The industry debate in recent years has centered on which companies can develop Navigate on Autopilot (NOA), achieve end-to-end functionality, and create urban pilot systems that mimic human driving most closely. In the coming years, a more critical question is likely to emerge: which company can make its system indispensable to users.
On April 11, 2026, Horizon Robotics' founder and CEO Yu Kai offered a prediction at a high-level forum on intelligent electric vehicle development: by the 2026 Spring Festival, vehicles equipped with Horizon's Super Driving (HSD) system already accounted for over 40% of total intelligent driving mileage. Once this share surpasses the critical 50% threshold, user reliance on intelligent driving will become irreversible.
He views this as an inflection point analogous to the transition from feature phones to smartphones. When users delegate the majority of their driving time to the system, intelligent driving ceases to be merely a selling point and becomes a standard feature. The company that masters this capability will have the opportunity to become the core of the next generation of the automotive industry chain.
**Overtaking the Competition**
Tesla once defined the first half of the global intelligent driving era. The significance of its Full Self-Driving (FSD) system extends beyond enabling automatic lane changes and parking; it convinced consumers that cars, like phones, could evolve continuously through software updates. This belief earned Tesla a valuation premium far exceeding that of traditional automakers.
However, industry insiders believe Tesla's lead was built on two premises: having the world's largest-scale data feedback loop and facing a prolonged absence of genuine competition. Both of these advantages are now diminishing.
In the Chinese market, Tesla retains strong brand appeal, but the competitive landscape has fundamentally changed. "The market here features intense price competition, complex road environments, rapid model iteration cycles, and a user base willing to try new technologies. The penetration rate of intelligent driving in China is already significantly faster than in most global markets," an industry professional stated.
The true beneficiaries of this trend may not necessarily be the automakers themselves, but rather the technology enablers, or the "shovel sellers." Yu Kai revealed that by August 2025, Horizon Robotics had cumulatively shipped 10 million advanced driver-assistance systems, with 4 million units delivered in the previous year alone. This indicates that Horizon has successfully crossed a common hurdle for Chinese tech firms: completing the closed loop from technological R&D to product development and large-scale commercial validation.
Financial results disclosed in March 2026 showed that Horizon's 2025 revenue reached 3.76 billion yuan, a year-on-year increase of 57.7%. While operating expenses also rose, Yu Kai openly acknowledged the significant spending, stating the company is continuously increasing its R&D investment. Annual R&D expenses amounted to 5.153 billion yuan, focused on personnel, cloud computing power, and technical service procurement. This translates to burning 1.37 yuan on R&D for every 1 yuan of revenue earned. At this pace, the timeline for Horizon to validate its business model is tight.
Horizon's most crucial asset is its HSD system. Yu Kai described HSD as the core product for winning the urban intelligent driving market and the technological foundation for progressing towards higher levels of autonomy. The system entered mass production in November 2025, initially deployed in the mainstream vehicle segment priced around 150,000 yuan. Within just eight weeks of launch, shipments exceeded 25,000 units. By the end of last year, HSD had been designated for over 20 vehicle models, with projected 2026 shipments reaching 400,000 units. This demonstrates that high-level intelligent driving is shedding its exclusive association with premium vehicles and beginning to penetrate the mass consumer market.
An executive from a new automaker noted that the industry previously assumed high-level assisted driving was a feature for vehicles priced above 300,000 yuan, an added luxury. Its emergence in the 150,000-yuan segment indicates that intelligent driving is gradually meeting the conditions for mass adoption. Yu Kai proposed a goal to increase the urban Miles Per Intervention (MPI) metric tenfold this year, reaching several hundred kilometers. "If we achieve this target, we will surpass Tesla," he claimed.
A key distinction is that Tesla operates a single-brand, closed-loop system, whereas Horizon is a platform player. Its system can be integrated into vehicles from multiple automakers across various price points. Each sale of a partner's vehicle helps Horizon collect data, train models, and improve algorithms. The more intense the competition among automakers, the stronger the platform becomes. This dynamic explains why Horizon's software and service revenue has already surpassed its hardware revenue. In 2025, within its automotive solutions business, product solutions revenue was 1.62 billion yuan, while licensing and service revenue reached 1.93 billion yuan, accounting for over half of the total. The gross margin for software licensing is around 90%, far exceeding that of the hardware business.
In essence, Horizon is transitioning from selling chips to selling systems, and further to selling ongoing services. Once intelligent driving mileage exceeds 50% and user dependency is established, the foundation is laid for subsequent revenue streams like software subscriptions, value-added features, and continuous over-the-air (OTA) update fees.
