The "new solution" of medical pain points in the post-epidemic era: from popular medical science to AI-assisted medical care

 "Technology for good" is the intended meaning of technology to help medical care

The "black swan" of the epidemic is catalyzing the real awakening of public health awareness.

According to the "2020 Public Health Behavior Insight Report" jointly produced by Tencent Medical Code and Ipsos Healthcare, 42% of the public believe that immunity is the most important health indicator; 67% of people said they are more willing to be vaccinated after the epidemic; When there is discomfort, 76% of respondents are willing to choose active intervention, including medical examination, online search for information, medication, etc.

Among them, the value of medical science information is particularly prominent. In addition to traditional offline hospitals, social media, search engines, news websites/apps, health-related apps, etc. have become the most important health science information channels today. The public's acquisition of authoritative medical knowledge is more frequent and active, and the demand for content is developing in terms of convenience and practicality.

The importance of medical science popularization is self-evident. Firstly, the inequity of medical information is a hidden danger in the doctor-patient relationship. Secondly, medical science popularization, especially online medical science popularization, will be a high-quality traffic portal for users to conduct medical treatment, medication and other activities, such as the United States. The medical health information platform WebMD is a typical example of this development path.

WebMD has formed a model of extending the time and space of medical services with the help of professional medical science popularization platforms on the Internet, from hospitals to families, communities and even the whole society. It is worth noting that Tencent Medical Classic has reached an exclusive strategic cooperation with WebMD, and has obtained a global exclusive Chinese license for more than 60,000 medical content.

The 3.0 Era of Internet Medical Science Popularization

Under the epidemic situation, Internet medical science is undergoing an unprecedented iterative upgrade. "Internet medical science popularization has gone from the 1.0 era of'bad money driving out good money' to the'deep water' 2.0 era, and now it has ushered in the 3.0 era-reshaping the connection of popular science and reconstructing popular science through socialized and scene-based communication methods "People-content-scene", knowledge production, dissemination, and forms are more open and diversified.” Huang Lei, vice president of Tencent Medical, explained at the 2020 Tencent Medical ME Conference on November 8.

The first is socialization, which has brought about changes in popular science carriers, popular science methods, and popular science producers. With the development of the Internet and social media, Internet science should be based on taking into account the authority of medicine, making fragmented knowledge like a puzzle, that is, there are partial and overall clues and connections.

During the epidemic, statistics from Tencent's major content platforms showed that more than 5,000 doctors set up online live broadcasts during the epidemic, with a total duration of more than 7,000 hours. At the same time, the patient community in the field of vertical diseases has also become a dense area for the spread of medical science. Knowledge is no longer just a one-way flow from doctors to the public, users can also reversely influence and participate in the production of popular science content.

The second is scenarioization. Through cloud connection, medical science knowledge can be extended to various online and offline application scenarios, and even hardware terminals. At the same time, the advancement of artificial intelligence and algorithms has enabled personalized screening and recommendation of massive content based on user needs, enabling accurate knowledge access in different scenarios.

The breaking of the information wall and the full coverage of popular science information enable users to obtain the information they want, use, and use anytime and anywhere. At the same time, the popular science platform can be deeply integrated with related services of different platforms through open content sharing.

Huang Lei said: "The medical science popularization scene will be more convenient, personalized, and service-oriented. The popularization of science will not only exist in paper books and APPs, but also in news clients, weather forecasts and other scenes. Even through the cloud connection, Expanding to smart terminals such as watches, users can get the information they want and can use anytime and anywhere."

In order to cope with the new changes in "people + content + scene" in medical science popularization, Tencent Medical Classic officially announced the launch of three major plans at the ME conference-the medical science officer plan, the popular science officer plan, and the content open plan. We hope to gather more Professional doctors, institutions, public power. This year's ME Conference takes "pain points" as the theme, and intends to find individual, medical, and social pain points and solutions.

Seamless connection between technology and medical care: AI becomes a life-saving "helper"

With the increasing public concern about health, medical data has increased dramatically. Among the many medical information, medical imaging is the most important source of information for disease screening, diagnosis, and treatment. At present, this field is also the most mature field of domestic AI and medical industry applications.

