HAIZHI TECH GP (02706) released its unaudited interim results for the six months ended June 30, 2026, on August 25. During the reporting period, the company achieved revenue of RMB 296 million, a year-on-year increase of 70.4%. Given that the same period in 2025 generated revenue of RMB 173 million, accounting for 27.9% of the full-year total, the company's annual performance is poised to sustain the robust growth trajectory seen in the first half.
Gross profit for the period reached RMB 132 million, surging 97.8% year-on-year, with the overall gross margin expanding by 6.2 percentage points to 44.7%. Net loss narrowed by 38.1% year-on-year, while adjusted net loss contracted by 31.5%, demonstrating simultaneous improvements in revenue scale and operational quality. Notably, the Atlas agent business achieved leapfrog growth, generating revenue of RMB 115 million during the period, a remarkable 135.6% year-on-year increase. This segment now contributes 38.8% of total revenue, officially cementing its status as the core growth engine. This milestone also marks HAIZHI TECH GP's strategic transformation from a graph solutions provider to an enterprise-grade AI infrastructure services company targeting complex business scenarios.
Synergistic Growth in Scale and Quality: Atlas Agents Forge the Primary Growth Engine
From a business structure perspective, the company's two major product lines continue to drive growth in tandem, with momentum increasingly shifting toward the high-value agent business. In the first half, Atlas graph solutions generated revenue of RMB 181 million, up 45.0% year-on-year, solidifying the foundational business. Meanwhile, Atlas agents contributed RMB 115 million, soaring 135.6% year-on-year, significantly outpacing traditional business growth and lifting its revenue share to 38.8%.
Alongside rapid revenue expansion, profitability quality has steadily improved. During the reporting period, the Atlas agent segment achieved a gross margin of 52.9%, up 4.8 percentage points year-on-year, driving the company's overall gross margin to 44.7%, a 6.2 percentage point improvement. With industry solutions continuously accumulating, delivery experience being repeatedly leveraged, and product standardization steadily advancing, HAIZHI TECH GP has entered a virtuous growth cycle where scale expansion and efficiency optimization advance in tandem.
The value of business synergies is clearly validated at the customer level. During the period, Atlas agents served 29 clients with an average transaction value of approximately RMB 4 million. The company has historically served over 430 enterprise clients in total. Among customers contributing to agent revenue, 69% had previously deployed Atlas graph solutions, confirming that the customer upgrade path of "building a data and knowledge foundation first, then extending to agent applications" has been successfully established. The data assets and industry insights accumulated through the graph business are continuously translating into growth momentum for the agent segment.
In the enterprise market, when clients extend procurement from infrastructure to intelligent applications, it typically signals that the product has embedded itself into core business processes. For HAIZHI TECH GP, graph solutions establish the customer base, while agents unlock higher-value growth opportunities, creating a robust product synergy and commercial flywheel.
Elevating the Technology Stack: Full-Stack System Reshapes Industrial AI Deployment Paradigm
As large language model capabilities rapidly evolve, enterprises' core demands for AI have fundamentally shifted from "whether it can answer questions" to "whether it can understand business, reliably execute tasks, and achieve stable deployment." The competitive focus of industrial-grade AI is also extending from standalone model capabilities to a systematic engineering capability that integrates data, semantics, processes, tools, and governance.
Aligned with this industry trend, HAIZHI TECH GP has built a complete industrial-grade AI infrastructure centered on "multi-modal database + Ontology + Atlas Harness," forming a full-stack technology loop spanning bottom-tier data computation, mid-tier business semantics understanding, and top-tier intelligent execution and governance. At the data foundation level, the company drives deep integration of graph, time-series, and vector multi-modal data, enabling cross-data-type joint analysis. This allows enterprise data to be organized, computed, and invoked around unified business objects, eliminating the need for matching, stitching, and transferring across multiple data spaces, thereby providing a consistent and efficient data foundation for long-chain reasoning and complex business decisions.
At the business semantics level, the company is advancing Ontology from a traditional data semantics model to an enterprise operating model covering business objects, states, rules, and actions. By incorporating business elements such as customers, contracts, accounts, equipment, and orders—along with their relationships, states, and executable actions—into a unified semantic framework, Ontology enables AI not only to understand "what" the enterprise has but also "how" it operates, serving as the business semantics layer connecting data with agents.
At the intelligent execution level, Atlas Harness is upgrading from a single-task agent orchestration framework to an industrial-grade agent runtime and standardized operating platform. This encompasses full-chain capabilities including context construction, long-task state management, tool and workflow scheduling, traceability, and effect evaluation, advancing agents from "capable of executing tasks" to observable, diagnosable, and optimizable production-grade systems. As of the reporting period, Atlas agents are compatible with over 100 large language models. The company has established strategic partnerships with Zhipu, Hygon Information Technology, and Bank of Ningbo, further integrating the "computing power–model–scenario" value chain and driving deep collaboration between domestic computing power, general-purpose models, and vertical industry applications.
Deepening Scenario Penetration: Benchmark Cases Build Replicable Capabilities, Accelerating Scale-Up Growth
Technological value ultimately requires validation through industrial scenario deployment. During the reporting period, HAIZHI TECH GP continued to deepen its presence in three advantageous industries—government affairs, finance, and energy—while achieving multi-point breakthroughs in emerging tracks such as intelligent mining, telecom operators, oil and gas, intelligent manufacturing, and pharmaceuticals.
In the intelligent mining sector, the company signed a project contract worth tens of millions of yuan with a major coal group, deploying Atlas agents to build a full-chain intelligent decision-making system covering production, safety, and operations. In the oil and gas field, it successively won bids for core business projects from multiple large state-owned enterprises involving production extraction, settlement, and operations. In finance, the company established a strategic partnership with Bank of Ningbo while continuously delivering solutions in specialized scenarios such as financial leasing and IPO prospectus review. In the electric power sector, it is building a graph data foundation around grid topology and time-series states to support core applications like fault location, impact analysis, and equipment maintenance.
Additionally, the company has established a Hong Kong local entity and dedicated team, broadening its financial industry coverage while steadily advancing government affairs scenario expansion. The cumulative base of over 430 industrial clients and multi-industry benchmark project experience provides rich samples for cross-industry replication of technical capabilities. For the industrial AI sector, the value of benchmark projects extends beyond individual orders—it lies in distilling customer needs, industry knowledge, business processes, and delivery experience into reusable ontology models, skill libraries, and scenario components.
HAIZHI TECH GP has clearly outlined its strategy to leverage flagship customer cases as a catalyst, transitioning its revenue growth model from "labor-driven project delivery" to "scaled replication of products and services," thereby optimizing marginal costs and enhancing revenue quality. The continued optimization of the capital structure also provides solid support for long-term R&D investment and business expansion. As of June 30, the company held total assets of approximately RMB 1.627 billion, equity attributable to shareholders of approximately RMB 1.199 billion, and combined cash and cash equivalents plus financial assets at fair value through profit or loss of approximately RMB 972 million. With a debt-to-asset ratio of only 26.3% and no external borrowings, the company maintains a robust financial position.
From bottom-tier graph computing to multi-modal data foundations, from business semantic ontology to Harness intelligent execution, and from the core government, finance, and energy markets to new scenarios in mining, oil and gas, manufacturing, and pharmaceuticals, HAIZHI TECH GP is continuously unlocking the complete "technology–product–customer–growth" chain. As industrial-grade AI enters its deep deployment phase, the company is well-positioned to leverage its full-stack technical moat, industry knowledge accumulation, and customer synergies to drive industrial intelligence services from point breakthroughs to a new stage of scaled growth.