It is too difficult for AI company to make money. Relevant reports show that nearly 90% of AI companies in the world are at a loss, and 10% profitable enterprises are basically technology providers. More than 90% enterprises in China AI industrial chain are also in the loss stage. AI four little dragons all lost money without exception, and one can lose more than the other. For example, Yi Tu Science and Technology 20 17-2020H 1 accumulated losses of 7.268 billion yuan; The accumulated loss of Shi Kuang Science and Technology in the third quarter of 20 17-2020130.6 billion yuan. The IPO of Shangtang Technology is not progressing smoothly. It is reported that the company will submit an application to the Hong Kong Stock Exchange in August. Although it is not clear how much Shangtang Technology has lost, the company, like Congyun Technology, is also at a loss.
Why is it so difficult for AI company to make money?
The main business of Congyun Technology is to provide customers with efficient man-machine collaborative operating systems and industry solutions. The former is to build a man-machine cooperative operating system with the core technology of artificial intelligence independently developed. Through the comprehensive connection of business data, hardware equipment and software applications, we can grasp the core entrance of artificial intelligence ecology and provide customers with information, digital and intelligent artificial intelligence services. The latter is based on the man-machine collaborative operating system, realizing application scenarios such as smart finance and smart travel, and providing industry solutions with artificial intelligence technology as the core for a wider customer base:
During the reporting period, the company provided customers with basic operating systems, application products, core components and technical services based on man-machine collaborative operating systems. Among them, the basic operating system can be directly sold to customers, generally delivered to enterprises with R&D capabilities and third-party software vendors, and put into use after secondary development by customers. The operating systems provided by the company include intelligent cloud platform, view aggregation analysis platform, Rong Zhi cloud platform and integrated biometric system. Based on different functions, it faces different application scenarios such as Internet of Things, government, public security and urban governance, finance and commerce.
It is worth noting that if the customer does not purchase the operating system from Cloud Technology in the early stage, the company will sell the operating system and application products to customers to ensure the effective operation of related application products. Core components are functional modules that can be delivered independently in the basic operating system. They are usually software packages that encapsulate core artificial intelligence functions. They are mainly delivered to customers with strong R&D capabilities and high requirements for software management and control, and integrated into their free systems for customers to use, and basically do not involve customized development. Technical services are mainly services provided by man-machine collaborative operating system except software product sales, including public cloud services, risk control services and intelligent operation and maintenance services.
Since its establishment, the man-machine collaborative operating system and its application products in cloud science have gone through three stages: V 1.0, which initially promotes the precipitation of the kernel of the man-machine collaborative operating system, V2.0, which integrates various business scenarios, and V3.0, which realizes the integration of basic operating systems in four key areas: smart finance, smart governance, smart travel and smart business. Version 4.0 of the company upgraded the Rong Zhi cloud platform in the field of smart governance and the comprehensive biometric system in the field of smart finance, and optimized the operating efficiency and user experience of the system through AI technology:
At the system level, the cloud has developed basic operating systems in different fields from the scientific and technological level, and enabled the application scenarios of AI technology through systems and components. Since 20 14, Defiance Technology has started the research and development of the AI productivity platform Brain++, covering the whole process from data generation, cleaning, preprocessing, marking and storage to algorithm architecture design, experimental link design, training environment construction, training, acceleration, model evaluation and model generation, model distribution, deployment and application. Brain++ integrates deep learning framework MegEngine (Tianyuan), deep learning cloud computing platform MegCompute and data management platform MegData, and integrates computing power, algorithms and data capabilities into an AI infrastructure to realize the whole process and large-scale supply from algorithm production to application;
Compared with the operating system+component mode of cloud-based technology, the Brain++ platform of defiance technology is different in integrating computing power, algorithms and data to realize the whole process of AI. For example, the company's Brain++ commercial version covers the whole process of algorithm production such as data management, model development and computing power scheduling, and can also provide customers with hardware delivery including cluster construction and deployment, so that customers don't have to worry about finding AI hardware suppliers and software and hardware adaptation, which improves the efficiency of AI. Brain++ platform and algorithm constitute the core AI capability of Defiance Technology:
In the business model, the basic operating system, components and application products of Cloud Technology can be sold separately, but the Brain++ platform of Defiance Technology is sold as a solution, which constitutes the business difference between the two companies.
From 20 18 to 2020, the revenue from cloud technology was 484 million yuan, 807 million yuan and 755 million yuan, of which the main business income was 483 million yuan, 780 million yuan and 75 1 10,000 yuan respectively. The decline of main business in 2020 is mainly due to the epidemic situation, which is related to its business model. During the reporting period, the company's other businesses mainly provided outsourced hardware and technology development services to a small number of customers. From 2065438 to 2009, the income from other businesses once reached 27 million yuan, but the proportion was still small.
In the main business, the income of man-machine cooperative operating system is 3,654,380,000 yuan, 654,380,300 yuan and 237 million yuan respectively, accounting for 6.2%, 22.7% and 365,438+0.3% respectively. The revenue of artificial intelligence solutions was 452 million yuan, 597 million yuan and 565.438+0.5 million yuan respectively, accounting for 93.6%, 74.0% and 68.2% respectively:
Defiance Technology is an AI company that focuses on the Internet of Things scenarios, takes the Internet of Things as the carrier of AI technology, and provides empirical solutions for the three core scenarios of consumer Internet of Things, urban Internet of Things and supply chain Internet of Things by building a complete AIoT product system. The company's business is divided into three categories: consumer Internet of Things solutions, urban Internet of Things solutions and supply chain Internet of Things solutions. From 20 17 to 2020Q3, the company's revenue was 304 million yuan, 854 million yuan,1260 million yuan and 7160,000 yuan respectively, of which more than 60% revenue came from urban Internet of Things solutions business:
It is worth noting that although the revenue of third-party software and hardware and intelligent AIoT devices in cloud technology revenue has dropped from 8 1.2% in 20 18 to 50.8% in 2020, it still accounts for half of the country. Known as the industry-leading AI company, in fact, half of its revenue comes from hardware products, which leads to a question: What does AI company make money from?
