From "usable" to "marketable": DIAGENS-B (02526) reports initiate a revaluation of the platform value of medical AI.

Determining whether a medical AI company has truly crossed the commercialization threshold cannot be assessed solely by looking at model parameters, product count, or the volume of press conferences. Instead, three more practical questions must be answered: Has AI capability become a primary source of revenue? Can the same technological foundation continuously generate new specialized tasks? Can the model pass high-level regulatory scrutiny and truly integrate into clinical workflows? DIAGENS-B (02526) provided relatively clear answers to these three questions with its interim results for the first half of 2026, released on August 7. During the reporting period, the company achieved total operating revenue of 108.7 million yuan, a year-on-year increase of 21.0%; gross profit was 80.5 million yuan, a year-on-year rise of 14.0%, maintaining a high overall gross margin of 74.1%. Notably, revenue from model services reached 94.5 million yuan, up 101.1% compared to the previous year, accounting for 86.9% of total revenue. If 2025 was still a validation period for Denshi's business model transformation, the fact that model service revenue represents nearly 90% implies that the company has crossed a critical threshold: the large model for medical imaging is no longer just a technical foundation or an ancillary product feature, but has become the core business driving revenue growth. Changing the Revenue Base: Model Services Contributing More than Total Net Increment The value underpinning Denshi's revenue growth is first reflected in the incremental structure. In the same period last year, the company generated technology licensing revenue of 46.96 million yuan, accounting for 52.3% of customer contract revenue. In the first half of 2026, adjusted model service revenue rose to 94.5 million yuan, increasing its share of total revenue to 86.9%. The additional revenue from model services in just six months was approximately 47.58 million yuan, while the company's overall net revenue increased about 18.89 million yuan, meaning the former amounted to 2.5 times the latter. This indicates that model services not only accounted for the entirety of the company's revenue increment but also effectively absorbed some short-term fluctuations caused by delayed budget approvals, tendering, and acceptance of certain medical imaging software and medical device projects. Denshi's revenue base has shifted from equipment and software sales to model capability output. The rebranding of "technical licensing" to "model services" does not alter the essence of contracts or the revenue recognition method. What is truly noteworthy is that the business content has expanded into three interrelated delivery models: model and technology licensing, iMedMaaS cloud services, and SCTI localized integrated training and deployment machines. These three models correspond to the varying needs of medical institutions regarding data security, private deployment, model training, and computing resources. Thus, what Denshi offers is no longer merely a ready-made algorithm, but a complete production service that transforms hospital imaging data and medical expert experience into specialized AI capabilities. Although the new and old categories are not entirely comparable, the model service revenue in the first half of 2026 has already surpassed the technical licensing revenue of 84.34 million yuan for the entire year of 2025, visually reflecting the rapid expansion of the related business scale. The True Core Asset Is Not the 158 Models, But the System for Producing Models Viewing Denshi merely as a company that possesses 158 specialized models would still underestimate the capabilities demonstrated in this interim report. Traditional medical imaging AI often follows a "one disease, one model" development approach: entering a new disease area necessitates collecting data, organizing annotations, training algorithms, and completing clinical validation anew. This model has long development cycles, requires substantial input from specialized personnel, and offers limited reuse between different projects. Denshi is addressing the scalability issue of medical AI production. The company uses the iMedImage medical imaging foundation model to provide reusable imaging understanding and reasoning capabilities; employs iMedStudio for data processing, professional annotation, manual corrections, multi-person verification, and quality control; relies on iMedMaaS for specialized model training, customization, evaluation, publication, and deployment; and evaluates model capabilities, safety, and applicable boundaries using DoctorBench. Launched in July 2026, the iMedLoop further connects the aforementioned capabilities, streamlining data access, intelligent annotation, quality control, model training, unified evaluation, publication and deployment, and application feedback, forming a full-process platform from medical imaging data to model application outcomes. As of the announcement, over 3,000 professionals have participated in the iMedLoop system, amassing approximately 28.95 million annotated samples. DoctorBench covers three categories of tracks: language models, multimodal models, and clinical task intelligent agents, across 25 medical scenarios and tasks, as well as approximately 12,920 testing projects. In the limited question sets and evaluation scope disclosed by the company, iMedImage ranked first overall, with all eight categories entering the top three. These data point to a fact more significant than "model quantity": Denshi has already organized previously dispersed data, tools, experts, and engineering processes into a production line for medical imaging AI. From Project Collaboration to Regulatory Products, a Clinical Closed Loop Begins to Form By the end of June 2026, Denshi had launched a total of 158 model projects in collaboration with 99 hospitals, including 65 top-tier hospitals, covering 43 human organs or application sites and 61 disease directions. These project collaborations do not equate to 158 commercial products or 99 paying clients; their deeper value lies in the continuous accumulation of task definitions, annotation standards, evaluation methods, deployment processes, and clinical feedback across different specialties. Each completed project not only adds a model to the platform but also a set of professional experiences available for reuse in subsequent tasks. For medical AI to generate true clinical and commercial value, it must also cross the critical threshold of regulatory approval. On May 19, 2026, Denshi's AI AutoVision chromosome karyotype image diagnostic software obtained the Class III medical device registration certificate from the National Medical Products Administration. This product aids in