The capital frenzy around general-purpose large models is gradually becoming more rational, with the market urgently seeking the next wave of hard technology growth curves. Against this macro backdrop, the medical AI sector is quietly undergoing a value reassessment, transitioning from the "application layer" to the "infrastructure layer." In recent years, the vast majority of medical AI companies have remained in the "application layer" of development, with their value ceilings consistently constrained by the capabilities of general-purpose large models, making it difficult to establish truly differentiated barriers. In contrast, a breakthrough into the deeper "infrastructure layer" means companies must undertake heavy investments in high-cost, long-cycle specialized foundational model pre-training, the construction of medical data flywheels, and private deployment architectures. This is essentially a shift from "working for general-purpose models" to "mastering the core foundation," evolving from reliance on external capabilities to autonomously defining technical standards, thereby establishing a formidable competitive moat. It is precisely based on this strategic judgment that DIAGENS-B (02526) has chosen a path that is heavier but also offers a deeper moat. On July 4th, the company officially launched the iMedLoop™ Global Medical Imaging Data Platform at the Diaoyutai State Guesthouse, further materializing the construction of these underlying capabilities. This is not merely a product launch; it is a watershed event marking the medical AI industry's shift from "artisanal, project-based exploration" towards "industrial assembly-line" style mass production.
For astute technology investors, what DIAGENS-B is showcasing is no longer a series of isolated "concept models," but rather a "productivity operating system" capable of continuously producing models, validating them, and achieving a commercial closed loop. As the industry's pain point shifts from "model accuracy" to "production efficiency," DIAGENS-B, through its deep reconstruction of the medical imaging data value chain, is attempting to define the core competitive logic for the next phase of medical AI.
The Efficiency Revolution: From Concept to Assembly Line
For a long time, the medical AI industry has been mired in a production paradox of "high input, low output." Dr. Song Ning pointed out during the launch that, using traditional methods, even if all 250,000 relevant doctors in the country dedicated one hour per day, it would take 1,200 years to complete the data annotation and model training for all medical imaging detection projects. This figure starkly reveals the industry's pain point: demand is industrial-scale, but supply remains at the artisanal stage. The past custom development model of "one disease, one batch of data, one project," while capable of creating exquisite "concept cars," cannot support the "mass production" needs of the entire healthcare system's intelligent transformation. This inefficient production relationship has been the core bottleneck restricting the realization of medical AI's commercial value. The true strategic value of the iMedLoop™ platform lies in constructing a complete medical AI productivity foundation, aiming to solve the efficiency problem at the root technology level. Through the iMedImage® foundational model for medical imaging, DIAGENS-B has reduced the annotated data required for training specialized disease models to 1/200 of traditional methods, shortened the development cycle to 1/12, and cut computing power costs to 1/10. These figures signify a significant breakthrough in the marginal cost of medical AI development, making scalable replication possible.
The accompanying iMedStudio intelligent annotation system, through human-machine collaboration, compresses 3D CT annotation from several hours to within one minute and improves expert annotation consistency from 65% to 99%. This extreme refinement of the "data production factor" directly elevates the performance ceiling of models. Looking back from the perspective of 2026, the launch of iMedLoop™ may well be medical AI's "Ford moment"—it no longer pursues building faster carriages but is dedicated to creating an efficient assembly line.
Supply-Side Reform of Data Value
In the investment narrative for medical AI, the "data barrier" was once considered the most solid moat. However, DIAGENS-B, through iMedLoop™, demonstrates a new possibility: when technology is sufficiently advanced, data is no longer a scarce resource but a liquid asset that can be efficiently processed. This shift in perception directly shatters the first ceiling of industry development. The iMedImage® foundational model, with its extremely low data dependency and high transfer learning capability, effectively establishes a universal "industrial standard" in the field of medical imaging. This means enterprises no longer need to "reinvent the wheel" for every minor disease but can quickly generate vertical applications on top of the foundational model through fine-tuning. This transition from "custom development" to "standardized production" will greatly enhance capital efficiency, potentially enabling medical AI companies to achieve manufacturing-like economies of scale.
Simultaneously, iMedStudio addresses the "last mile" problem for the commercial deployment of medical AI. It liberates doctors from repetitive manual labor, allowing them to focus on reviewing and adjudicating complex cases. This "human-machine collaboration" model not only increases annotation efficiency by tens of times but also ensures annotation consistency through algorithmic intervention, thereby providing a solid data foundation for model clinical registration and commercial promotion. A more profound impact is that the closed-loop ecosystem built by iMedLoop™ is forming a self-reinforcing commercial flywheel. As more hospitals and experts connect to the platform, the volume of data and number of models on the platform will grow exponentially. The continuous circulation, iteration, and optimization of this data and these models within the platform will generate significant network effects.
Redefining Industry Organization
A noteworthy detail from the iMedLoop™ launch ceremony was the presence not only of leaders from the technology sector but also representatives from regulatory authorities, authoritative media, industry associations, data exchange institutions, and top-tier hospitals. This gathering of stakeholders from government, industry, academia, research, medicine, and application itself sends a strong signal: the development of medical AI has transcended the scope of individual companies, rising to the level of national strategy and industry consensus. The platform constructed by DIAGENS-B through iMedLoop™ is essentially playing the role of an "industry connector." It links upstream data resources, midstream technology development, and downstream clinical applications, reducing transaction costs across the entire industry chain through standardized interfaces and protocols. Within this ecosystem, hospitals can more easily access advanced AI tools to improve diagnostic and treatment efficiency; research institutions can leverage the platform's data and computing power to accelerate medical discoveries; and DIAGENS-B captures ongoing platform value by providing infrastructure services.
Furthermore, the agenda of this launch event effectively outlined a macro roadmap for medical AI industry evolution. The entire process, from policy guidance and trusted data space construction at the strategic level, to academic consensus from leading experts on cutting-edge clinical scenarios like large models and smart hospitals, to the official unveiling of iMedLoop as the underlying technological driver, followed by ecosystem partnership signings and panel discussions, represented a high-level integration of policy direction, academic consensus, technological foundation, and industry collaboration. This was not an isolated product promotion but a systematic alignment of industry elements. It sends a clear signal to the capital market: the medical AI industry is transitioning from the early stages of "single-point breakthroughs" and "project-based showcases" to a new cycle characterized by "platformization, industrialization, and ecosystem development."
For the capital market, the value logic of DIAGENS-B is undergoing a qualitative change: it is no longer merely a "company that makes medical imaging AI," but is becoming an "industry organizer" that orchestrates the means of production, tools, and scenarios for medical AI. As the industry shifts from a battle of models to a battle of productivity, DIAGENS-B, with its platform strategy, has already seized the core position in the value chain. From a capital market valuation perspective, DIAGENS-B's potential value ceiling is clearly quantifiable. Public data suggests the long-term market space for integrated platforms covering domestic medical imaging data, model training, and clinical services in China exceeds HKD 100 billion. DIAGENS-B's iMedLoop builds an industry-scarce, fully closed-loop productivity foundation, addressing the development needs for models across thousands of diseases for thousands of hospitals. Combined with multiple revenue streams from MaaS subscriptions, model licensing, and medical device registration commercialization, the company possesses a long-term industrial foundation and growth space capable of supporting a market valuation in the hundreds of billions. It represents a rare platform-type investment target in the Hong Kong market with the potential to reach a valuation in the hundreds of billions within the large model and 18A medical AI sectors.
Comments