--- title: "Horizon Robotics (4Q25 Trans): 2026 auto rev. target +60%---" type: "Topics" locale: "zh-HK" url: "https://longbridge.com/zh-HK/topics/39384663.md" description: "We are confident we can keep GPM above 60%. That would be a high level." datetime: "2026-03-19T16:21:32.000Z" locales: - [en](https://longbridge.com/en/topics/39384663.md) - [zh-CN](https://longbridge.com/zh-CN/topics/39384663.md) - [zh-HK](https://longbridge.com/zh-HK/topics/39384663.md) author: "[Dolphin Research](https://longbridge.com/zh-HK/news/dolphin.md)" --- > 支持的語言: [English](https://longbridge.com/en/topics/39384663.md) | [简体中文](https://longbridge.com/zh-CN/topics/39384663.md) # Horizon Robotics (4Q25 Trans): 2026 auto rev. target +60%--- **Below is Dolphin Research's curated transcript of**$HORIZONROBOT-W(09660.HK)**'s FY2025 earnings call. For our earnings take, cf. '**[**China's 'NVDA alt'? Is Horizon at risk?**](https://longbridge.com/en/topics/39384481)**'.** **I. Key takeaways from the results** 1\. **Guidance**: Management raised its multi-year Avg. revenue growth outlook to 60% (prior 50%). Auto-related revenue growth is set to accelerate to approx. 60%. The 2026 HSD shipment target is ~400k units. 2\. **Key financials**: 2025 revenue RMB 3.76bn (+57.7% YoY). Blended GPM 64.5% with auto GPM at 67.2%. Product solutions revenue RMB 1.62bn (+144% YoY). Licensing and services revenue RMB 1.94bn (+17.4% YoY). 3\. **R&D**: R&D expense RMB 5.2bn (+63.3% YoY), mainly for HSD and cloud training. **Management will keep ramping R&D to build a tech moat.** 4\. **Operating loss**: Adj. OP loss RMB 2.37bn (vs. RMB 1.5bn in 2024). Net loss RMB 10.5bn (vs. RMB 2.3bn net profit in 2024), mainly due to fair value changes of CARIAD CBs. 5\. **SoC shipments**: Auto-grade SoC shipments exceeded 4mn units (+39% YoY). ASP rose over 75% YoY. **SoCs supporting NOA reached a 45% mix, up nearly 5x YoY.** **II. Earnings call details** **2.1 Management highlights** **1\. Smart driving industry trends** a. In 2025, ADAS penetration in China’s PV market hit a record 68%. Higher-end models with highway and city NOA reached a 43% mix, doubling vs. 2024. b. In the sub-RMB 200k mass market, the sales mix of China-brand NOA models rose from 5% at the start of 2025 to over 50% by year-end. This underscores rapid mainstreaming. c. Among China-brand smart EVs, higher-end NOA models made up 62% of sales. That is far above JVs at 13%. **2\. Market share and competition** a. With a 47.7% share, Horizon remains No.1 ADAS solutions supplier to China-brand OEMs. The lead is solid. b. In higher-end smart driving, Horizon holds 14.4% share, nearly tied for No.2 with Huawei at 15.2%. Together with NVIDIA, the trio forms the first tier with a combined 89% share, a 'one dominant, two strong' structure. c. In sub-RMB 200k China-brand NOA, Horizon has 44% share. It ranks first in the segment. **3\. SoC shipments and mix** a. Full-year shipments topped 4mn units (+39% YoY). NOA-capable SoCs were 45% of units and contributed over 80% of revenue. b. ASP rose over 75% YoY, achieving volume and price gains. Over 95% of NOA SoCs were jointly delivered via ecosystem partners. c. Over 110 new program wins were added during the period. Coverage now spans major China brands and JVs. **4\. HSD (Horizon SuperDrive) city NOA** a. HSD entered mass production in Nov 2025, first launching in the ~RMB 150k mainstream. It is China’s first one-stage end-to-end production large model for smart driving. **Shipments exceeded 22k units in just over a month post-launch.