The Zhitong Finance App learned that on November 17, CITIC Securities released a research report. The unit price and gross margin of Haizhi Technology Group (02706) smart cars are higher than those of the Atlas business, and the release of demand from existing customers and the implementation of new industry projects are expected to jointly drive revenue growth and profit improvement. With the increase in the scale of revenue, the increase in the share of revenue from smart companies, the gradual dilution of period expenses, and the decline in share payment fees, it is expected that the company's net profit returned to the mother will correct in 2029 and enter an inflection point in profit statements. The reasonable equity value of HK$19.014 billion was given to the company, corresponding to PS 18.6 times 2026 and 12.8 times PS in 2027. Target price is HK$49.0. First coverage, giving a “buy” rating.
CITIC Securities's main views are as follows:
Haizhi Technology has been deeply involved in map computation and knowledge mapping for more than ten years, and has now formed two business lines: Atlas Map Solution and Atlas Agent. As enterprise AI moves from knowledge questioning to core business processes, the importance of trustworthy knowledge and controllable execution continues to increase; we expect the compound growth rate of China's enterprise-level AI smart device market to be around 99.5% from 2025 to 2029. The company's Atlas agent relies on AtlasGraph to process enterprise data and complex relationships, uses Ontology to organize business objects and rules, and calls business tools through AtlasSharness to make task execution evidence-based and processes traceable. The smart procurement needs of existing Atlas customers have gradually been fulfilled, and new scenarios such as mining, oil and gas have been implemented one after another. The unit price and gross margin of smart sports customers are higher than those of the Atlas business, and the release of demand from existing customers and the implementation of new industry projects are expected to jointly drive revenue growth and profit improvement.
Company summary: The map establishes a solid basic market, and smart entities open up a second growth curve. Haizhi Technology is the earliest industrial AI solution provider for layout calculation and knowledge maps in China. The core team covers the Internet platform, industrial customer service and high-performance computing backgrounds. The company's products have gradually evolved along the “data governance-knowledge map-graph database-industrial intelligence”. After years of development, it has now formed two business lines: Atlas Map Solution and Atlas Agent. In the first half of 2026, the company's revenue was 296 million yuan, an increase of 70.4% over the previous year. Among them, Map/Smart's revenue increased by 45.0%/135.6% respectively. Smart companies' share of revenue rose to 38.8% from 28.1% in the same period last year, and its gross margin was 52.9%, 13.4 percentage points higher than that of the Atlas business. Combined with improvements in operation and contract execution efficiency, the company's overall gross margin increased 6.2 percentage points to 44.7% year on year. The Atlas business continues to grow, and smart devices are gradually becoming the main incremental increase in revenue growth and profitability improvement.
Industry analysis: Enterprise AI is moving towards trusted execution, and the integration of graphics and models opens up space for industrialization. 1) Graph model integration improves knowledge reliability: Knowledge maps organize enterprise data with entities, relationships, and rules, and graph integration further improves knowledge construction efficiency and model reliability. According to our estimates, the market size of enterprise knowledge mapping solutions in China is expected to increase from 10.44 billion yuan in 2024 to 30.12 billion yuan in 2029, with a compound growth rate of about 24.1%. Replication of mature industries and expansion of new scenarios support market growth.
2) Harness ensures controllable execution of intelligent tasks: Agents understand tasks through models and call tools to complete work. Harness is responsible for authority, status, and process management to ensure stable and controllable execution. We expect China's enterprise-level AI smart device market to exceed 330 billion yuan in 2029, with a compound growth rate of about 99.5% from 2025 to 2029. Increased payment penetration rate and deepening individual customer purchases will drive market growth.
3) Platform reuse and governance capabilities have become the focus of competition: industry competition shifts from project delivery to industry knowledge accumulation, product reuse, and execution and governance; policies simultaneously promote large-scale application and standardized development, and platform-based manufacturers with data governance, knowledge organization, and execution and control capabilities are expected to benefit.
Core advantages: Leading image and model integration technology, industry body sedimentation to build barriers. 1) The high-performance graph calculation base supports complex relationship analysis: The company's self-developed AtlasGraph database broke the world record at the time in the 2023 LDBC test, and AtlasGraphRag was launched on this basis. It can also search for text information, business objects, and related relationships related to the problem, adding factual evidence and relationship paths to the big model.
2) The accumulation of business entities and execution rules enhances customer stickiness: Atlas agents integrate Model, Ontology, and Harness and are compatible with more than 100 major language models; when customers change models, existing business entities, tool interfaces, and execution rules can still be retained, which helps reduce repeated development and migration costs.
3) Existing customer product upgrades and new industry replication support growth: The company deposits financial, public service and energy industry experience into knowledge models and tool components to promote Atlas customers to purchase additional smart products; at the same time, the company is expanding mature solutions to new scenarios such as mining, oil and gas, and advanced manufacturing to expand the scope of smart business.
Risk factors: The development of AIAgent and knowledge mapping technology falls short of expectations; AI technology has flaws or is being misused; AI computing power supply falls short of expectations; industry competition intensifies; enterprise data security risks, etc.
Profit forecast, valuation and rating: We expect the company's 2026/2027/2028 revenue to increase by 46.8%/45.3%/40.3% year-on-year, and net profit to mother of -1.11/-0.63/-0.24 billion yuan, respectively. Excluding adjustments such as share payments and listing fees, the core net profit was 0.39/0.57/0.76 billion yuan, respectively, and profits have been achieved at the core management level. With the increase in the scale of revenue, the increase in the share of revenue from smart companies, the gradual dilution of period expenses, and the decline in share payment fees, we expect the company's net profit to return to mother in 2029 to correct and enter an inflection point in profit statements. In terms of valuation, taking into account the PS Act and the DCF Act, we gave the company a reasonable equity value of HK$19.914 billion, corresponding to PS 18.6 times in 2026 and 12.8 times PS in 2027. Target price is HK$49.0. First coverage, giving a “buy” rating.