Horizon Robotics (HK:09660), China's leading autonomous-driving chip and software supplier, reported first-half 2026 revenue of over 2 billion yuan (about $282 million), up 32.9% year-on-year, according to Chinese tech outlet Huxiu. However, adjusted net loss widened 25.4% to 1.67 billion yuan (about $235 million), and the company's shares have fallen more than 10% since the results were published, reflecting investor concerns about its growth ceiling.
The company's market position remains strong. Horizon holds more than 50% of China's market for basic assisted-driving solutions (ADAS) for domestic brands, a 22.8% share in urban NOA (city advanced driver assistance), ranking second, and a 31.9% overall share of China's intelligent driving market, ranking first. Yet the financials reveal a structural challenge: revenue is still far too small to absorb the company's heavy research-and-development spending, which reached 5.15 billion yuan in 2025, equivalent to 137% of that year's revenue, and 2.76 billion yuan in H1 2026, or 134% of revenue.
Two business models, diverging fortunes
Horizon operates two distinct revenue streams. The first is a hardware-oriented model, where it sells chips, algorithms and tools as a bundle delivered with vehicles, charging per vehicle installed. The second is a software-oriented model called "HSD together," where customers license algorithms and data services on Horizon's base platform rather than buying hardware outright. The software model has historically been far more profitable, with gross margins around 90% versus roughly 36% for hardware in H1 2026.
Software revenue peaked at 71.3% of total revenue in 2024, but fell to 54.4% in 2025 as hardware sales surged 144% to 1.62 billion yuan, driven by the success of the Journey 6 (Zhengcheng 6) processor series and the mass production of HSD, Horizon's full-scenario urban NOA solution. In H1 2026, software contributed 75.3% of automotive gross profit, with a gross margin of 90.4%.
Horizon argues that integrated hardware-software solutions unlock the full potential of its technology and are the foundation for its licensing business, drawing an analogy to restaurant franchising: self-operated stores must thrive before franchise fees can be collected. But automakers increasingly prefer a mixed strategy of self-development, joint development and external procurement, aiming to retain control over product definition and iteration. Many are buying Horizon's underlying capabilities, such as its BPU architecture and base models, rather than complete hardware packages, creating a mismatch between Horizon's expectations and customer choices.
Losses persist despite margin improvements
Operating losses have shown little improvement over the past two and a half years. The operating loss rate stood at 81.3% in H1 2026, down from 89.5% in 2024 and 88.8% in 2025, but the gap between gross margin (66% in H1 2026) and total expense ratio (167.3%) has actually widened slightly. The company's operating loss reached 1.67 billion yuan in H1 2026, compared with 3.3 billion yuan for full-year 2025.
One positive development came from Horizon's robotics subsidiary, D-Robotics (also known as Digua Robotics). After multiple financing rounds diluted Horizon's stake to 40%, the loss-making robotics business no longer drags on the parent's results and generated a one-off gain of 2.78 billion yuan in H1 2026. Still, analysts note that Horizon's core challenge remains scale: until revenue grows enough to spread R&D costs thin, profitability will stay elusive, and its leading market share in a crowded Chinese autonomous-driving sector may not guarantee the growth investors seek.
