AI Highlights · The full text is about 2613 words, and it takes 8 minutes to read

1. China's medical AI stocks collectively stormed on February 14, pushing the sector index to rise by more than 12% in a single day.

2. Generative AI technology provides a new narrative for the medical AI field and is expected to solve existing pain points.

3. Pharmaceutical companies, medical information companies, and Internet diagnosis and treatment companies have accessed the DeepSeek model.

4. However, the healthcare AI market still faces many challenges, such as data complexity, physician trust, data security, and patient privacy.

5. In the future, the threshold for medical AI applications is expected to be lowered, and more small-scale dark horses are expected to emerge.

The above content is generated by Tencent's hybrid model and is for reference only

Author / Boiled under the stars

Edit/Spinach's Starry Sky

Typesetting/Coriander under the stars

After the Chinese AI company DeepSeek released DeepSeek V3 on December 26, this open-source model with low training cost that subverts industry perception quickly occupied the front page of the global AI field. This mysterious force from the East not only caught Silicon Valley off guard, but also gave a huge shock to the capital market, which has never been moved by hot spots. Needless to say, A-share concept stocks have jumped up and down, and even the market value of many Chinese companies such as overseas-listed Alibaba (BABA) is also tumbling.

For a time, the argument that Chinese assets are in a revaluation moment has filled every corner of the stock community.

On February 14, the medical AI stocks in the A-share market collectively went berserk, with more than 20 stocks rising by 10%+, boosting the sector index to rise by more than 12% in a single day, and Jiahe Meikang (688246), Anbiping (688393) and other stocks are happy to raise the limit. Hong Kong stocks are naturally synchronized, Yidu Technology (02158) and Ali Health (00241) rose by more than 10%, while Ping An Good Doctor (01833), Jingdong Health (06618) and other stocks also followed suit.

Jiahe Mekang recent trend chart Source: Internet, compiled by the author

In fact, the concept of medical AI is not very new, but the phenomenal explosion of DS this time has indeed given this track another chance to tell a story.

1. Medical AI once went down a narrow path, and generative AI provided a new narrative

The direct catalyst for this change comes from Huawei's just-released Ruijin pathology model, which claims to improve the efficiency and accuracy of pathological diagnosis and alleviate clinical pain points. Coincidentally, JD Health, MedFound and other companies have also recently released large medical models.

Cathie Wood, known as the "female version of Warren Buffett", led the team to release an annual report called "Big Ideas 2025", highlighting that healthcare is the most underestimated AI application field, and it is expected to achieve revolutionary cost reduction and efficiency increase in AI-driven drug development and other fields.

In short, after the release of DeepSeek, medical AI, which has always been like a tree without roots, seems to be really about to enter reality this time.

Of course, medical care has always been one of the main landing scenarios of AI+.

The application field of medical AI in China has long expanded from image-assisted diagnosis to drug research and development, health management and clinical decision support, and a number of manufacturers in subdivided fields such as Hongbo Pharmaceutical (301230) have emerged.

AI medical application field Source: Jiazi Light Year, Soochow Securities Research Institute

But the embarrassing thing is that in fact, the commercialization process in each direction is not too smooth. AI-assisted imaging technology has become more and more mature, but there are also relatively many competitors. AI-assisted drug R&D is on the eve of an outbreak, but it will take time for data accumulation and experimental transformation. The payment side of AI genetic testing is unstable, and the concept of AI health management has become so much that it seems like a cliché in the primary market.

The author believes that this is mainly due to the technical limitations of traditional AI. Traditional AI is mainly based on the use and reasoning of existing data, focusing on efficiency improvement, and has long been mainly used as an auxiliary means, which can be said to be better than nothing. However, when medical AI is advanced to the stage of pharmaceutical research and development, which requires innovation, it is destined to appear to be more than enough and its value is limited.

Fortunately, generative AI, represented by DeepSeek, is maturing and pulling a savior to a medical AI that has no new story to tell.

Generative AI is a type of AI model that focuses more on generating and creating data content, so that knowledge can be efficiently deduced and summarized. This kind of AI has significant advantages in logical reasoning and multimodal data processing, so it has the ability to push the original single-link artificial intelligence intervention to the intelligence of the whole process.

