From data asset compound interest to smart service efficiency release: A re-examination of the medium- to long-term value anchors of Minglue Technology (02718)

In early trading on August 3, the stock price of Minglue Technology (02718) experienced a strong rebound. At one point, the intraday increase exceeded 20%, reaching a high of HK$55.9. At the close, the company's stock price was HK$53.0, up 15.62%, with a total market capitalization of HK$8.103 billion. On July 31, the ban on large-scale restricted shares was lifted for the first time since the company first experienced its listing. When the selling pressure is phased out, the focus of the market's attention will return to a more fundamental question: as Agentic AI accelerates from a technical concept to an industrial inflection point for large-scale commercial use, is the closed technology loop and policy card position previously completed by Mingluo Technology sufficient to support medium- to long-term value revaluation?

Currently, the competition rules for enterprise services are being fundamentally rewritten — value anchors are shifting from model parameter scale to implementation of business results, and pricing power is shifting from functional benchmarking to distribution of effects. Minglue Technology has a multi-dimensional layout at the technical, product, and business levels: the end-side model Mano-P and the inference framework Cider were successively open sourced, AI Native hardware Octo was officially released, and Octo, the world's first open source trusted agent collaboration network, was launched, creating a complete closed loop from perception, reasoning, execution to collaboration. At the same time, the implementation of the end-side AI filing system and the entry of intelligent devices into the national industrial roadmap for the first time mark that the industry has officially entered a new stage of compliance and large-scale development. The policy framework resonates with the first-mover advantage of technology. Minglue Technology's growth narrative is moving from the continuous deepening of the data intelligence infrastructure to the accelerated implementation of the second growth curve for smart services. On July 29, the company announced that it plans to acquire about 19.07% of Pulian Software's shares, incorporating scenarios such as financial sharing and treasury management of large group companies into the business layout, further broadening the implementation space of intelligent agents in the field of advanced enterprise management.

Technology card position: software and hardware full-stack layout and open source strategy to seize the core card position of Agentic AI infrastructure

Agentic AI is moving from technical paradigm verification to a key inflection point for large-scale commercial use, and the pace of the industry is clearly accelerating. Minglue Technology has formed a technology chain covering various aspects such as end-side models, inference engines, agent collaboration platforms, and hardware entrances.

The foundation of in-depth technology lies in breakthroughs at the model level. The company's open source end-side GUI-VLA model Mano-P and supporting inference framework Cider have jointly constructed a closed loop infrastructure for end-side AI from “operable” to “mass-producible”, laying the prerequisites for a higher-level collaborative ecosystem.

In June 2026, the company released Octo, the world's first open source trusted agent collaboration network, which is open source based on Apache 2.0 and other protocols and supports private deployment. Octo is positioned as an interconnection infrastructure between agents. Currently, the network has covered about 1,500 employees within the company, supporting more than 4,200 agents to invest in actual business processes.

The most long-term advantage of Octo Design is that every human-robot collaboration is not only the completion of a task, but also a structured accumulation of organizational knowledge. Through the asset deposition flywheel of “dispatch tasks → acceptance feedback → deposit preferences and skills → be more efficient next time”, Octo has built a unique positive cycle, so that organization-level productivity infrastructure is naturally enriched with every collaborative interaction, achieving true capacity accumulation and intelligent upgrading. This design elevates AI from a “tool” to a “carrier of organizational memory.” For enterprises, their organizational knowledge is systematically recorded, verified, and passed on. As the number of collaborations increases, the depth of the system's understanding of the business context continues to accumulate, and migration costs increase accordingly, forming a compound interest barrier.

The release of Octo marks that Minglue Technology has built a collaborative platform that can schedule, collaborate, and amplify the efficiency of intelligent services. Through Octo, different agents can cooperate in an orderly manner in complex business flows, integrating scattered AI capabilities into a continuous, controllable, and evolvable closed loop of productivity to support large-scale implementation of intelligent services in enterprise scenarios.

Policy catalysis: End-side filing and smart device industry roadmap implementation to accelerate the pace of commercial penetration and the construction of compliance barriers

The official implementation of the end-side AI filing system and the introduction of intelligent devices into the national industrial roadmap for the first time mark the industry's transition from technological exploration to an institutional framework period of compliance and large-scale development. The opening of this policy window has essentially created an institutional entry barrier for manufacturers with privatized deployment capabilities and full-stack self-developed technology.