At present, there are still many pain points in the field of medical imaging: firstly, the imaging resources of primary hospitals and tertiary hospitals in first- and second-tier cities are in a situation of "ice and fire." The number of misdiagnosis is 57 million, mainly in primary medical institutions, while the number in the United States is 12 million. In addition, there are problems such as repeated filming, difficulty in storing past image data, and large gaps in doctor training.

AI has made technological breakthroughs in the field of image recognition. For example, AI can identify patients' medical images and automatically mark key information such as lesions, thereby improving diagnosis efficiency and accuracy. Therefore, AI is also regarded as a "good prescription" to solve the problems in the domestic medical imaging field.

In this epidemic, AI images have played a very good role. First of all, in the face of the diagnosis and treatment of a large number of new coronary pneumonia patients, the demand for CT examinations has surged, but the number of CT equipment in the frontline epidemic area is limited. How to help frontline doctors to improve efficiency has become an urgent problem.

In February this year, emergency CT equipment equipped with the latest "Tencent Miying" AI rushed to Hubei. This set of CT equipment can complete AI pattern recognition within seconds after a patient's CT examination, and can provide doctors with an auxiliary diagnosis reference within 1 minute, which greatly relieves the pressure of insufficient local CT screening capabilities. At the same time, "Tencent Miying" also optimized the image recognition model of new coronary pneumonia. AI will perform accurate image segmentation of the pneumonia lesion area, provide quantitative analysis of the pneumonia lesion, and help doctors accurately assess the severity of the patient's pneumonia and its development process.

"Tencent Miying" is Tencent's first AI medical imaging product, which aggregates the capabilities of Tencent's internal artificial intelligence teams including Tencent AI Lab, Youtu Lab, and Architecture Platform Department, which can assist doctors in disease screening and diagnosis. According to Tencent's official data, "Tencent Miying" has reached cooperation with more than 100 domestic top three hospitals to promote the research and application of AI in the medical field through the joint construction of artificial intelligence joint medical laboratories.

In addition, predicting the development of patients with new coronary pneumonia has become the key to early triage. To this end, the team of Academician Zhong Nanshan and Tencent AI Lab jointly conducted a study. The research results are: based on the survival model established by artificial intelligence deep learning, the 10 clinical characteristics of COVID-19 patients at the time of admission are analyzed, which can be predicted separately The probability of critical illness in 5 days, 10 days and 30 days helps medical staff to accurately triage patients.

Of course, the application of AI in medical imaging still faces challenges, such as the robustness of algorithms and the quantity, quality, and privacy of data. At present, AI imaging needs to be based on clinical applications, enhance clinical value attributes, and become doctors’ assistants to reduce the risk of missed diagnosis.

In addition to the field of medical imaging, AI is becoming a "helper" for the diagnosis and management of some difficult diseases, such as Parkinson's disease. According to statistics, there are currently as many as 3 million Parkinson's disease patients in China. Experts from the World Health Organization predict that China's Parkinson's disease patients will reach 5 million by 2030. With the aggravation of the population aging problem, the evolution of Parkinson's disease has turned into an important social problem.

In response to the increasingly severe disease development trend, Tencent AI Lab and the team of Professor Wang Jian from Huashan Hospital affiliated to Fudan University used artificial intelligence as the starting point to help analyze the exercise videos of Parkinson's disease patients through AI.

This is the world's first attempt to achieve AI-assisted diagnosis of Parkinson's disease through video motion capture analysis technology. This technology does not require the patient to wear any sensors, just stand in front of the mobile phone camera, follow the instructions to complete a series of actions, and the analysis of the patient's motion video can be completed within 3 minutes.

“This provides a very important material for doctors to diagnose the condition. At the same time, AI can assist doctors in quantitatively assessing the severity of Parkinson’s disease.” Professor Wang said that doctors encourage patients to regularly take pictures of their actions every 3 months as required. , And upload via mobile phone.

Whether it's "Tencent Miying", AI prediction research for COVID-19 severe disease, AI-assisted diagnosis or AI new drug research and development, behind these products are the technological crystallization of Tencent AI Lab, and it is also the display of Tencent's achievements in creating "AI that will save lives."

