--- title: "Hyundai's First Semi-Annual Report Post-IPO: Multispectral AI Starts Making Money in the \"Invisible\" Places" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/43784655.md" description: "On August 28, Haiqing Zhiyuan released its first interim performance report since going public; on September 4, the full interim report was published on the HKEX; and starting from September 7, the company's H-shares were officially included in the list of eligible securities for both the Shanghai-Hong Kong Stock Connect and the Shenzhen-Hong Kong Stock Connect. These three milestones highlight a rather rare AI company: revenue is generated from data centers, power systems, and energy facilities, with technology processing multi-band information including infrared, ultraviolet, and visible light. Its product portfolio spans modules and perception terminals, as well as large model services tailored for specific scenarios. Haiqing Zhiyuan only listed on the HKEX on June 22..." datetime: "2026-09-07T08:45:41.000Z" locales: - [en](https://longbridge.com/en/topics/43784655.md) - [zh-CN](https://longbridge.com/zh-CN/topics/43784655.md) - [zh-HK](https://longbridge.com/zh-HK/topics/43784655.md) author: "[财报研究](https://longbridge.com/en/profiles/2152743.md)" generator: "portal-rs" --- # Hyundai's First Semi-Annual Report Post-IPO: Multispectral AI Starts Making Money in the "Invisible" Places On August 28, Haiqing Zhiyuan released its first interim performance report post-listing; on September 4, the full interim report was published on the HKEX; and starting September 7, the company's H-shares were officially included in the Shanghai-Hong Kong Stock Connect and Shenzhen-Hong Kong Stock Connect target security lists. These three milestones highlight a rather rare AI company: revenue is derived from data centers, power systems, and energy facilities; its technology processes multi-spectral information including infrared, ultraviolet, and visible light; and its products range from modules and sensing terminals to large model services tailored for specific scenarios. Haiqing Zhiyuan only listed on the HKEX on June 22, making this semi-annual report akin to its first operational health check since going public. Revenue growth has been rapid, with large model services becoming the primary income source; however, profit growth lags significantly behind revenue, and gross margins have also contracted. This provides a realistic benchmark for the commercial value of multi-spectral AI: customers are willing to pay, orders are materializing, but scalable replication, profit quality, and collection efficiency still require validation. For investors, this is precisely where Haiqing Zhiyuan's most significant value lies. As AI enters physical spaces, the segments that will likely realize large-scale revenue first may be hidden within sensors, industry-specific data, and security operations & maintenance (O&M) systems. **Behind the 410 million yuan in revenue, large model services have replaced hardware as the main engine** In the first half of 2026, Haiqing Zhiyuan achieved revenue of 410.5 million yuan, an 84.5% YoY increase; multi-spectral AI large model service revenue surged from 66.32 million yuan in the same period last year to 320 million yuan, a 382.5% YoY jump. Based on disclosed data, this business segment now accounts for nearly 78% of total revenue. Meanwhile, multi-spectral AI module revenue fell 33.9% to 69.1 million yuan, and multi-spectral AI sensing terminal revenue dropped 63.8% to 16.9 million yuan. This data paints a clear picture of the company's evolving business mix. Historically viewed as a supplier of optical modules and smart terminals, Haiqing Zhiyuan's revenue focus has clearly shifted toward large model services. These models are not general-purpose Q&A tools for consumers but integrate UV, IR, and visible light data with industry algorithms to identify risks such as abnormal temperature rises, arc discharges, and equipment failures. AI in physical spaces operates on a different logic. While general models process text, images, and video, multi-spectral AI must first solve "what machines can see." Cameras can detect smoke but may miss pre-fire temperature anomalies; manual inspections can spot existing faults but struggle to continuously cover data center racks, power infrastructure, and new energy assets. Multi-spectral sensors convert invisible information into data, which AI then analyzes, judges, and alerts on. Sensing devices, industry data, and algorithmic services thus form a closed loop. Citing Frost & Sullivan data, the Chinese multi-spectral AI market grew from 6.3 billion yuan in 2020 to 20 billion yuan in 2025, and is projected to reach 79.4 billion yuan by 2030; the multi-spectral AI large model service segment is expected to grow from 1.5 billion yuan in 2025 to 8.4 billion yuan by 2030. Whether these forecasts hold remains to be seen, but the projected 41.1% CAGR for large model services outpaces the overall multi-spectral AI market's 31.8%, signaling a shift in industry value toward algorithms, data, and continuous services. Haiqing Zhiyuan has capitalized on this structural shift. According to Frost & Sullivan, the company ranked No. 1 in China's multi-spectral AI market by revenue in 2025, with a 3.3% share; its share in the multi-spectral AI large model service market reached 23%. The juxtaposition of these two figures is telling: while the broader market is highly fragmented with modest leader shares, in the more specialized large model service niche, Haiqing Zhiyuan has established a clear lead. Capital markets may reward scarce businesses with high attention, but sustained operations depend on project renewals, industry replication, and service subscriptions. In H1, Haiqing Zhiyuan secured large model service deployment orders from a state-owned enterprise client in Shanghai and a listed company client in Shenzhen, primarily targeting early-stage risk warnings in IDCs. If these benchmark projects can replicate across more data centers, power facilities, and industrial parks, order visibility for large model services will gradually improve; if the business relies long-term on a few large custom projects, revenue growth could become volatile. **Revenue is surging, but profits and cash flow haven't fully caught up** The most critical aspect of Haiqing Zhiyuan's semi-annual report is the gap between revenue growth and profit performance. Gross profit for H1 stood at 83.3 million yuan, with the combined gross margin declining from 27.5% in the prior-year