**Strategic Bet**
After securing a foothold in the intelligent driving sector, Yu Kai is preparing for a strategic pivot. He predicts that integrated cockpit and driving domains will be the next major trend, and Horizon is going all-in.
Informed sources revealed that Horizon plans to play two trump cards: the Agentic CAR SoC (an integrated cockpit-driving full-vehicle AI chip) and the Agentic CAR OS (an AI agent operating system). The codename for the former is "Starry Sky." It is a cockpit-driving fusion chip capable of running large language models on the vehicle and is scheduled for official unveiling on April 22, 2026, with mass production and vehicle integration planned within the same year. At the forum, Yu Kai previewed that "Starry Sky" will support an intelligent agent OS similar to models like "OpenClaw." The internal plan is to reveal this card at the Beijing Auto Show.
This signifies that Horizon is not content with being merely an intelligent driving chip supplier but aims to become the central brain of the entire vehicle. The electronic architecture of cars is currently undergoing the most significant shift in the past decade. A R&D professional from an automaker explained that traditional intelligent cars have separate systems for the cockpit and driving domains: two chips, two controllers, two memory sets, and two software teams. The advantage is independence; the disadvantages are higher costs, poor coordination, and slower development. The core logic of cockpit-driving integration is merging these two systems into a single central computing platform.
Yu Kai calculated the benefits: controller fusion can save 50% of chip space, while sharing components and computing power can reduce component costs by 30%, translating to savings of 1,500 to 4,000 yuan per vehicle. Given current profit margins, these savings are highly attractive to automakers. Latest data from the China Passenger Car Association shows that while China's automotive industry revenue reached approximately 111.8 trillion yuan in 2025, the annual sales profit margin fell to 4.1%, with the single-month margin in December dropping to just 1.8%. Price wars have compressed automaker profits to extremely low levels, making every thousand yuan saved critically important, especially in the mainstream 100,000 to 200,000 yuan market where cost reduction directly impacts product competitiveness.
Therefore, while cockpit-driving integration appears to be a debate over technical routes, it is fundamentally a battle for profitability first. Data from Gaogong Intelligent Automotive Research shows that in 2025, 1.5636 million passenger vehicles in China were equipped with pre-installed integrated cockpit-driving computing units, a year-on-year increase of nearly 50%. In just the first nine months, nearly a million new vehicles with this integration were delivered. This indicates the market has moved beyond the "proof-of-concept" phase into the scale adoption phase. By 2030, institutions project the penetration rate of cockpit-driving integration will exceed 30%, with single-chip solutions becoming a mainstream configuration in the 100,000 to 200,000 yuan segment.
The window of opportunity for integrated cockpit-driving SoCs is opening. However, the challenge is formidable. Qualcomm currently maintains a firm grip on the domestic cockpit chip market with a share exceeding 70%. Realistically, a direct confrontation by Chinese manufacturers against Qualcomm is difficult. Yu Kai's strategy is to avoid the mature market and instead target the next-generation market. He stated that past attempts by some Chinese companies to directly challenge Qualcomm's cockpit dominance yielded unsatisfactory results. Horizon plans to leverage underlying technological innovation and system-level optimization to bridge the architectures of intelligent driving and cockpit chips, connect the upper-layer software ecosystem, and essentially create its own blue ocean market.
In other words, Horizon intends to bypass Qualcomm's strongest front and compete in the future where Qualcomm is not yet fully entrenched. This is Horizon's unique advantage compared to pure-play cockpit chip vendors. Yu Kai's reasoning is that entering intelligent driving from the cockpit side is difficult due to the safety redundancy and real-time decision-making requirements. Conversely, extending from intelligent driving into the cockpit presents a lower barrier. However, Qualcomm's solutions are also moving towards cockpit-driving fusion, and its cockpit-side ecosystem is far more mature than Horizon's. Whether Horizon can build a sufficiently robust software ecosystem on the cockpit side remains an open question.
The industry chain is already reorganizing around this trend. Xpeng merged its autonomous driving center and intelligent cockpit center into a General Intelligence Center this year. Li Auto integrated its autonomous driving team into the intelligent space system, restructuring into three main teams: Base Model, Software Body, and Hardware Body. Organizational structure often serves as a leading indicator of trends. When automakers internally no longer distinguish between cockpit and driving domains, it signifies their acceptance that the future car has only one system: the AI system. Horizon's bet on "Starry Sky" is a wager on precisely this future.
Tesla defined the first half of the smart car era, and Qualcomm dominated the cockpit chip period. In the next phase, Chinese companies are vying to define the new order, which will also determine how long Horizon's ambitious narrative can be sustained.
Comments