Technology for Good: Technology Empowers Medical "Background"

The past year was also the first anniversary of Tencent's new mission and vision of "User Vision, Technology for Good". Specifically, the meaning of Tencent's new mission and vision is: everything is based on user value, integrating social responsibilities into products and services; pushing technological innovation and cultural heritage, focusing on upgrading various industries, and promoting the sustainable development of society.

Tencent poses a big problem for itself, because in many cases "technology for good" may mean running counter to interests. As a business giant, when such contradictions occur, the direction of its thinking and actions will be the biggest test. Similar to Tencent’s vision, Google’s initial value of "do not be evil" was widely praised by the outside world, and was later revised to "do the right thing."

Some evaluations say that Tencent's new vision has completed the second definition of the ethical standards of science and technology-from Google's "do not be evil" neutral to "goodwill." And Tencent's various explorations in the medical field demonstrate its intended meaning of "technology for good"-using technology and products to help solve social problems.

Popular medical science products represented by Tencent Medical Dictionary are expected to bridge the information gap. With the continuous enrichment and expansion of content and technological innovation, at the same time, we will fully cooperate with the government, authoritative medical institutions, professionals and other sectors to establish effective connections between doctors and patients. Eliminate the asymmetry of information between doctors and patients, making it easier for patients to seek medical treatment and for doctors to practice medicine more easily. This is also the intended meaning of "technology for good".

Long before this epidemic, Tencent paid close attention to medical workers, and was committed to being a "comrade" instead of a "bystander" fighting alongside them. In 2019, Tencent played the power of "Technology for Good" and joined hands with the Chinese Medical Doctor Association to build an intelligent, personalized, online and offline physician service platform for doctors-"Chinese Physician's Home", helping China 400 Many doctors grow up.

Adhering to "technology for good" also means that companies will increase investment to improve social well-being and do not put business returns first. In November of this year, Tencent once again launched the "Tencent Science Week", which mainly included four events including Tencent Science WE Conference, Tencent Medical ME Conference, Scientific Exploration Award Ceremony, and X-Talk, a popular science event for the public.

Among them, the "Medical ME Conference" invited the world's top medical scientists to focus on the most concerned disease topics such as cancer, depression and AIDS, and showcase cutting-edge medical breakthroughs. Ma Huateng, Chairman and CEO of Tencent, once stated that the original intention of the Tencent Medical ME Conference is "to guide more people to pay attention to life and self-understanding, and to improve public health through authoritative medical science and reduce the suffering of illness."

This year, the ME Conference takes "pain points" as the theme, intends to explore the pain points and solutions of individuals, medicine, and society, reveal the reflection and evolution of life and civilization, and promote the popularization of medical science for all. Academician Zhong Nanshan, director of the National Respiratory Diseases Clinical Medicine Research Center, mentioned in his message to the ME conference, "Mr. Ma Huateng once said that the task of the Internet is to solve the pain points of users. This is very similar to the mission of medicine. We are the most fundamental solution for human beings. The pain point is the pain of life." Academician Zhong Nanshan also said that the future medical research and development practice should also actively invest in cutting-edge science and technology.

The ME Conference is dedicated to creating a stage for the same frequency resonance for technology and medicine. Tang Daosheng, senior executive vice president of Tencent and president of the cloud and smart industry business group, said at the meeting, “We believe that science and technology should be good and combined with medicine to help eliminate the pain points of mankind. Use science and technology to protect the light of life and let everyone’s Health becomes measurable, manageable, and improveable.

In fact, in today's digital society, it is not only necessary for large companies to adhere to the background of "technology for good", but also for everyone to hold on to the bottom line of "goodness". As stated in the "A Thousand Miles Journey·Technology for Good White Paper 2020" published by Tencent Research Institute: The second half of the large-scale social experiment of "Technology for Good" has just begun. Stay out of things.

The loss of 790 million yuan in three and a half years has continued to expand. How can AI unicorn Yun Zhisheng support a valuation of 1.2 billion US dollars?

 The continued expansion of losses contrasts with high valuations and unicorn halo.

As a company that started with AI voice, Yunzhisheng is fortunate to be sought after by the capital market during the AI ​​investment frenzy. Behind the high valuation, it is also trapped in the vicious circle of "the higher the valuation, the more difficult it is to prove self-worth with income". Even in the fierce market competition, constantly changing directions, but in the face of the uncertainty of the artificial intelligence industry and continuous losses, it is still facing a huge test.