** b. For launch models, the high-trim configuration centered on HSD accounted for 83% of sales. This breaks the traditional automotive 'pyramid' structure. c. Over the CNY holiday, smart driving mileage share reached 41%, nearing human-driven share. Management believes once it passes 50%, dependence becomes irreversible, enabling a subscription model. **5\. Globalization** a. Overseas: 11 OEMs and 40+ export models secured for mass production. Cumulative lifecycle export nominations reached 2mn units. Mainstream overseas models of China’s top-5 exporters carry Horizon tech. b. JVs: 9 JV OEMs with 35+ program wins cumulatively. Models outside Volkswagen Group now exceed 60% of the mix. c. Intl OEMs: Won three international OEM programs for markets ex-China. Total lifecycle volumes involved reach 10mn units. **6\. IP licensing and services** a. **Beyond the VW JV, Japan’s largest OEM and auto parts group has started algorithm and software licensing. It already became a top-5 customer in 2025.** **7\. Product outlook** **a. Journey 6M single-SoC end-to-end city NOA is now in mass production on vehicles. In 2026, with partners, city NOA will move from the ~RMB 150k segment down to the ~RMB 100k mass segment.** **b. The Journey 6P + HSD combo will launch HSD 2.0, targeting a 10x uplift in MPI. The goal is to set the industry benchmark.** c. China’s first full-vehicle AI Agentic SoC and full-vehicle AI Agentic OS will be unveiled, enabling cockpit-driving fusion and concurrent open-cloud plus HSD city smart driving. Details will be announced at the Apr product event. **d. The Journey 7 family, built on a next-gen VMM architecture, is under development with targets comparable to Tesla’s next-gen SoC. It is slated for a 2027 launch.** **8\. L4/Robotaxi frontier** **a. In 2H25, together with ecosystem partners, Horizon plans pilot Robotaxi operations in China using the HSD base model. This prepares for the next stage of tech evolution and model commercialization.** **2.2 Q&A** **Q: What drives the high revenue growth, and how to sustain a 60% CAGR? What is the 2026 HSD shipment target?** **A: Auto-related revenue growth will keep accelerating. It is 54% for 2025, and we are confident it will reach around 60% in 2026. From both volume and price:** **On units, 2025 is the first year of mass adoption, but most OEMs will only add highway NOA or above in 2H or even Q4. We shipped 4mn smart-driving SoCs last year, with nearly half from the higher-end Journey 6 series. We also added 100+ new program wins in 2025, mostly mid- to high-end, which ramp in 2026 and build a strong pipeline. From last year’s 4mn base, total shipments should grow about 35% this year to ~5.4mn.** **More importantly, the mix is shifting. Last year, AD SoCs were 45% of units and ADAS 55%. This year, AD will rise to 65%+. Looking at AD alone, unit growth will be substantial.** **On ASP, AD is priced well above ADAS. Last year, SoC revenue divided by 4mn units implied an ASP below $60, still low in absolute terms. There is ~50% upside vs. current higher-end NOA, and up to 10x vs. city NOA ASP. Over the next few years, ASP gains will contribute more to revenue growth than unit growth.** **In short, unit growth will be several tens of percent this year, while ASP will rise by an even higher several tens of percent. SoC revenue will more than double this year. IP licensing should remain stable.** **On HSD, we had 20 mass-production nominations by end-2024, covering China’s top-selling OEMs. We are in advanced talks with leading OEMs and expect record HSD wins. HSD shipments are expected to reach about 400k units this year.** **Q: How do you see GPM over the next few years, and can it stay above 60%? DRAM price spikes this year—impact and mitigations?** **A: With a 60% Avg. revenue growth outlook, we are confident GPM stays 60%+. First, on competition, our peers are NVIDIA and Huawei—top-tier tech players—so this is not a low-margin arena.** **Second, our core model is SoC plus software licensing. SoC GPM runs 40–50%, while software licensing is near 100%, yielding a high-margin mix. Early HSD builds used DCU delivery to speed installs and reduce coordination, but that won’t last; this year we hand DCU hardware to Tier-1s and stay focused on SoC + software.