Horizon Robotics HSD V2.0 advances full-scenario intelligent driving upgrades with a dual-engine architecture
AI summary card
Horizon Robotics HSD V2.0 advances full-scenario intelligent driving upgrades with a dual-engine architecture
Deutsche Bank assigns a Buy rating to Horizon Robotics (9660.HK), believing that Superdrive 2.0 comprehensively improves autonomous driving experience and safety through its world model, reinforcement learning, and OCC active safety capabilities.
- HSD V2.0 shifts from a single-stage end-to-end modular design to a dual-engine framework of "world model + reinforcement learning."
- The world model is trained on real expert driving data and can simulate complex long-tail scenarios while predicting future traffic conditions.
- The reinforcement learning model supports continuous self-iteration of the system, covering driving, parking, safety, and long-tail scenario handling.
- OCC is applied to active safety functions such as AEB, AES, and AMAP, enhancing the ability to recognize and respond to unconventional obstacles.
- The company disclosed that its technology accumulation includes more than 50 top-conference papers, over 100,000 academic citations, and more than 2,000 global patent applications over the past three years.
Report interpretation
Overview
This report focuses on Horizon Robotics' launch of the second-generation Horizon Superdrive (HSD V2.0) full-scenario autonomous driving system. Deutsche Bank believes the core change in the system is the upgrade from a single-stage end-to-end modular design to a dual-engine architecture of "world model + reinforcement learning," aiming to achieve a more complete capability leap in complex roads, parking, active safety, and in-cabin interaction.
Core views
The core view is that HSD V2.0 gains spatiotemporal understanding and complex long-tail scenario simulation capabilities through the world model, and achieves continuous self-iteration through reinforcement learning; meanwhile, OCC active safety extends recognition logic from traditional "whitelist" object recognition to occupancy network perception, helping improve response capability to unknown and irregular obstacles. The report maintains a positive view on the stock, with the current rating at Buy.
Analysis framework
The report combines technical product interpretation with verification of the company's R&D capabilities: it first analyzes the architectural changes in HSD V2.0, then breaks down its functional upgrades in driving experience, parking, hazard avoidance, and user interface, and finally uses papers, citations, patents, and laboratory布局 to demonstrate the company's long-term technology accumulation.
Methodology notes
The world model is responsible for understanding and simulating complex traffic environments, while reinforcement learning is responsible for continuous self-iteration.
This framework is used to explain the architectural leap of HSD V2.0 relative to the previous-generation system, as well as its potential improvements in long-tail scenarios, prediction, defensive driving, and navigation strategy.
An upgrade from traditional whitelist object recognition to occupancy network perception.
The report states that OCC is applied to active safety functions such as AEB, AES, and AMAP, helping identify and respond to unconventional obstacles.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Horizon Robotics(9660.HK)Company covered by the report and investment target
- Strengths
- HSD V2.0 features technical highlights such as a world model, reinforcement learning, OCC active safety, and Journey 6P toolchain optimization; the company also has accumulated papers, citations, patents, and multi-laboratory R&D capabilities.
- Weaknesses
- The source materials do not disclose financial forecasts, target price, current price, order size, or mass-production timeline, making it difficult to judge the pace of commercialization based on this report alone.
- Comparison
- The report mainly emphasizes the architectural upgrade of HSD V2.0 relative to the previous-generation single-stage end-to-end modular design, without providing quantitative comparisons with peer companies.
- Risks
- Attention should be paid to risks related to mass production of intelligent driving, customer onboarding, intensifying competition, active safety reliability, regulatory requirements, and valuation volatility.
Key data
- Report institutionDeutsche Bank ResearchDisclosed on the report cover page.
- Report date2026-06-30Disclosed on the report cover page as Date 30 June 2026.
- Ticker9660.HKThe Reuters code is 9660.HK and the Bloomberg code is 9660 HK.
- RatingBuyThe report cover page discloses Rating Buy.
- AnalystsBin Wang;Wei HuangThe report cover page lists the research analyst and research associate.
- Core productHorizon Superdrive (HSD V2.0)The second-generation full-scenario autonomous driving system.
- Architectural changeWorld model + reinforcement learning dual-engineShift from a single-stage end-to-end modular design to a dual-engine framework.
- R&D accumulationMore than 50 top-conference papers, over 100,000 academic citations, and more than 2,000 global patent applications over the past three yearsUsed in the report to support the company's technological depth.
- Specialized laboratories4 laboratoriesFocused on general AI, embodied intelligence, agents, and acoustics.
Impact & implications
If HSD V2.0's dual-engine architecture and OCC active safety capabilities can be successfully mass-produced and commercialized, Horizon Robotics is expected to strengthen its competitive position in intelligent driving systems and the automotive intelligent chip ecosystem. For investors, the key question is whether the technology upgrade can translate into customer adoption, vehicle model deployment, revenue growth, and a sustainable competitive moat.
Risks
- The report input does not disclose target price, current price, or detailed valuation assumptions, leaving the investment view without a valuation anchor.
- The technical advantages of HSD V2.0 still need to be validated through mass-production vehicle models, customer adoption, and real-road performance.
- Autonomous driving and active safety functions involve risks related to safety liability, regulatory compliance, and user trust.
- Competition in the intelligent driving industry is intense, and algorithms, chips, toolchains, and OEM customer ecosystems may all affect the company's commercialization.
- Deutsche Bank discloses that its research views may be adjusted as market conditions and company fundamentals change.
What to watch
- The deployment pace of HSD V2.0 in mass-production vehicle models and the number of customers.
- The actual improvement in computing efficiency and cost from the Journey 6P chip and toolchain optimization.
- The real-world performance of OCC in active safety scenarios such as AEB, AES, and AMAP.
- The speed at which the company's patents, papers, and laboratory R&D成果 are converted into commercial products.
- Whether subsequent reports disclose target price, earnings forecasts, order progress, and valuation methodology.