In mid March, after landing on the Hong Kong stock exchange for less than half a year, Yingtong technology, the first stock of medical AI, handed over its report card in 2021. Excerpt from Yingtong technology's 2021 Financial Summary (data source: Yingtong technology's 2021 annual report) The whole annual report has four highlights: the first is the revenue, 115 million yuan, an increase of 141.6% year-on-year. Yingtong technology has crossed the threshold of 100 million income Club; The second is the loss. In 2021, it lost 143 million, with a year-on-year increase of 79.8%. The sharp increase in R & D costs, sales costs and administrative costs is the main reason for the loss; The third is the number of customers, increasing from 85 to 244. Excluding the income of about 46 million provided by customer a of physical examination center and customer B of optometry center, each customer contributes an average income of 85000 yuan; The fourth is the service fee of AI. According to the estimation of Yingtong technology, AI will be called 4.86 million times in 2021, and the fee charged for each AI test is 19.9 yuan (calculated by dividing the income from providing software solutions based on artificial intelligence by the number of tests). Yingtong technology's revenue of 115 million yuan mainly comes from physical examination center, optometry center and insurance institutions. When price access is only available in some provinces and cities, Yingtong technology hospital has fewer users, only 38; The revenue of physical examination center and visual center is strong, and the revenue contributed by a single user is more than 10 million. Since the income of the physical examination center and the time center has a linear relationship with the number of software calls, it can be inferred that the AI adjustment amount of a single major customer of Yingtong technology is in the range of 300000-1 million times. Fundus examination is a kind of routine examination. Taking the annual physical examination of about 18 million people in meinian health in 2021 as a reference, the AI call volume of Yingtong technology accounts for only 2% - 5%, which has great room for improvement. Last year was the first year of the sales of the fundus screening software and hardware integrated machine airdoc aifundus (1.0). Yingtong technology sold 18.7 million in 41 hospitals and 36 community clinics. New imaging equipment combining software and hardware may have great prospects in the future. In terms of losses, Eagle pupil technology continued to increase in various expenses, and R & D expenses, sales expenses and administrative expenses increased by more than 50%. However, the income of one yuan invested in 2020 is only 0.6 yuan, which rose to 0.8 yuan in 2021. At present, Yingtong technology still has 1.8 billion cash and cash equivalents on its book. If the current development trend remains unchanged, it is likely that it can turn losses into profits in a few years. In 2021, the number of Yingtong technology customers changed greatly. With the cooperation of optometry chain platforms such as Jinggong glasses, Lean glasses and glasses 88, the number of hospital customers also increased to a certain extent, and the overall number of customers increased by nearly three times. However, the small contribution of a single customer's revenue is a serious problem faced by Yingtong technology. With the passage of cooperation time, if Yingtong technology can bring value to the actual operation of enterprises and medical institutions, or obtain the further recognition of C-end consumers, the income of Yingtong technology will further increase. Finally, focus on the AI call cost of Yingtong technology. We often say that pay per view is the ideal profit model of artificial intelligence in the future. Can the 20 yuan medical imaging AI support the 100 billion market? After all approval, the business model is still an obstacle As of the release of the benchmark AI review and approval policy of the guiding principles for the registration and review of artificial intelligence medical devices in early March, 33 types of medical artificial intelligence products in China have been recognized by the China food and drug administration. At this time, some leading enterprises have won 2-3 class III certificates and multiple class II certificates. The review and approval process is no longer a key "barrier" that hinders the progress of medical enterprises. According to the operation data of medical AI after 2020, each enterprise submitting the Prospectus has achieved several times of revenue growth and reached tens of millions after obtaining the certificate. There are two reasons behind it. One is that the sales base before approval is too low, and the other is that the layout of medical AI has been gradually released in the past few years. 2022 is a watershed for medical AI enterprises. The hospital price catalogue in various provinces means that more medical institutions can purchase AI products through bidding and other means, and the sales volume of enterprises will increase. However, the sales in 2021 consumes a certain proportion of stock users. To land the same users without considering the epidemic situation, more sales costs must be paid in 2022. The shifting revenue and costs make it difficult for us to judge the revenue growth trend of medical AI, but based on the AI call cost of 20 yuan per time, we can roughly estimate the ceiling of this market. According to the data of the number of national medical and health institutions by the end of October 2021 released by the statistical information center of the National Health Commission, by the end of October 2021, there were 3147 tertiary hospitals and 10664 secondary hospitals in China. In addition, combined with the CT scanning volume sorted out by the arterial network according to the public data, the daily CT scanning volume of tertiary hospitals ranges from 200 to 500 times, and that of secondary hospitals ranges from 50 to 200 times,. Based on the above data, assuming that the number of scans is 300 and 100 times respectively, and AI needs to be called for each CT scan, medical institutions need to spend 14.673 billion yuan a year for CT related AI enterprises. This means that if pay as you go can be realized, even if the real data is only 20% of the ideal data, medical AI enterprises can still reap a good scale of revenue. Reality is still far from the ideal sales model. Yingtong technology creatively incorporated the two flux scenes of physical examination center and visual center into the system, and was able to harvest 4.86 million calls. However, for AI enterprises such as pulmonary nodules and CTA, it is necessary to cross the two mountains of price access and medical insurance access from entering a well regulated hospital through bidding and selling in the form of software buyout to calling on demand and realizing more realistic payment on a case by case basis. So far, the leading medical AI is still in the