Hai Zhi Technology (02706) 2025 Annual Report: From Graph to Intelligent Body, Countdown to Revenue Structure Reversal

After 44 days of listing, Haizhi Technology (02706) faced its first major test. Listed on the Hong Kong Stock Exchange on February 13, 2026, the company saw a surge of 246% on its first day of trading, with a public offering oversubscribed by 5065 times. The market's enthusiasm for Haizhi Technology was not just about a graph database company, but about an AI intelligent body growth logic driven by "graph model fusion". On March 27, the company released its first annual report since going public. Total revenue was 621 million yuan, a 23.4% increase year-on-year; adjusted net profit was 24.15 million yuan, a 42.6% increase year-on-year. The core growth engine Atlas intelligent body revenue reached 146 million yuan, a significant increase of 68.4% year-on-year. Analyses in the market have mentioned the "68.4% growth rate, establishment of the second growth curve." This article attempts to provide more precise numbers to answer three questions: What is the quality of this curve? How far is it from truly changing the company's revenue structure? Which indicators should investors focus on to continuously validate? Comparing the two business lines Haizhi Technology's revenue consists of two main segments - the Atlas graph solution (core base) and the Atlas intelligent body solution (second growth curve). The full data for 2025 is as follows: The difference in growth rates is clear, but what is more worthy of contemplation are a few easily overlooked details. The average price of the intelligent body is 3.6 million yuan, nearly 30% higher than the graph. This indicates that customers are not just "testing the waters," but are "betting," as a purchase decision of 3.6 million yuan in any large organization requires multiple layers of approval. Being able to sell at this price means the product has entered the customer's core business processes. What's more, such large purchases are not usually sporadic departmental demands, but are strategically positioned by the customer's management as a core strategic level of AI intelligent body construction. Compared to regular software purchases, strategic projects correspond to longer-term budget planning, and the subsequent scalable budget space for full scenario deployment and feature iteration upgrades will be broader. The gross profit margin of the intelligent body is 53.2%, approximately 14 percentage points higher than the graph. In the B-to-B software industry, the gross profit margin is a hard benchmark for measuring productization. Project-based companies typically have margins between 20-30%, achieving over 50% means high standardization, less custom development, and strong technology reuse. The 53% gross profit margin of the intelligent body directly pushed the company's overall gross profit margin from 36.3% to 43.3%, a 7 percentage point increase in a year, which is extremely rare in the B-to-B field. What truly catches the eye in the numbers is hidden in the customer structure: 50% of intelligent body customers are converted from graph old customers. The importance of this number cannot be overstated. It implies that Haizhi has established a path of customer upgrading from "selling infrastructure first, then selling upper-level applications" - customers who have already deployed graph databases and knowledge graphs naturally possess a technical foundation and data assets for evolving into the intelligent body. This significantly reduces customer acquisition costs and improves delivery efficiency. The annual report also disclosed that five customers have made repeat purchases since the launch of the intelligent body product. In the enterprise market, the "first sale" may rely on chance, but the "second sale" can only rely on the product. Being able to get customers to pay twice is the real business validation. Countdown to the reversal: When can the intelligent body take the lead? Once the quality of the curve is confirmed, the next question naturally arises: at this speed, when can the intelligent body surpass the graph and become the company's largest business? The current ratio is 77:23, with a gap of over 3 times. We conducted three scenario simulations: Optimistic scenario: Intelligent body maintains 65% growth, graph maintains 15% growth In the optimistic scenario, by 2028, the revenue share of the intelligent body will be close to half; by 2029, the intelligent body will surpass the graph. Baseline scenario: Intelligent body growth rate gradually slows down to 40%, graph maintains 12% growth In the baseline scenario, by 2030, the share will approach 45%, with a complete reversal in 2030-2031. Conservative scenario: Intelligent body growth rate drops to 30%, graph growth rate 10% Even in the most conservative assumption, the share of the intelligent body is expected to exceed 40% by 2030. All three scenarios point to the same conclusion: the reversal of the revenue structure is directionally determined, with the divergence only in 3 years or 5 years. Moreover, there is an acceleration factor in the simulations that has not been accounted for - the conversion potential of graph customers. At the end of 2025, there were 172 graph customers, with only about 20 converting to intelligent body customers, a conversion rate of about 12%. Another 150+ customers make up a large "conversion reservoir." If the conversion rate increases to 30-40%, the growth of the intelligent body will gain additional internal momentum, potentially bringing forward the reversal timeline. With 1 billion cash on hand, what's the next move? Growth requires ammunition, and Haizhi is not short on this front. As of February 28, 2026, the company's total cash and financial assets exceeded 1 billion yuan. For a company with an annual revenue of 620 million yuan, this means having a reserve equivalent to about 1.6 years of income - in an environment where there is generally a need for funds in the AI industry, this financial strength is solid. Ammunition mainly came from the IPO fundraising (net amount of about 655 million Hong Kong dollars) and accumulated operational reserves. Based on the signals from the annual report, the management's spending priorities are clear: First, prioritize research and development. In 2025, R&D expenses increased by 64.6% year-on-year to about 100 million yuan, far exceeding the growth rate of