On November 3, Yunzhisheng Intelligent Technology Co., Ltd. ("Yunzhisheng") formally submitted a prospectus to the Science and Technology Innovation Board, and plans to be listed on the Science and Technology Innovation Board. 

Yunzhisheng was first established in 2012. Since its establishment, it has completed 8 rounds of financing, with a valuation of 1.2 billion US dollars, and put on the unicorn halo early. Yunzhisheng, which started with voice technology, is no longer a single smart technology supplier, but a comprehensive IoT solution provider, expanding the general AI chip and cloud edge market. 

In 2019, according to the "China Unicorn Report 2019" released by the China Evergrande Research Institute, Yunzhisheng is valued at US$1.2 billion, ranking it among the unicorns. 

In the prospectus recently disclosed by Yunzhisheng, this unicorn with a valuation of up to 1.2 billion US dollars showed its true appearance for the first time. 

Judging from the report data, Yunzhisheng has accumulated a loss of 790 million yuan in the past three and a half years. The risk warning section of the prospectus mentioned that “the company suffered a significant loss during the reporting period and is not expected to achieve profitability in the short term, and the uncovered loss will continue to expand. If the unprofitable state continues, it may trigger delisting conditions.” 

At present, the commercialization of Yunzhisheng has not yet fully completed. The continued expansion of losses is in contrast to its high valuation and unicorn halo.

The hidden worries behind a loss of 790 million yuan in three and a half years

Yunzhisheng was first established in 2012, and it took eight years to become an AI unicorn in the voice field. But there are hidden worries in the process of becoming a unicorn. 

According to the prospectus, from 2017 to 2019, Yunzhisheng's revenue was 61.140 million yuan, 197 million yuan and 219 million yuan, and its corresponding net losses were 173 million yuan, 213 million yuan and 279 million yuan, respectively. In the first half of 2020, it achieved revenue of 84.689 million yuan and a net loss of 106 million yuan.

In three and a half years, the accumulated loss amounted to 790 million yuan. 

According to the prospectus, as of June 30, 2020, Yunzhisheng has undistributed profits of -463 million yuan. "In the next period of time, it is expected that the company's unrecovered losses will continue to expand, and the division of cash will not be possible, which will have a certain degree of adverse impact on shareholders' investment income." The prospectus stated. 

The trouble is that the epidemic has also slowed down Yunzhisheng's business development speed and seriously affected its financial performance. 

According to the prospectus, in the first quarter of 2020, Yunzhisheng achieved main operating income of 23,458,500 yuan, a 46.44% decrease from the same period last year. 

The most direct impact is that Yunzhisheng’s business for hospitals, home appliances, and commercial real estate customers, including Yunzhisheng’s voice medical record system for hospitals, medical record quality control systems and other products, has been delayed; home appliance customers’ demand for smart voice accessories fall; and a wine shop Chi can upgrade requirements of delay and so on. 

The more serious data than continuous loss is the performance of cash flow generated by operating activities.

According to the prospectus, the net cash flow generated by Yunzhisheng’s operating activities has been negative, and the net cash flow outflow from 2017 to 2019 has been increasing, reaching 340 million yuan in 2019, indicating that there is a relatively Risks of excessively high accounts receivable or high inventory levels.

In terms of inventory value, in the first half of 2020 alone, Yunzhisheng’s inventory value increased by 9.26 million compared with last year. Inventory mainly includes raw materials, inventory goods, and products issued, including chips used in the production of Internet of Things voice interactive products, PCBA board microphones and other hardware components, and unsold voice interactive products. 

Yun Zhisheng did not specifically explain the inventory situation in the first half of 2020 in the prospectus. However, it is stated in the prospectus that there are more hardware in the overall solution in 2019, and the acceptance is not completed at the end of the year, resulting in a rapid increase in the amount of unissued goods. 

In the absence of profitability, before the listing, the source of funds for Yunzhisheng's operating business was mainly external financing. Since its establishment, it has completed 8 rounds of financing, setting a record for the highest single round of financing in the voice technology field, and the highest round of financing in D. 