DeepSeek can empower to solve the pain points of AI medical treatment Source: Lei Feng.com, OMAHA Alliance, Artery.com, New Media Technology Review, Observer.com

Second, listed companies move quickly

The pace of players in the biomedical sector "embracing" DeepSeek is also very fast. The third-party testing manufacturer Berry Gene (000710) arranged the research in the Chinese New Year's Eve, and then announced that it would use the data processing and analysis capabilities of DS to improve the detection efficiency. Subsequently, at least a dozen listed companies in the biomedical field have generated and connected to the DeepSeek model.

Access to DeepSeek by pharmaceutical companies Source: Internet, compiled by the author

Hengrui Pharmaceutical (600276) issued the "Notice on Comprehensively Carrying Out DeepSeek Applications within the Company" just after the Spring Festival, claiming to directly link the application of AI with the assessment, with a clear attitude. Innovent Biologics (01801) announced that it will access DS to build a service-oriented information database. Fosun Pharma (600196) has also integrated the DS model into the PharmAID decision agent platform, which is independently developed in-house.

As we mentioned in previous articles, AI technology is expected to greatly shorten the pharmaceutical development cycle in stages such as target discovery and compound screening. Generative AI can further reduce the cost of inference and improve operational efficiency, and the prospects in this regard are promising.

In addition, medical informatization and Internet diagnosis and treatment companies are also actively deploying. Weining Health (300253) directly shouted the slogan of "AI Everywhere Full-Scene Empowerment", and together with Zhiyun Health (09955), Eagle Eye Technology (02251) and other companies, it connected its various medical models to DeepSeek.

Compared with traditional pharmaceutical companies, these companies, which are already rooted in information platforms, are the first to benefit from the progress of AI technology. For example, after JD Health's model is connected to DS, it can achieve multimodal (text, image, and gene) integration in the whole process of triage, consultation, and follow-up, providing greater support for personalized diagnosis and treatment.

Of course, due to the short time since DeepSeek was born, it is difficult for us to link the implementation of AI with the scale of performance, and it is not possible to judge which ones can really achieve business collaboration and which ones are just shouting slogans.

DeepSeek thought about it for a while, and finally "swallowed" the author's question Source: Compiled by the author

Third, the pain points are still there, and there are potential overseas markets

After all, there are still many drawbacks that are difficult to solve for a while in order for medical AI products to be truly implemented.

On the one hand, even if the clinical manifestations are similar between different patients, the underlying conditions and physiological characteristics may be very different, and the resulting medical data is highly complex, and the use of such data requires large models to have high generalization capabilities, so it may be difficult for medical AI products to accurately respond to complex cases. A momentary misdiagnosis in other areas may not affect the overall use, but medical decisions are very serious, and any small misdiagnosis or missed diagnosis can change a patient's life.

Therefore, it is difficult for medical AI to gain the full trust of doctors in clinical practice, and the market penetration is slow. Issues such as data security and patient privacy are so old that I won't go into details.

In short, when medical AI is still in its infancy, the domestic market is difficult to open for a while. The author believes that in fact, the overseas market is also a potential blue ocean that domestic companies can pay attention to.

It has now been proven that Chinese AI companies are in no way inferior to their international counterparts in terms of technology. However, in mature overseas regions, commercial insurance is developed, and the intervention of AI may be beneficial to cost control, so the willingness to pay is expected. However, China relies on medical insurance, and for these emerging products with uncertainty, medical institutions are relatively unwilling to pay. Even the relatively mature AI images are only 30% of the top 3 hospitals that can be purchased on a regular basis, not to mention other emerging AI formats.

In addition, overseas hierarchical diagnosis and treatment is thorough, and the patient's consultation time is relatively long. Medical AI uses family doctors as the carrier to cut into the diagnosis and treatment scenario, and it is much better than the system dominated by large public hospitals in China. Domestic start-ups such as Zhiyuan Huitu have obtained EU CE certification around 2020 and have begun to lay out overseas markets.

Before generative AI gained attention, medical AI had been accumulating technology and promoting scenarios around the proposition of "efficiency improvement" for many years, and had also built a certain industrial foundation, while generative AI's breakthrough capabilities in complex reasoning and multimodal generation have further injected new momentum into the industry, and the market potential has been affirmed by capital.

The story of medical AI is by no means exhausted, and I believe that one of the most important trends in the future is to lower the threshold for application scenarios such as medical AI. On the basis of the current disputes between traditional medical information companies and Internet giants such as Ali Health, there may be more small-scale dark horses emerging, which will give the market a thunderstorm like a deep search.

Note: This article does not constitute any investment advice. The stock market is risky, and you need to be cautious when entering the market. There is no harm in buying and selling.

View original image 58K

Personal opinion, for reference only