revenue. The company took the lead in undertaking a national major project, focusing on key technologies for the next generation multi-mode database, while also being selected as the sole lead in a key project in Beijing. The new version of AtlasGraph graph database achieved significant enhancements in dozens of functions such as temporal graphs and data lifecycle management, and the GraphRAG technology also upgraded its parsing capabilities for structured and semi-structured text. The R&D efforts are building momentum for the next generation of products. Second, expand territories. The company's industry map is expanding beyond its core areas of finance and public safety. The annual report mentions several new initiatives: building a unmanned device ontology model for a unmanned area management center in the Guangdong-Hong Kong-Macao Greater Bay Area, covering drones, unmanned ships, and unmanned vehicles; deploying road network operation analysis intelligent body and infrastructure maintenance intelligent body for a city's tunnel company; a large state-owned telecom operator has made repeat purchases for three consecutive years, expanding from graph databases to GraphAgent and other ecosystem tools. High-end manufacturing, healthcare, and intelligent mining are also listed as key next directions. Third, explore international markets. The company has already established an intelligent body application platform, including large-scale integrated models, for a foreign non-bank financial institution, and has explicitly stated its intention to explore the Southeast Asian market leveraging its Hong Kong listing platform. Although overseas revenue is currently small, the Hong Kong listing naturally opens a door for international capital operations. Meanwhile, operational efficiency is also improving. Excluding listing expenses, the net cash flow from operations shrank from 38.5 million yuan in 2024 to 30.2 million yuan in 2025. Moat: Why is it difficult for others to replicate? The market is not lacking in AI companies, nor in players shouting the slogan of "intelligent body". What makes Haizhi hold onto its territory? First, consider its positioning: According to Frost & Sullivan, based on revenue in 2024, Haizhi ranks fifth among industrial-grade AI intelligent body providers in China - ahead of giants like Baidu, Alibaba, and Huawei. However, in the specific field of "AI intelligent body providers based on graph core", Haizhi ranks first with a market share of over 50%. This difference reveals Haizhi's competitive strategy: rather than directly competing with large companies in the general big model space, it focuses on the intersection of graph model fusion, leading in the absolute top of a niche market. As of now, there are three layers that competitors will find difficult to replicate in the short term: Firstly, the underlying technology. AtlasGraph graph database broke the world record in the 2023 LDBC international test with a 45% advantage over the previous record, ranking first in overall performance. Graph databases are long development cycle, high engineering complexity foundational infrastructure that cannot be rushed with money alone. The new version in 2025 continues to widen the gap. Secondly, industry deposits. Serving over 400 customers, covering 100+ industry scenarios, the most significant achievement is in completing the domestic substitution of graph databases for four out of six large state-owned commercial banks - these knowledge ontologies, data governance experiences, and industry methodologies accumulated during delivery are genuine hidden barriers. Thirdly, the business flywheel. The upgrade path from "graph->intelligent body" has already been established, with a 50% customer conversion rate and continuous repeat orders forming a positive cycle. Players who can cover the entire stack are very few - pure big model companies lack the customer base and data foundation like graph databases, while pure database companies lack the application capability of upper-level intelligent bodies. Of course, risks must be considered. Microsoft has already released the GraphRAG framework, and mainstream big model manufacturers are enhancing their graph computing capabilities. If the graph capabilities of general tools improve significantly, it may compress Haizhi's technological differentiation space. However, within the foreseeable 2-3 years, it is almost impossible for general frameworks to achieve the industry depth of Haizhi in specialty fields such as energy grid scheduling, financial fund supervision, and high-end manufacturing with extensive knowledge and high security requirements. How will the market price? Pay attention to these five numbers Returning to the initial question - the expectations behind the 5065 times oversubscription, how much of it has been fulfilled by this annual report? The direction is confirmed. The technology route of graph model fusion has completed a commercial closed-loop in the industrial end, with a customer upgrade path of "first sell graph, then sell intelligent body" supported by a 50% conversion rate and tangible repeat orders. The rhythm is steady. With a revenue growth of 23.4%, a gross profit margin of 43.3%, and an adjusted net profit of 24.15 million, Haizhi has provided a balance of growth and profits in the context of AI industry where increased revenue does not automatically translate to increased profits. Continuous profitability for two years is a rare attribute in the same race track. The imagination space lies in the evolution of the revenue structure. Starting from 23%, the intelligent body may become the largest business in the next 3-5 years. The reserve of 1 billion cash, continuous increase in R&D investment, expansion into new industries and overseas markets, provide upward elasticity for the growth curve. As to whether this logic can continue, each subsequent financial report will be the validation. It is recommended to track five core indicators in the long term: the growth rate of intelligent body revenue, the speed at which the number of intelligent body customers expands from 40, how close the overall gross profit margin can get to 50%? When will the operating cash flow turn positive? When will overseas revenues further breakthrough? A company that is profitable in the AI intelligent body race, has barriers, and is repeatedly endorsed by customers, is worth investors giving it some time.