stage of price access, and after entering the price, it needs to run for 1-2 years before it is expected to enter the scope of medical insurance. At present, no enterprise has achieved a breakthrough in medical insurance access. Medical AI enterprises still face a long journey. When the market is cold, enterprises need to wait for the opportunity to go public In order to raise all kinds of expenses required for the next stage and help PE who have been waiting for many years to find an exit way, in 2021, a number of medical AI companies submitted prospectuses to seek listing on the Hong Kong stock exchange. Over the past year, in addition to the successful listing of Yingtong technology, Keya medical, Shukun technology and hypothetical medical have all ended with invalid registration documents. Recently, Shukun technology publicly said that it would resubmit the prospectus in the near future. Arterial network has consulted all AI enterprises recently. Whether they have applied for listing but failed, or those at the C-end facing listing, they all say that the enterprises are in a silent period and it is inconvenient to answer in detail. Summarizing the sporadic statements of various enterprises, the listing problems faced by medical AI enterprises do not appear in the business, but at the macro level. Under the influence of many factors, such as soaring stamp duty on Hong Kong stocks, capricious epidemic situation and the increasing trend of de globalization under the geopolitical crisis, the valuation of enterprises in the primary and secondary markets plummeted, the public's pessimistic expectations took shape, and the investment sentiment of the whole market fell to the freezing point. Crisis ridden, enterprises and investors are unwilling to enter the next stage at this moment. For AI enterprises planning to go public, the delay in entering the secondary market will not have a great impact on their business. On the eve of IPO, there are still more than 1 billion enterprises in the market. Before the recovery of the secondary market, should we lie flat or seek ways to make a living? In the year of crisis, medical AI enterprises walked out of different roads. For some medical AI enterprises with small steps and not much capital boost, they choose to promote their own AI products in local and local connected provincial and municipal hospitals. Such companies either have close cooperation with hospitals in the region, or cooperate with listed companies such as Weining health to provide new service support for hospitals and realize small-scale profits with the help of medical AI. Their playing method is consistent with the traditional informatization. Enterprises with large-scale capital but not in the listing stage, such as AI enterprises such as medical quasi intelligence and Zhiyuan huitu, still focus on the deepening of products. Medical quasi intelligence takes full image as its strategy, which focuses on the detection of breast and lung diseases. Zhiyuan huitu is similar to the playing method of Yingtong Technology, focusing on the detection of fundus diseases. In contrast, Huiyi Huiying and Shenrui medical began to explore the construction of industrial ecology, manufacture the collaborative closed-loop of medical IT infrastructure, medical technology and clinical departments, and emphasize the closed-loop of data and application flow. Under this layout, enterprises bet on the digitization of the process of clinical application in the future. Huiyi Huiying's product system has a big data cloud platform, a data center, and AI auxiliary diagnostic tools such as aorta, bone and breast. The whole closed loop has begun to run. Shenrui is very strong in auxiliary diagnosis, won multiple class III certificates, and a large number of medical AI related papers landed in authoritative journals. After the acquisition of Yitu medical in July last year, the layout of Shenrui medical in medical it has been very mature. In the face of the market value crisis of Yingtong technology, Shukun technology, Keya medical and Bodong medical are trying to expand their own business scope as much as possible. On the one hand, they are expanding their business horizontally from "Ai image" to "Ai informatization" and "Ai health management". On the other hand, they are expanding vertically from separate software to the combination of software and hardware. For example, Keya medical participates in the research and development of balloon and other products, and Shukun technology is involved in the chronic disease management scene. Either way, such enterprises are trying their best to explore new scenarios for AI landing, win more market share and push up the valuation as much as possible. In the process of production line expansion, the problem of "three high" costs faced by Yingtong technology in its annual report has also appeared in the prospectus of the above three companies, especially the sales expenses, which will increase year by year. According to a medical AI company, the medical AI products sold through bidding are attached with a three-year service fee free strategy, and the hospital will not charge for update and maintenance within three years. In other words, when the hospital purchases the products of the medical AI company, it is difficult to have the opportunity to purchase other products within three years. Does the low unit price hurt the innovation of medical AI? When talking about medical AI before 2020, entrepreneurs can attribute all kinds of commercialization problems faced by the industry to the restrictions on registration and access. However, with the breakthrough of this level, the black hole once questioned began to dissipate and replaced by various concrete problems. This is a good thing for medical AI enterprises. The deeper cooperation between doctors and developers eliminates information asymmetry. Like traditional medical devices, enterprises can carry out meticulous design according to the clinical and scientific research needs of doctors, and the application of capital will become efficient in this process. The early rapid breakthrough development of medical AI in various fields has passed, replaced by slow and detailed innovation. In any case, doctors and patients do benefit from such intelligent changes. They have more time for rest or scientific research, and patients get more and faster treatment opportunities because of the acceleration of medical efficiency. Back to the first question. For disruptive innovation projects such as medical AI, it requires high R & D investment to promote technical iteration and clinical trials, as well as high sales investment to establish consensus between doctors and patients. Without a large amount of capital investment, how to break the limitations of products and cognition? From this point of view, the medical image AI of 20 yuan per time under the current domestic medical system may be a little higher than the high benefits brought by the innovation of pharmaceutical equipment