The prospectus shows that as of the first half of 2020, Yunzhisheng has a cash and cash equivalent balance of 205 million yuan and a debt-to-asset ratio of 9.16%, which is only suitable for short-term and daily non-interest-bearing debt. 

Long-term sustained losses and excessive accounts receivable have almost become pain points for AI companies. In view of the industry’s high technical barriers, long R&D cycle and high investment, if Yunzhisheng’s R&D direction and product positioning do not meet market expectations, it is likely to cause huge risks. 

R&D investment equals revenue 

With continuous losses, Yunzhisheng's R&D investment costs are quite high. It’s no surprise that R&D investment is high in the AI ​​industry, but Yunzhisheng’s R&D investment can almost be equal to revenue. 

The prospectus shows that Yunzhisheng’s cumulative R&D investment accounts for 107.4% of cumulative operating income, and the proportion of R&D expenditures is higher than the industry level, and the scale of R&D expenditures has grown rapidly.

In 2017, Yunzhisheng invested nearly 100 million yuan in research and development expenses, accounting for 163.55% of operating income. In 2019, R&D expenditures remained high, with R&D investment reaching 258 million yuan, accounting for 117.78% of operating income. Research and development expenses mainly include salary and labor service fees, and technical service fees. 

In the first half of 2020, the average number of R&D personnel of Yunzhisheng is 325, and the per capita salary of R&D personnel is 174,300 yuan. R&D personnel accounted for 68%. 

However, where did the large investment in research and development go? 

In the prospectus, Yunzhisheng listed the status of projects under development as of June 30, 2020, and the heart of the machine only selected a few projects with a large investment amount, mainly focusing on the "cloud side". 

• The AIOS service open platform is mainly used in the fields of smart travel, smart education, and smart home. It has invested 51,775,100 yuan with 79 people. 

• The research and development of artificial intelligence accelerators and chips for the Internet of Things is mainly used in interactive scenarios and product directions such as smart homes. It has invested 62.764 million yuan and 96 people have been invested. · 

• Vehicle-level language artificial intelligence chips and solutions. Design and develop vehicle-level voice AI chips for the automotive pre-installation market, and verticalize this chip with Yunzhisheng’s artificial intelligence software engine and vehicle-level artificial intelligence solutions. Integration to form a complete cloud-core integrated solution. Mainly used in the field of smart travel, 33,0021 million yuan has been invested, 112 people have been invested, and the budget is 55 million yuan. 

In addition to the solution, it is worth noting that Yunzhisheng has also invested heavily in artificial intelligence chips. 

As early as 2018, Li Xiaohan, vice president of Yunzhisheng IoT Business Unit, told the media that he would invest in chip research and development regardless of cost. At present, the AI ​​chips "Swift" and "Hummingbird" series for IoT scenarios have begun mass production and large-scale sales, and "Snow Leopard" automotive-grade chips are still undergoing stability tests. 

In the AI ​​chip and car-level chip market for IoT scenarios that Yunzhisheng is targeting, competitors are also constantly sneaking in, and the utility needs to be tested. 

At the same time as the listing acceptance is approved, Yunzhisheng intends to raise about 912 million yuan of funds, of which 317 million yuan is planned to be used for the construction of the "artificial intelligence chip research and development platform construction project for edge computing of the Internet of Things". The construction period is three years. , The construction front is longer.

Currently, Yunzhisheng currently focuses on three major businesses, including intelligent voice interactive products, intelligent IoT solutions, and artificial intelligence technology services. It provides cross-hardware platforms, cross-application scenarios, and cloud-core integrated IoT voice interaction products. And for industry-level customers. 

In the field of smart IoT solutions, it mainly provides integrated smart IoT solutions for specific scenarios such as hotels, communities, residences, and hospitals. This part of the business has become the mainstream income. 

When it comes to intelligent voice interactive products, they are mainly delivered in the form of an integrated software and hardware. Yunzhisheng itself does not engage in hardware production, but instead entrusts OEM and ODM suppliers to OEM production. 

Yun Zhisheng mentioned in the risk warning that there is a certain risk of improper product quality control when it involves the production of hardware production links entrusted to OEMs. 