** **On DRAM, price hikes are meaningful industry-wide, but we locked in memory pricing through end-2025 as one of the largest smart-driving solution shippers. DRAM volatility will not further impact our GPM this year. And as our model centers on SoC + software, other hardware swings do not affect the model.** **More importantly, we counter competition through step-change innovation. When peers compete at the SoC level, we add software and algorithms. When industry profit is pressured, we offer integrated cockpit-driving fusion. This year’s fusion solution merges two memory systems into one, saving OEMs thousands of RMB on DRAM and further reducing costs on harness, thermal, and PCB. This boosts competitiveness through innovation, not price wars. All in, we are highly confident in sustaining 60%+ GPM.** **Q: How does cockpit-driving fusion differentiate vs. Qualcomm and NVIDIA? What is HSD’s roadmap from L2 to L4/Robotaxi and the key milestones?** **A: Fusion is the inevitable tech path and enables a seamless experience. With rapid AI Agent progress, we aim to turn cars into mobile AI Agents via more compute and richer hardware.** **Entering the cockpit raises margin and ASP for Horizon. OEMs get better value and a new selling point. Consumers get an in-vehicle AI assistant with hundreds of TOPS, rich cameras and sensors, and high-quality display and interaction.** **On competition, few players truly enter cockpit from the smart-driving domain. Qualcomm has tried to move from cockpit into AD, but its China AD share is very low. Models with Qualcomm AD SoCs have not delivered impactful AD systems, which is as much about software ecosystems as hardware.** **NVIDIA is more interesting. It has long attempted fusion for a luxury brand overseas, but that project has not mass-produced. Largely, it lacks a strong cockpit-driving fusion solution and third-party system providers; building a full-stack in-house does not fit its model.** **By contrast, Horizon can deliver an in-house integrated stack from SoC to AD software to Agentic OS. Entering cockpit from AD is a higher-dimensional play and more advantageous than the reverse. We see limited competition in AD systems because the capability remains scarce.** **For HSD’s path from L2 to L4, our disclosed MPI goal this year is a 10x jump—from tens of km last year to hundreds this year. By 2027 it should reach the thousands-of-km range, and in specific zones exceed 10k km.** **Combined with remote human intervention to exit edge cases, this enables L4 Robotaxi operations in those zones. In the U.S., Tesla’s FSD has achieved 1,000+ km MPI overall and is testing Robotaxi in some areas, implying 10k+ km MPI with near-zero safety takeovers in certain zones.** **HSD should follow a similar path in 2–3 years. As a tech provider, we will support partners to commercialize Robotaxi in select regions in China.** **Beyond MPI, we track smart-driving mileage share. The 41% share over CNY gives us confidence consumers truly like HSD. If it crosses 50%, machine-driven miles exceed human-driven, and users will accept subscriptions, enabling a business model reset for Horizon.** **** **Risk disclosure and disclaimer:**[**Dolphin Research disclaimer and general disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### 相關股票 - [Tesla (TSLA.US)](https://longbridge.com/zh-HK/quote/TSLA.US.md) - [Huawei (HUAWEI.NA)](https://longbridge.com/zh-HK/quote/HUAWEI.NA.md) - [HORIZONROBOT-W (09660.HK)](https://longbridge.com/zh-HK/quote/09660.HK.md) - [Volkswagen AG (VOW.DE)](https://longbridge.com/zh-HK/quote/VOW.DE.md) - [NVIDIA (NVDA.US)](https://longbridge.com/zh-HK/quote/NVDA.US.md) - [HorizonRobo HK SDR1to1 (HHZD.SG)](https://longbridge.com/zh-HK/quote/HHZD.SG.md)