Horizon Robotics Swaps Profit for Market Share as 2025 R&D Spend Deepens $1.45 Billion Deficit
Horizon Robotics reported a net loss of RMB 10.46 billion (US$1.45 billion) for 2025, reversing a previous-year profit as the Chinese autonomous driving chipmaker sacrificed short-term margins to bankroll an aggressive expansion in high-end driver assistance systems.
The financial pivot, disclosed on March 19, underscores a brutal land grab within China’s automotive supply chain. While total revenue surged 57.7% to RMB 3.75 billion, research and development expenditures outpaced top-line growth, jumping 63.3% to RMB 5.15 billion as the company raced to deploy its end-to-end AI models and secure next-generation vehicle platforms.
Investors and supply chain analysts are weighing the steep deficit against structural gains in Horizon’s revenue mix. The company’s hardware-centric product and solution sales skyrocketed 144.2%, signaling a successful transition from a pure software licensing model to a dominant hardware-software supplier in China's fiercely competitive electric vehicle sector.
Shifting Revenue Mix Validates Hardware Strategy
The 2025 earnings report reveals a fundamental restructuring of Horizon's cash generation. Product and solution revenue reached RMB 1.62 billion, increasing its share of total revenue from 28% two years ago to 43% last year. This indicates that automakers are increasingly adopting Horizon's integrated hardware rather than merely licensing its algorithms.
Licensing and service revenue, historically the company's cash cow, grew a modest 17.4% to RMB 1.93 billion. However, this segment secured a critical geopolitical and commercial hedge: alongside its joint venture with Volkswagen AG, Horizon onboarded Japan's largest auto parts conglomerate as a top-five client in 2025, establishing a beachhead in the conservative Japanese automotive supply chain.
Despite the intense price war squeezing Chinese automakers, Horizon maintained a robust gross margin of 64.5%, with its core automotive business yielding 67.2%. The RMB 10.46 billion net loss was largely driven by non-cash items, including fair value changes of preferred shares, alongside the RMB 5.15 billion R&D outlay. A liquidity buffer of RMB 20.2 billion in cash and equivalents—up 31.3% year-over-year—provides the runway needed to sustain this cash burn.
Capturing the Mid-Market ADAS Boom
Horizon’s aggressive spending correlates directly with its dominance in China's mass-market advanced driver assistance systems (ADAS). In 2025, the penetration rate of Navigate on Autopilot (NOA) in Chinese passenger vehicles doubled to 42.6%. Horizon captured 44% of the sub-RMB 200,000 vehicle segment, ranking first domestically.
Total shipments of the Journey (征程) series chips reached 4.01 million units, a 38.8% increase. More critically, shipments of mid-to-high-end chips capable of supporting NOA surged nearly 500% to 1.8 million units. These high-tier processors now account for over 80% of the company's product revenue. The November 2025 mass production of Horizon’s HSD—an end-to-end autonomous driving model—resulted in over 22,000 deliveries within a single month, targeting the highly contested RMB 150,000 vehicle tier.
Projecting ASP Growth and Robotaxi Ambitions
Looking ahead to the remainder of 2026, Horizon CEO Yu Kai projects automotive revenue growth to accelerate to 60%, driven by a 35% increase in total chip shipments. The strategic focus is shifting from volume to value: the average selling price (ASP) of Horizon's chips currently sits below US$60. Management anticipates significant ASP expansion as autonomous driving (AD) chips increase their volume share from 45% to over 55%.
The company is also laying the groundwork for post-ADAS revenue streams. Horizon plans to launch an Agentic CAR System-on-Chip (SoC) and operating system later this year, aiming to consolidate cabin and driving computational domains. Furthermore, the company will initiate Robotaxi pilot operations in select Chinese cities by the second half of 2026, leveraging the foundational AI models developed during its 2025 R&D sprint.