After Tempus soared, the AI4S value of Deshi Technology (02526) is being rediscovered

The Zhitong Finance App learned that on September 17, the intersection of AI and life science once again became the focus of the capital market. At the close of the Hong Kong stock market, Jietai Technology (07666) rose 12.70%, and Insilicon Smart (03696) rose 6.86%; in the US stock market, Tempus AI (TEM.US) once reported $81.12, an increase of 15.93%. On the same day, DSE (02526) closed up 0.39%. The rise and fall of a single day cannot sum up the value of a company, but it provides an observation window worth paying attention to: when the market starts paying higher prices for medical data, scientific workflows, and AI platforms, is the ability Deshi is building still underestimated?

Tempus's operating data gave a clue. The company's data and application business revenue reached US$93.2 million in the second quarter of 2026, an increase of 28% year on year; the company disclosed that the Insights-related business increased 36% year over year. The additional relevant license contracts of US$200 million were added during the quarter. This is the contract amount, not all confirmed revenue for the current period. Tempus management emphasized at the Morgan Stanley Global Healthcare Conference on September 15 that barriers come from hospital connections, compliance agreements, data pipelines, and matching clinical records with molecular, pathological, and imaging information. The company then announced that it will build a platform containing 100,000 genome-wide and related vertical clinical information in the next few years, and expand the target to 1 million. The market's focus is shifting from individual AI products to systems capable of continuously producing high-quality medical data and models.

This is the core direction of AI4S (AI for Science). In common industry definitions, AI4S links machine learning with scientific data, mechanism knowledge, simulation calculations, and experimental verification to improve the efficiency of hypothesis generation, prediction, and knowledge discovery. AlphaFold 3 uniformly predicts multiple biomolecular interactions, and GNOME shows the path from candidate discovery to stability assessment; China's “Artificial Intelligence +” initiative also lists artificial intelligence and scientific research as a key direction. AI4S ultimately relies on repeatable verification and continuous iterative confirmation. As a result, the closed loop of data quality, expert knowledge, and feedback becomes an infrastructure.

If you study Desu along this line, you can see a company that is closer to the essence of AI4S. Deshi organizes real medical problems, imaging data, expert judgment, model training, evaluation, and application feedback into the same production chain. iMedLoop was launched on July 4, 2026, and integrates data access, professional labeling, quality control, training evaluation, deployment release, and application feedback; iMedStudio supports AI pre-labeling, expert revision, multi-person result comparison and dispute arbitration, and related functions have been used in clinical research and internal R&D projects.

This set of procedures precipitates hidden knowledge in medical practice into data and model development capabilities that can be reviewed, trained, and reused. The model assists experts in processing materials, and experts correct model errors. Within the scope of authorization, quality-controlled data can be used for training and verification; in the face of misreports or boundary cases, the platform can design targeted samples based on this, and evaluate the improvement results using independent tests in the project. Every real task is an opportunity to accumulate task definitions, data standards, quality control rules, and deployment experience.

As of the evening of September 17, the iMedloop website page showed that about 29.015 million samples, 466.8 TB of images, 222 active tasks, and 3,172 certified experts had been marked. The user agreement makes it clear that there is no transfer of intellectual property rights to the uploaded data, and the use and benefits are specifically authorized; currently, it is still possible to experience free points, and the actual paid service will be announced separately. Deshi's core competency lies in the organization, governance, and transformation of data, providing a foundation for long-term collaboration with hospitals, experts, and R&D institutions.

Deshi already has an observable commercial foundation. In the first half of 2026, the company's model service revenue was 94.541 million yuan, an increase of 101.1% year-on-year, accounting for 86.9% of total revenue. iMedloop was launched in July, later than the reporting period for the first half of the year. After platforming, it is necessary to observe whether model iteration can lead to more usage and payment, whether data and tools can reduce development and delivery costs, and whether the same platform can be reused across specialties.

These three paths correspond to revenue, cost, and business expansion space, and also form the most imaginative part of Deshi. As tasks increase, the platform may establish methods to continuously improve the model; as methods are reused, the time for new problems to become usable tools is expected to be shortened, and the collaboration efficiency of data and expert networks is also expected to improve. Deshi is building a closed loop of “finding problems — organizational data — training model — verifying results — redelivery” that AI4S values.

Expert feedback, data growth, and retraining alone do not prove that autonomous recursive self-improvement has been achieved, nor are they a substitute for independent testing and deployment monitoring. The evidence supporting the valuation revaluation will come from model iterations, authorized collaboration, ongoing payments, and cost efficiency. Tempus's market reminds the market that only when high-quality medical data enters the actual workflow and generates commercial returns can there be a better basis for obtaining higher pricing. It is the next-generation model production system currently under construction in addition to existing products that Deshi deserves to be re-examined.

If you want to find a target on the AI4S circuit that is close enough to scientific questions and has a platform-based imagination space, judging from public data, Deshi has the key conditions to become the purest and most imaginative target. Its “purity” comes from AI acting directly on medical research tasks; its “imagination” from every real task may become the starting point for the next round of model capabilities and commercial efficiency.