vice - president of product, and the CIO, have successively left the company. According to a report in Time Weekly, before the "dissolution storm", the number of current employees at Haomo.AI has dropped to only 280. Compared with the scale of over 1500 employees two years ago, the total number of employees has shrunk by more than 80%.

According to a report by Dongchedi, after contacting Haomo.AI employees, it was learned that there are currently only more than 200 employees in the company, "but all the technical backbones have left". Dongchedi also reported that Zhang Kai, the chairman of Haomo.AI, and Gu Weihao, the CEO, are still in their positions.

According to people close to the company, Haomo.AI has a "large - scale organization" internally, almost equivalent to a "miniature car manufacturer". There are more than a dozen first - level departments, such as product engineering, intelligent hardware, artificial intelligence, product marketing, and capital markets, which report directly to the CEO. This has also trapped Haomo.AI in the "heavy - organization dilemma" and made it unable to transform quickly in line with the industry's rhythm.

According to a report from Lei Feng Network, a team within Haomo.AI proposed to layout an end - to - end large model at the beginning of 2023, but the proposal was rejected on the grounds that "the cycle was unpredictable and the resource investment was too large". However, by the end of 2024, end - to - end had become an industry consensus. XPeng, Li Auto, and DeepRoute.ai have successively mass - produced VLAs. It was only then that Haomo.AI re - evaluated its technical direction, but it had already lagged behind the industry by an entire cycle.

Meanwhile, when the technical route was still unclear and the company's resources should have been focused on core business, Haomo.AI still simultaneously promoted the construction of the Changxing Smart Driving Domain Controller Factory and the Chengdu Intelligent Robot Manufacturing Factory in 2024. This scattered investment due to unclear strategy further accelerated Haomo.AI's dissolution.

For a track like smart driving that requires rapid decision - making, rapid iteration, and rapid scaling, Haomo.AI's organizational structure is too fragmented. This has led to its inability to streamline itself internally at the critical moment of the industry's cycle change and its lack of sufficient organizational resilience to cross the threshold of the new cycle.

The supply chain in the smart driving industry is shifting from "multiple scattered choices" to "binding with leading players". To control costs and risks, car manufacturers are more inclined to choose a small number of stronger and larger - scale suppliers for in - depth cooperation, which has shrunk the survival space for players in the middle. Since Haomo.AI is neither a super - leading enterprise nor a low - cost, lightweight supplier, it is naturally difficult for it to find its place in the new cycle.

Cost Narrative: When Smart Driving Enters the Next Trench

Currently, the smart driving industry has entered the deep - water area of competing on cost and commercialization from the first half of competing on technology and financing. Whether an enterprise can achieve a sustainable business model under a reasonable cost structure has become a key question that determines its survival. Against this background, industry players have begun to differentiate into clear survival paths and competition echelons.

At the top are super players with scale, ecosystem, and technology closed - loops, such as Huawei, Tesla, and XPeng, which have large models and data systems. These companies can span cycles and have triple barriers in terms of team scale, data volume, and R & D capabilities. They have the ability to maintain a partial paid model in the future.

The second layer consists of mainstream car manufacturers that actively choose to bind with leading suppliers. For most car manufacturers, smart driving is no longer a technology that requires independent innovation, but a part of the "supply chain". Most brands will choose to bind with one or two leading solutions, or adopt a hybrid combination of self - development and outsourcing to ensure a balance between cost and experience and keep risks under control.

The third layer includes smart driving companies that focus on niche markets, such as Pony.ai's in - depth development in the Robotaxi scenario. They do not confront smart driving large models head - on but seek commercial breakthroughs in specific scenarios.

Those forced to leave the market are players like Haomo.AI in the middle. These enterprises do not have ecosystem - level capabilities, cannot compete on cost, and cannot win more customers with flexibility. They generally have problems such as slow technology iteration, heavy team structure, high cost structure, and lack of self - hematopoietic ability. The first sign of the industry entering the "liquidation period" is the downfall of such companies.

From a longer - term perspective, Haomo.AI's shutdown is actually a sign of the industry moving towards health. Because smart driving cannot rely on capital for a long time, nor can it survive in the narrative of "burning money for scale" for a long time. Smart driving ultimately depends on a real business model, a balance between cost and experience, scale to improve products and algorithms, and self - hematopoietic rather than relying on shareholder investment.

In the smart driving industry in the new cycle, technology is not the only moat, and capital is not a reliable enough backer. What ultimately determines whether a company can survive is the synchronization ability of its organization, rhythm, cost, and business model. Those who can provide a good enough experience at a controllable cost can stay in the game. Those who cannot reduce costs will be the first to be eliminated. In other words, the first batch of fallen unicorns fell in the old narrative, while the surviving players still need to prove themselves in the new business logic.