Horizon Robotics (Trans): Urban NOA Accelerates, Breakeven in 2028

Below is Dolphin Research's compiled Trans of $HORIZONROBOT-W(09660.HK) 1H26 earnings call

I. Key takeaways from the results

1. Guidance

a. Full-year revenue to exceed RMB 5 bn. Based on RMB 2.1 bn in 1H, this implies 2H of ~RMB 2.9 bn or more.

b. Chip shipments of ~5 mn units this year, with a target of 7 mn+ next year. Shipments were 2.2 mn units in 1H (+12.1% YoY), with deliveries clearly back-end loaded.

2. Revenue and GP

a. Total: 1H revenue was just under RMB 2.1 bn, up 32.9% YoY. GP was ~RMB 1.4 bn, also up 32.9% YoY, with GPM flat at 66%, and both revenue and GP at the high end of the Jul guidance range.

b. By segment: Product Solutions revenue rose 14.8% YoY, with GPM at 36.2%. Licensing & Services revenue was RMB 1.1 bn, up 52.7% YoY, with GPM at 90.4%.

c. Scope: to accelerate early HSD customer ramp, bundling domain controllers with core products diluted Product Solutions GPM. Excluding this one-off, GPM would be 48.1%, up 300 bps YoY.

d. Customers: the JV with Bosch and one of the world's largest auto parts groups continued to license algorithms and software. Both remained among the top five customers.

e. Scope variance: the CEO verbally cited revenue of RMB 2.455 bn and Product Solutions of RMB 225.6 mn. Back-solving from segment GPM implies Product Solutions of ~RMB 900 mn and total of ~RMB 2.0 bn, which aligns with the 32.9% growth; this Trans adopts the CFO figures.

3. Opex and OP loss

a. R&D expense (incl. SBC) was slightly above RMB 2.7 bn, while S&M plus G&A totaled RMB 634 mn.

b. Total opex was RMB 3.3 bn, up ~RMB 650 mn YoY. Most of the increase came from cloud training costs.

c. OP loss was RMB 1.7 bn. On a non-IFRS basis, adding back SBC and other non-cash items, OP loss was RMB 1.3 bn.

4. Net income metrics and breakeven

a. IFRS net profit was RMB 3.8 bn, mainly driven by gains on fair value changes of convertible loans.

b. Excluding that, adding back SBC and deducting the one-off deconsolidation impact of Digua Robotics, Adj. net loss was RMB 1.67 bn.

c. Management still targets breakeven around 2028.

II. Details from the call

2.1 Management remarks

1. Industry backdrop and penetration

a. 1H domestic PV retail sales declined over 20% YoY (CFO cited 20.2%), with some local OEMs down over 30%. Meanwhile, vehicle exports rose over 65% YoY, hitting a record high.

b. PV intelligent AD penetration reached 76.1%, up 850 bps vs. 2025. JV OEMs hit 80% for the first time.

c. Models with urban NOA had ~23% penetration, measured against all PV sales rather than smart-driving sales. Penetration in May–Jun already exceeded end-2025 levels.

2. Competitive position

a. ADAS market share rose from below 5% in 2022 to above 50% in 1H for the first time. This is double the No.2 player.

b. In urban NOA compute platforms among Chinese brands, share rose to ~23%, up ~500 bps, lifting rank from No.3 to No.2 and now trailing only a U.S. tech company. That company's share fell nearly 10 pts vs. last year.

c. Management believes fewer than 20% of top OEMs can fully in-house development. Most OEMs will adopt independent third-party solutions, and price wars will accelerate this shift.

d. Cumulative investment: RMB 5 bn+ in 2021–2023 for Journey 5 and rule-based algorithm solutions. RMB ~15 bn in 2024–2026 for five Journey 6 variants, one cockpit-driving SoC, and HSD 1.0/2.0.

e. HSD 1.0 was released in Nov 2025, and 2.0 is slated by year-end. Management said third-party disengagement rates already beat peers.

3. Biz. model: Wintel and ARM+Android dual track

a. Direct revenue is SoC plus software, analogous to Wintel. The software platform maps to Microsoft, and at L4/L5 it is expected to shift to per-mile or subscription fees.

b. Expansion revenue is IP licensing, analogous to ARM+Android, enabling customer-designed chips with stickiness from the ecosystem lock-in.

c. IP licensees include the world's largest NEV maker, one of the world's largest auto parts groups, and China's largest JV OEM.

d. From next year, combining in-house SoC share (direct) and OEM self-developed chips using its IP (expansion), the goal is to take No.1 in the high-end segment.