Ambiguous customer 

In the "Partners" column of Yunzhisheng's official website, well-known Internet and server vendors such as China Telecom, Intel, Lenovo, and Qualcomm are listed. Official website shows, cloud-known sound current partners have over 20,000, covering the Internet, mobile Internet, smart appliances, automotive systems, online education, intelligent customer service, smart chips and other fields.

However, among the top five customers disclosed in the prospectus, the partners shown on the official website are not included. Yunzhisheng stated in the prospectus that in the process of changing from a smart single product supplier to a smart IoT solution provider, there are certain changes in major customers and suppliers every year. 

In the first half of 2020, the top five customers of Yunzhisheng are mainly in the real estate development and medical fields. 

In 2019, 5.49% of Yunzhisheng's revenue came from Tianjin Jinnan New City Real Estate Development Co., Ltd. Until the first half of this year, the company's revenue contribution increased to nearly 30%, making it the largest customer. Real estate giant Shimao Group has consistently ranked among the top five customers in 2019 and the first half of 2020, contributing nearly 20 million in revenue. 

In the cooperation with Shimao Group, Yunzhisheng mainly provides AI capabilities, integrating Yunzhisheng’s voice interaction capabilities and smart home solutions. In 2019, Yunzhisheng also established a joint venture with Shimao Group to help real estate realize smart transformation. 

It remains to be seen whether the cooperation between real estate and AI companies and the subsequent implementation of the solution can really be used, and how effective it is. 

According to Almost Human is understood that in the process of seeking real estate intelligent upgrade, there are cases of real estate cooperate with AI algorithms, security companies, extending to the wisdom of the community, the wisdom of the home court of the scene, there are many Baidu , millet and other Internet giants players, competition fierce. 

Yunzhisheng also needs to find its own direction in the fierce market competition, and the commercialization effect of its cooperation with real estate needs to be tested, making its prospects full of uncertainty. 

Observe that Yunzhisheng, as a company that started with AI voice, is fortunate to be sought after by the capital market during the AI ​​investment frenzy. Behind the high valuation, it is also trapped in the vicious circle of "the higher the valuation, the more difficult it is to prove self-worth with income". Even in the fierce market competition, constantly changing directions, but in the face of the uncertainty of the artificial intelligence industry and continuous losses, it is still facing a huge test.

Focusing on the improvement of chip manufacturing yield, "Zhongyiyun" is conducting tens of millions of yuan in Pre-A round of financing

 Provide domestically produced alternative solutions for yield control for chip manufacturers in the semiconductor industry chain.

The "Proposals of the Central Committee of the Communist Party of China on Formulating the Fourteenth Five-Year Plan for National Economic and Social Development and Long-Term Goals for 2035" put forward that the system will deploy new infrastructure and accelerate the fifth generation of mobile communications, industrial Internet, and big data centers. Wait for construction. Artificial intelligence, quantum information, and integrated circuits will become the "troika" of priority development.

The 36Kr interviewed " Zhongyi Cloud " is a company based on industrial interconnection, big data and artificial intelligence, focusing on providing big data platforms and yield control solutions for chip manufacturers (Foundry and IDM) in the semiconductor industry chain. The team has been deeply involved in the industry for 15 years, and has reached in-depth cooperation with customers such as SMIC , Hua Hong Hongli , and has been successfully applied in many mass production lines.

Its ADC (Automatic Defect Classification System), DMS (Defect Management System) and YMS (Yield Rate Management System) products that integrate big data and artificial intelligence are the leading professional engineering software for the semiconductor wafer industry in China, which can provide more in-depth Data mining capabilities, more powerful information analysis capabilities, and smarter problem management and control capabilities have achieved full coverage from the full dimensions of yield improvement (discovery, analysis, and disposal), realizing localized substitution.

Semiconductor wafer manufacturing industry chain 

The BDYAP platform independently developed by "Zhongyiyun" is a yield big data platform based on the Industrial Internet. The industrial App it provides involves wafer defect detection and yield improvement in chip manufacturing, production planning and scheduling, equipment and personnel scheduling , Predictive equipment maintenance, intelligent quality inspection, asset monitoring and operation and maintenance, etc. It can achieve high ROI and multi-dimensional improvement of manufacturing efficiency and quality. The average price of similar software provided by international leading manufacturers is more than US$2 million, while the price of "Zhongyiyun" is less than half. The use of "Zhongyiyun" products can achieve "domestic substitution" with high quality and excellent price.