cutting, counting, identifying, arranging, and flagging suspicious abnormalities in karyotype images from peripheral blood and amniotic fluid samples. The importance of this Class III certificate lies not only in the approval of an additional product but also in validating Denshi's ability to convert the medical imaging foundation model into a regulatory-grade medical device, thereby opening up a complete path from "base modelspecialized developmentclinical validationregistration applicationcommercial delivery." While single algorithms can be caught up to, high-quality medical data, clinical collaboration experience, regulatory registration capability, and hospital deployment systems require long-term accumulation. What Denshi has accomplished is precisely the integration of these high-barrier elements into a single production platform. The Platform Flywheel Is in Motion, Growth Is No Longer Dependent on a Single Product Denshi's future growth potential should not be calculated using a static method of "158 models multiplied by single model revenue." Model projects, diagnostic tasks, and revenue contracts do not have a one-to-one correspondence; a simplistic multiplication would overshadow the true value of the platform's business model. A more reasonable way to observe is to see whether two types of reuse can occur simultaneously. Horizontally, there is task reuse. iMedImage provides generic foundational capabilities; iMedStudio, iMedMaaS, and DoctorBench provide standardized toolchains. New specialized tasks can leverage existing model capabilities, data governance methods, evaluation systems, and deployment experiences, thereby expanding the range of organs, diseases, and imaging modalities that the platform can undertake. Vertically, there is institutional reuse. Medical institutions can start from model training or technology licensing and gradually extend to local deployment, system access, model iteration, and new task creation. The three models of cloud services, technology licensing, and localized integrated machines offer multiple entry points for institutions of varying scales and data security requirements. More importantly, practical applications will also feedback into core capabilities: compliance data, testing results, deployment experiences, and physician feedback generated after specialized models enter the clinic can be used for model improvement and subsequent task development, thus forming a closed loop of "foundation modelspecialized modelservice and productreal-world application feedbackmodel iteration." The richer the application, the more mature the models and engineering systems become; the stronger the foundation, the higher the efficiency of new task development and delivery. This cycle of accumulation is the core platform effect that differentiates Denshi from single-point medical AI product companies. Resonating Policies and Investment, Industrialization Window Is Opening The external environment for medical imaging AI is also changing. The five departments of the state have issued "Implementation Opinions on Promoting and Standardizing the Development of Artificial Intelligence + Medical and Health Applications," clearly stating that by 2030, they aim to promote the general application of AI technologies such as intelligent assisted diagnosis in medical imaging at secondary hospitals and above. This sets a clear timeline for demand on the supply side for medical imaging AI. Similarly, the data supply side is rapidly improving. In June 2026, the National Healthcare Security Administration released the "Basic Specifications for the Medical Insurance Imaging Cloud," promoting cross-institutional and cross-regional interoperability of imaging examination data. By the end of June, nearly 440 million pieces of data had been indexed nationwide in the medical insurance imaging cloud. The former opens application demand and the latter builds the data foundation, leading medical imaging AI from single-hospital pilots to a more systematic stage of infrastructure construction. Denshi is also continuously investing in this industrial window. In the first half of the year, the company's R&D costs reached 64.118 million yuan, a year-on-year increase of 67.4%, accounting for about 59% of the same period's revenue. Of this, computing service expenditure was 45.792 million yuan, accounting for approximately 71.4% of R&D costs, mainly focused on upgrading foundational models, data governance, professional workflow development, unified evaluation, and core product development. The company's losses for the period expanded to 55.883 million yuan, mainly due to increased R&D investment, expenses related to listing, rising sales and administrative costs, and reduced other income. From an operational structure perspective, Denshi currently presents a development characteristic of "core business high growth alongside high investment in platform construction," rather than a lack of revenue momentum. By the end of the period, the company held approximately 655 million yuan in cash and cash equivalents, with net current assets of about 701 million yuan and a debt-to-asset ratio of approximately 13%, providing ample financial space for continued R&D, product registration, and commercialization expansion. Conclusion: Denshi Is Selling the Productivity of Medical Imaging AI The most significant message from Denshi's interim report is not just that model service revenue has doubled year-on-year or that it accounts for nearly 90% of total revenue, but rather that several key links surrounding the commercialization of medical imaging AI have begun to operate simultaneously: Foundation models can continuously breed specialized tasks, the data platform can support specialized production, regulatory certificates validate the conversion capabilities of regulatory-grade products, and multiple service models provide real income interfaces for model capabilities. As a result, Denshi's identity has become clearer. It is not merely a large medical model company or just a provider of chromosome diagnostic equipment and software, but is constructing a medical imaging AI research and production acceleration platform that connects data, experts, models, regulations, and clinical applications. The first half of the medical AI journey addressed "Can the model be used?" while the latter half must answer "Can the model continuously produce, deliver at scale, and generate revenue?" The fact that model service revenue accounts for nearly 90% indicates that Denshi has taken the lead in entering this latter stage. As more specialized tasks, medical institutions, and regulatory products connect to the same platform, what Denshi outputs will no longer be isolated AI tools but the foundational productivity necessary for the intelligentization of medical imaging. This is the long-term value behind the company's interim report that warrants greater attention.