4. JV brands

a. JV brand share has been stable around ~35% for nearly two years. VW, Toyota and Honda account for 50% of JV sales; including Tesla, about 75% are mass-market models, with luxury slightly above 20% combined.

b. Through the VW JV CARIZON, high-end solutions based on Journey 6 and core IP will launch on seven new VW models this year. Next year it will extend to VW's mainstream platforms, totaling ~20 models.

c. Management said future VW models priced at RMB 40k–50k may also carry its urban NOA. The quoted price band magnitude is subject to caution.

d. Capital: the company redeemed part of VW's convertibles, reducing dilution for all shareholders by 4.9%. VW early converted and became a 9.9% strategic shareholder with a voluntary 12-month lock-up.

e. The Toyota program entered mass production in 1H, first on GAC Toyota's best-selling entry model. Through DENSO, it will serve mainstream platforms and is expected to contribute volumes starting the year after next.

5. Overseas

a. The top six groups account for over 70% of China vehicle exports, while only ~3 smart-driving suppliers can truly support their globalization. The company has won 24 brands and nearly 60 export models, covering the top six groups.

b. Bosch and another Tier-1 secured multi-million and million-level awards for next-gen entry products based on Journey 6, spanning China, Europe, ANZ, LatAm, the Middle East and APAC. This entry SoC (noted as J6V in transcript) has ~20 mn cumulative awards.

c. Overseas remains ADAS-centric, and the company first locks in customers and the HW/SW base with entry products. Journey 6E and 6M have entered multiple awarded programs on sale overseas, and French, German and Japanese OEMs are planning high-end projects.

6. 2H operating goals

a. By year-end, target 70%–80% share on core customers' entry AD models. High-compute SoCs plus HSD must complete mass-production deliveries on their urban NOA models.

b. The integrated cockpit-driving solution (transcript notes it is based on the Star chip) is scheduled to ramp in Q4 as planned. It will go from award to SOP in just a few months.

7. Robotics and frontier apps

a. Associate Digua Robotics has enabled 100+ downstream robot categories and serves 400+ customers. It covers more than half of domestic embodied-intelligence companies.

b. The open-source embodied model has been tested or used by 100+ institutions and enterprises, including Nvidia, Meta, Microsoft, Alibaba, Unitree, and Tsinghua and Zhejiang universities. A next-gen brain model combining a world model and a general understanding model will be released within six months.

c. L4 Robotaxi is under test with partners, and management believes it shares the same foundation as L4 private vehicles.

2.2 Q&A

Q: How are awards progressing on major NEV platforms in 2H, and which product line will drive revenue growth more?

A: Shipments scale on J6M, while revenue growth is driven by high-compute 6P/6H plus HSD. In 2H, an entry AD platform at a leading NEV OEM will scale, with J6M as the shipment workhorse; but volume growth is only part of revenue growth, and higher-value Journey 6 (especially 6P and 6H) plus HSD is the main revenue accelerator.

There have been recent breakthroughs at two leading Chinese OEMs (management named Chery as already deploying HSD, the other was not fully captured in notes), and the HSD platform now covers the top five Chinese brands by sales and the top two JV OEMs. This almost fully covers the highest-volume customer base, and management claims no other company has simultaneously secured the top five local brands and the top two JVs.

These awards will move into mass production by year-end and contribute meaningful revenue next year. The transcript also notes a new VW ID series model carrying the company's SoC and HSD is already on sale, though model naming was incomplete.

Q: With R&D stepped up, how do you balance ROI and when does profitability inflect?

A: Breakeven remains targeted around 2028, with saturated investment maintained until then. Management intends to keep GPM at a high level, as high GPM enables sustained high-intensity investment; 1H revenue is a monetization of past years' R&D, and current R&D determines the next few years' P&L.

On investment magnitude: if R&D spend were only RMB 1–2 bn, industry barriers would be too low and competition easier. For a hypothetical RMB 10 bn revenue with 60% GPM, allocating the entire RMB 6 bn GP into chips and AI R&D would exceed what most OEMs can bear, making it more likely that independent third parties like Horizon lead the technology.

Q: How do you see HSD volumes in 2026/2027, and what are the key milestones for mass-production ramp?

A: Chip shipments of ~5 mn units this year, with a target of 7 mn+ next year. HSD has secured awards at all top-five Chinese brands and covers most of their sales; at a leading NEV customer, both entry AD and urban AD are covered and already in mass production.

On the JV side, core program awards have been won at the top two JV OEMs, meaning customer mix is no longer overly reliant on local brands. Management did not provide a standalone HSD install base metric, and only gave total chip shipment guidance as above.

Q: What is the timeline from tape-out to customer SOP for Journey 7, and how will it contribute to revenue and shipments?

A: Tape-out is planned for early Q2 next year, with a 2027 launch, and no quantified contribution provided. Journey 7 targets L3/L4 scenarios with a major compute uplift, and it is natively optimized for next-gen larger-parameter HSD models and local LLMs for the cockpit; R&D is progressing well.

Multiple leading OEMs and Tier-1s have proactively engaged even while the product is still in development. Management did not answer the tape-out-to-SOP interval or provide revenue/shipment contribution metrics, and only stated the goal to make Journey 7 a global benchmark for autonomous-driving chips.

<End of text>

Risk disclosures and statements:海豚研究免责声明及一般披露