To solve the yield problem in wafer manufacturing, data analysis based on the Internet of Things is required. When collecting unstructured data from a mass production line for POC, it is necessary to use FDMS (Foundry Domain Expert System) to clean and manage the data and intelligently analyze and match the data. Train models through algorithms. One model requires multiple algorithms. Support, not an algorithm that eats all over the world. For example, there are two dimensions to measure the quality of AI ADC models: accuracy and purity. When the comprehensive index reaches 85%, it is possible to import the entire production line.

In addition, "Zhongyi Cloud" can also solve the problems of data tracking and data interaction in the wafer manufacturing process. For example, its Lot Tracing system can realize the data tracking of Foundry factory lot's full life cycle, which can improve the decision-making efficiency of production, management, and operation. The adopted Hadoop and Flink technical architecture have the advantages of high controllability, low cost, and timeliness. In addition, the SIIP industrial interconnection platform of "Zhongyiyun" can help foundry companies realize a business interaction and data interaction management platform that runs through the entire life cycle of business stages from design, tapeout, manufacturing, outsourcing and processing to after-sales. It is used to solve the problem that most Foundry does not have a unified industrial Internet platform to manage and track business chain breaks and data islands caused by data interaction with external partners.

Currently, "Zhongyiyun" yield control technology has been applied to three scenarios: AI+ADC (automatic defect recognition and classification), AI+Metrology (measurement) and AI+Mechanism Trace (tracing analysis). In AI automatic defect recognition scenarios, it can achieve an accuracy rate of greater than 95%, defeating KLA and AMAT; and the usual report output time in AI measurement scenarios is shortened from six hours to about one hour; in AI retrospective analysis, the Problems that cannot be solved by deposition can also achieve a success rate of more than 50%.

In terms of business, "Zhongyiyun" BD mainly has three channels: one is the most important, that is, the industry reputation after verification by the head manufacturer, and the successful case is a must-have ticket to enter other customer sites; the second is the entrepreneurial team for 15 years The accumulation of deep industry experience and contacts, the third is to establish the company's extensive market and industry awareness by participating in world and national competitions.

In August 2020, the State Council set a goal of achieving 70% chip self-sufficiency rate by 2025, but the current self-sufficiency rate is only about 20%. The entire industry's production capacity has at least three times more room for growth. Therefore, software and hardware products and solutions for big data and yield control also have sufficient growth potential. According to SEMI estimates, by the end of 2020, the overall 8-inch wafer supply capacity in China will reach 1.3 million pieces per month, and the monthly output of 12-inch wafers is also estimated to be 750,000 pieces per month.

In the future, different from "production automation" software, "Zhongyi Cloud" will continue to develop the "engineering automation" software market in Foundry wafer manufacturing and large silicon wafer production, including three sets of carriages of industrial interconnection, big data and AI yield control . Calculated based on the minimum market size of 100 Chinese Foundries and 5 large silicon wafer manufacturers during the "14th Five-Year Plan" period of 2020-2025 (equivalent to 1/20 of the benchmarking international manufacturers), and the minimum market size of each manufacturer’s annual R&D investment of 40 million. The investment amount is 21 billion yuan. In the domestic semiconductor industry, there are usually only 2-3 suppliers in the subdivisions. Therefore, according to the estimated market share of 20%, "Zhongyiyun" is expected to achieve about 4 billion yuan in revenue in 2025.

Li Haijun, the co-founder of "Zhongyiyun", told 36Kr that under the overall situation of self-control and control, the team established a vision to build a AiPC (AI Based Process Control) company with a market value of 100 billion in five years. The practice of the world's leading manufacturers in the semiconductor industry has shown that the advancement of Moore's Law by relying only on the evolution of processes and materials itself has become increasingly slow and difficult, and artificial intelligence has become a battleground for military experts.

"Zhongyiyun" expects to complete the Pre-A round of financing on the eve of New Year's Day in 2021. The funds will be used for industry promotion, research and development based on customer needs, and to strengthen the docking with the state-owned industrial chain. At the same time, the team is recruiting ( yinjiao@allinabc.com ).