The outline of the policy framework has taken shape rapidly recently. In July, end-side AI was officially incorporated into the national standardized management track, clearly requiring model deployment to have technical characteristics that are traceable, auditable, and controllable. Earlier, the Ministry of Industry and Information Technology issued the “Implementation Opinions on the Innovation and Development of Artificial Intelligence+Information Communication”, incorporating “smart devices” into the national industrial roadmap for the first time; on July 23, Beijing issued the first special smart policy “Certain Measures to Support Smart Devices to Lead Development”, from common infrastructure, native applications, terminal integration, security management to open source ecosystem system deployment.

Minglue Technology's private AI infrastructure built with the core concept of “open source, can be privately deployed, and can be audited in a white box” is in line with current policy requirements for data security and controllability, and has certain compliance advantages in enterprise procurement in data-sensitive industries.

Market data verified the commercial transformation efficiency of this system's dividends from the demand side. Jost Sullivan predicts that the end-side AI market will jump from 321.9 billion yuan in 2025 to 1.22 trillion yuan in 2029, with a compound annual growth rate of 40%; China's enterprise-grade AI smart device market is expected to exceed 332 billion yuan in 2029, with a compound growth rate of 107% from 2024 to 2029. On the high growth circuit, the policy framework sets the entry threshold, and the technical base determines the depth of penetration — with privatized compliance base product deployment, Mingleu Technology can meet some of the requirements of enterprise customers in this type of procurement. This adaptability forms a certain pre-emptive condition in project bidding in data-sensitive industries.

Achieving Growth: Compounding Effects of Data Assets and Collaborative Network Customer Sticky Agent Services Catalyzing the Second Curve

Enterprise AI competition has moved from an arms race for model power to a deep competition between data asset compound interest and ecological locking effects. The multi-modal data assets accumulated by Minglue Technology in 20 years establish an industry business context for smart devices; the Octo collaboration network continues to accumulate organizational experience through every human-computer interaction, forming a “more used and thicker” barrier. Based on this, DeepMiner's full-stack architecture transforms data insights into a scalable AI workforce and drives the evolution of the business model from software subscriptions to performance sharing. The volume of revenue from smart services is transforming the compounding effects of data assets and customer stickiness into quantifiable growth and implementation.

On top of data assets and collaborative networks, intelligent integrated services are reshaping marketing delivery logic. Minglue Technology injects long-accumulated multi-modal data and AI capabilities into key links, uses algorithms to improve the degree of standardization of core operations, supports large-scale supply with multi-modal content generation, and encapsulates business experience and strategic capabilities into professional judgments for large-scale output, opening up an end-to-end closed loop from strategy to delivery. It contributed more than 100 million yuan in revenue in the first year of 2025, verifying the viability of the model.

The deeper the brightness penetrates into the customer execution link, the larger the budget pool that can be tapped. According to the first coverage report of Huaxing Securities, the company's smart service revenue is about to reach an inflection point. It is expected to expand to about 500 million yuan in 2026 and further jump to about 1.2 billion yuan in 2027, which is more than 5 and 10 times the 2025 level, respectively. Based on recognition of its high growth potential, Huaxing Securities valued the smart device business at a market sales rate of 35 times.

When the data-compounding driving model continues to be optimized, the marginal value of the collaborative network increases due to scale expansion, and the results-oriented pricing model gains premium space as ROI is realized, the triple momentum forms a positive feedback cycle. Minglue Technology's transition from an intelligent data supplier to an intelligent service provider is driving the evolution of its business model to a monetization model based on results.

In summary, within the current price range, the market may be entering an observation window for the repricing of Minglue Technology's industrial logic. The anchors of medium- to long-term value are: the continuous increase in the external adoption rate of the Octo open source ecosystem, the expected compound growth in smart service revenue, and the pace of commercialization of end-side AI resonates with policy drivers. As short-term sales pressure clears up and market attention returns to industrial logic, Minglue Technology's card position in the field of intelligent services will continue to be transformed into a core driving force for the revaluation of the valuation system as the agent economy expands exponentially.