RoboSense: LiDAR Volume Leader Eyes Full-Stack Robotics Play From a Shaky Profit Base

RoboSense (HKEX: 2498) shipped 303,000 robotics LiDAR units in 2025 — a 1,142% year-over-year increase — and claims the No. 1 global position in robotics LiDAR by volume. Those numbers, if confirmed by audited financials expected in H1 2026, would represent a genuine inflection in non-automotive robotics adoption. The harder question is whether the Shenzhen-based company can convert that volume into a sustainable business before its cash position forces a strategic reset.

Business Overview

Founded in 2014 and publicly listed on the Hong Kong Stock Exchange in January 2024, RoboSense employs approximately 1,500 people across operations in China, Germany, and the United States. The company raised roughly $207.7 million in pre-IPO financing and an additional $45 million in disclosed post-IPO funding. It operates across two primary segments: automotive ADAS, which serves 350+ OEMs and Tier 1 suppliers, and non-automotive robotics, which now counts 3,400+ customers across delivery, warehousing, cleaning, lawnmowing, and humanoid platforms.

Q2 2025 consolidated revenue reached RMB 460 million, up 24.4% year-over-year and 38.9% quarter-over-quarter. The robotics segment contributed RMB 150 million of that total on approximately 34,400 units at a 41.5% gross margin — meaningfully above the consolidated gross margin of 27.7%. On a last-twelve-months basis, however, the company carried a net loss of approximately RMB 383 million against RMB 639 million in R&D expenditure. Path to profitability depends substantially on software monetization that has not yet been quantified in any public disclosure. MODERATE CONFIDENCE on financial trajectory.

Technology Stack

RoboSense’s most defensible asset is vertical integration at the semiconductor level. The company designs four chip categories in-house: RISC-V-based data-processing SoCs, digital large-area SPAD-SoCs, 2D addressable VCSELs, and 2D MEMS devices. This portfolio reduces bill-of-materials costs, tightens LiDAR integration, and limits exposure to third-party supply chain disruptions — a structural advantage over sensor competitors reliant on merchant silicon. These claims are company-stated and have not been independently verified through teardown analyses in available sources. MODERATE CONFIDENCE.

Manufacturing is centralized at the Mars Intelligent Manufacturing Base, a ~48,000 m² facility operating at a claimed 95% automation rate. The site holds automotive-grade certifications including ISO 26262, ISO 21434, IATF 16949, and ASPICE CL2 — qualifications that create meaningful entry barriers for OEM supplier programs and are rarely achieved by sensor-focused startups at this scale.

Beyond sensors, RoboSense is expanding into perception software bundled with LiDAR hardware, an Active Camera series (AC2, launched 2025, deployment status: limited), a VTLA-3D AI planning model capable of executing more than ten sequential autonomous actions (deployment status: prototype), and a multi-degree-of-freedom dexterous hand for embodied intelligence applications (deployment status: prototype). No paying customer volumes or revenue figures have been disclosed for the latter two products.

Market Position

The robotics LiDAR segment is RoboSense’s clearest competitive ground. Named 2025 production partners include Coco Robotics (last-mile delivery), LionsBot (commercial cleaning), Rhino Robotics, Kua Robotics, New Stone Robotics, and EasyGo Smart Driving. Bulk-production agreements with unnamed European and North American warehousing companies indicate initial international traction, though contract economics remain undisclosed.

The 41.5% robotics gross margin in Q2 2025 is the most significant data point in the bull case: it demonstrates that service robotics commands higher per-unit economics than automotive ADAS, and that mix shift toward robotics should structurally improve consolidated margins as volumes scale. The risk is commoditization. LiDAR ASP compression is already visible in cost-sensitive indoor mobile robotics segments, and camera-only or radar-fusion architectures continue to attract investment from integrators seeking to reduce sensor costs.

The strategic expansion into cameras, AI planning models, and dexterous hands introduces execution risk and potential channel conflict. Integration partners who currently source LiDAR from RoboSense may reconsider that relationship if RoboSense begins competing at the perception software or end-effector layer.

Outlook

Three catalysts will determine whether RoboSense’s 2025 volume story translates into an investable thesis. First, the 2025 audited annual report — expected H1 2026 — needs to confirm the 303,000-unit figure, segment margins, and provide any quantification of software revenue. Second, named multi-year contracts with disclosed economics from EU or North American warehouse automation customers would validate international scaling. Third, a breakeven quarter, which becomes plausible if robotics mix continues to shift consolidated gross margins toward the 30–35% range.

Geopolitical exposure is a structural overhang. As a China-headquartered sensor and AI company scaling in the EU and North America, RoboSense faces regulatory scrutiny, data localization requirements, and potential procurement restrictions that are difficult to price but impossible to ignore. The company’s HKEX listing provides capital market access but does not resolve the underlying country-of-origin risk for defense-adjacent or critical infrastructure customers in Western markets.

RoboSense is a credible, operationally mature supplier with demonstrated manufacturing scale and a genuine robotics volume inflection. It is not yet a profitable business, and several of its most ambitious product lines remain unproven at commercial scale. Rating: CONTENDER.