period to 20.3%, a drop of 7.2 percentage points. The company attributed this to certain large multi-spectral AI large model service orders containing a higher proportion of hardware components, which dragged down overall margins. This means the term "large model services" cannot yet be treated as pure software; delivery still involves equipment, deployment, and engineering implementation. While revenue scales rapidly, profit elasticity has not yet synchronized. Attributable profit to owners was 7.65 million yuan, down 24.7% YoY; after excluding listing expenses and share-based payments, adjusted net profit rose 21.9% YoY to 27.58 million yuan. R&D expenses increased to 35.11 million yuan (+125.3% YoY), with the R&D expense ratio rising from 7.0% to 8.6%. Increased R&D investment isn't inherently negative for a tech-expanding firm. Haiqing Zhiyuan must simultaneously handle sensor chips, optical imaging, edge inference, model compression, and industry algorithms, so heavy R&D spend is unavoidable. However, profit pressure isn't solely due to R&D. Margin contraction also impacted profitability quality; operating profit fell from 14.32 million yuan in the prior-year period to 2.65 million yuan, and financial expenses rose alongside increased borrowing. Going forward, as large model service revenue continues to grow, whether the hardware share in project delivery can decline; whether the same algorithms can be reused across clients; and whether maintenance, subscription, and model upgrade revenues can increase post-deployment are key. If Haiqing Zhiyuan can transform more custom projects into standardized products, margins may see marginal improvement, allowing profit growth to catch up with revenue. Cash flow also demands scrutiny. Operating cash flow for H1 was -34.19 million yuan, a significant narrowing from -68.67 million yuan in the prior-year period (improving by nearly half), but it remains negative. Cash and cash equivalents at period-end reached 595.6 million yuan, driven mainly by proceeds from H-share issuance; meanwhile, total borrowings increased from 201.3 million yuan at end-2025 to 301.2 million yuan. Expanding operations are consuming capital. Inventory at cost rose from 87.82 million yuan to 116.6 million yuan; credit loss provisions for accounts receivable increased from 12.15 million yuan to 17.85 million yuan; and financial asset impairment losses for H1 rose from 2.16 million yuan to 4.71 million yuan. These changes don't invalidate the logic but directly impact cash flow quality. For project-based tech firms like Haiqing Zhiyuan, order scale, acceptance cycles, and collection speed must be viewed together; focusing solely on revenue growth risks overestimating commercial maturity. Customer structure also needs optimization. The top customer contributed 188.1 million yuan in H1, and the second contributed 44.29 million yuan, totaling ~56.6% of total revenue, with the top customer alone accounting for ~45.8%. Large clients can rapidly amplify revenue but also increase earnings volatility. Expanding direct customer coverage and reducing reliance on single projects will be crucial for transitioning from high-speed growth to stable growth. **Multi-spectral AI has proven there's demand; the next hurdle is turning projects into platforms** Haiqing Zhiyuan's opportunity is clear. As data centers expand, rack density, power load, and cooling pressures rise synchronously, making traditional smoke detectors and manual inspections inadequate for covering all early risks; power grids, energy storage, and new energy facilities similarly require higher-frequency, lower-latency unmanned inspection. Multi-spectral AI addresses safety issues in high-value scenarios, where customer willingness to pay typically exceeds that for standard visual recognition. The company is also integrating multi-spectral modules with embodied intelligence. In June 2026, Haiqing Zhiyuan partnered with Cloud Deep Technology to explore multi-spectral AI applications in electric power inspection robot dogs; in July, it signed a strategic cooperation agreement with Singapore's LINKWISE, covering data centers, industrial manufacturing, and new energy scenarios. Here, robots act more as mobile carriers, while the multi-spectral sensing system serves as the "eyes" for identifying temperature rises, discharges, and equipment anomalies. This path is closer to real-world orders than simply pitching robot units and creates clearer customer value. Haiqing Zhiyuan is also advancing edge computing validation based on NVIDIA Jetson AGX Orin, aiming to migrate models from centralized servers to field terminals, reducing network transmission dependency and improving real-time response. If lightweight deployment succeeds, the company's delivery costs, deployment cycles, and customer coverage could improve, potentially extending its business from large projects to more mid-sized clients. Ultimately, Haiqing Zhiyuan has answered the first question: multi-spectral AI generates revenue, and growth is rapid. The harder questions remain: Can large model services shed their heavy hardware delivery burden? Can margins recover? Can benchmark orders replicate cross-regionally? Can operating cash flow sustainably improve alongside revenue expansion? Physical AI may not initially land at scale through general-purpose humanoid robots; sensors, industry models, and security O&M systems are already generating revenue in data centers, power, and energy facilities. Haiqing Zhiyuan holds a leading position in its niche and has demonstrated commercialization speed; the next watershed moment will be transforming this fast-growing project company into a platform enterprise with standardized products, reusable models, and stable cash flow. The semi-annual report offers a promising opening, but the content that truly determines long-term value lies in future orders and income statements. This article is compiled based on publicly available materials including Haiqing Zhiyuan's 2026 interim report, interim performance announcement, prospectus, and HKEX disclosure information, for informational purposes only and does not constitute any investment advice. Markets carry risks; invest cautiously. ### Related Stocks - [01392.HK](https://longbridge.com/en/quote/01392.HK.md) - [00388.HK](https://longbridge.com/en/quote/00388.HK.md) - [80388.HK](https://longbridge.com/en/quote/80388.HK.md) - [HKXCY.US](https://longbridge.com/en/quote/HKXCY.US.md) --- > **Disclaimer: This article is for reference only and does not constitute any investment advice.**