In terms of gross profit margin, although the gross profit margin of cloud technology's main business increased from 265,438+0.5% to 43.2% during the reporting period, it was still significantly lower than that of Yi Tu Technology and Defiance Technology. The gross profit margin of Eto Technology's main business increased from 57.4% in 2065,438+07 to 765,438+0% H 65,438+0 in 2020, the highest among these companies.
Subdivided into specific products or services, we can see that the gross profit margin of cloud-based technology man-machine collaborative operating system is above 75%, which is at a high level. The gross profit margin of software authorization business of man-machine collaborative operating system is above 80%, mainly because most of the software authorization business involves installation, debugging or customized development, resulting in corresponding expenses. During the reporting period, the gross profit margin of technical services decreased from 99.45% to 40%. Since the financial risk control business involves purchasing data services from abroad, in 2020, the new intelligent operation and maintenance services in the data center need to be entrusted to a third party to provide services, which reduces the gross profit margin.
The gross profit margins of artificial intelligence solution business, which accounts for the largest proportion of cloud technology revenue, are 17.76%, 23.43% and 28. 19% respectively, mainly because this kind of business needs to outsource some supporting software and hardware products or services according to customer demand, and the cost of outsourced materials is high, which squeezes the gross profit margin space. The gross profit margin of the company's artificial intelligence solutions is also significantly lower than that of comparable enterprises. For example, during the reporting period, the gross profit margins of Yi Tu science and technology software and the combination of software and hardware were 64. 1%, 8 1.9%, 87.5%, 86.8% and 1.3%, 32.8%, 54.3% and1.
The gross profit margin of Tian Yun Li Fei and Yunzhisheng solution business is lower than that of Yi Tuo Science and Technology and Defiance Technology, and is equivalent to that of Congyun Technology. For example, the gross profit margin of Tian Yun Li Fei digital city cloud hiding management business and human settlement ambient intelligence upgrade business decreased from 42.27% and 63. 16% to 38.23% and 44.43% respectively, mainly due to the need to purchase hardware in the solution and a certain proportion of installation service fees, especially the increase in the proportion of hardware equipment will drag down the gross profit margin level of related businesses:
Ignoring the gross profit margin level of the technology business, the gross profit margin of the consumer IoT solution business is above 80%, but its revenue ratio has dropped from 45.9% in 20 17 to 18. 1% in 2020Q3, and the gross profit margin of the Internet of Things in the city with the largest revenue ratio has dropped below 30%, thus dragging down the company's gross profit margin level:
Defiance Technology mentioned in the prospectus that the consumer Internet of Things solution is the company's traditional core advantage business, mainly using face recognition technology to provide cloud SaaS and mobile terminal solutions, with software as the main cost and the highest gross profit margin. Urban Internet of Things solution business is mainly smart city and smart building management. With the accumulation of industry experience and the continuous improvement of project design delivery capacity, it is reasonable that the company has the ability to improve gross profit margin. However, Defiance Technology mentioned that due to the increase in the proportion of hardware in the project cost, the gross profit margin decreased:
Combining the business models of Congyun Technology, Tian Yun Li Fei Technology and Defiance Technology, we can see that the company can maintain a high gross profit margin if it only relies on shipping operating systems and other businesses. In the future, as the business continues to mature and the cost is reduced, the company is likely to make a profit. However, at present, AI companies such as Defiance Technology and Yi Tu Technology still focus on solution business, which involves the procurement and installation of some hardware, resulting in a corresponding decline in gross profit margin.
The four little dragons of AI are all losing money without exception, and one can lose money than the other. The cumulative loss of Yuncong Science and Technology during the reporting period is 2.684 billion yuan, which seems to be a lot, but it is still weak in the face of contempt for science and technology and Yi Tu Science and Technology.
The net profit of Yi Tu Science and Technology from 20 17 to 2020H 1 was166 million yuan,1.61billion yuan, 3.642 billion yuan and/kloc respectively. Shi Kuang Science and Technology lost 775 million yuan, 2.80 billion yuan, 6.639 billion yuan and 2.846 billion yuan respectively during the third quarter of 20 17-2020, with a cumulative loss of1306 million yuan. The IPO of Shangtang Technology is not progressing smoothly. It is reported that the company will submit an application to the Hong Kong Stock Exchange in August. Although it is not clear how much Shangtang Technology has lost, the company, like Congyun Technology, is also at a loss.
The main business of CAMBRIAN is the research, development, design and sales of AI chips, which is obviously different from that of Congyun Technology. However, from 20/kloc-0 to 7-2020, the company still accumulated losses of more than 2 billion yuan. The CAMBRIAN losses will be greatly reduced in 2020, but it is still far from turning losses into profits:
Industry leaders suffered serious losses, and small and medium-sized AI companies also suffered heavy losses. For example, in the third quarter of 2007-2020, Li Fei, Tian Yun, which provides application scenario solutions for the operation and management of digital cities and the upgrading of human settlements in ambient intelligence, accumulated net profit loss was 654.38+607 million yuan. In the first three quarters of 2020, the company's revenue was 267 million yuan, and the accumulated revenue during the reporting period was only 680 million yuan, which was not too big a loss.
Why is it so difficult for AI company to make money?
Let's talk about the direct causes of these companies' losses.
From 20 18 to 2020, the gross profit of cloud technology increased from/10 10 yuan to 328 million yuan, and the gross profit rate increased from 2 1.5% to 43.2%, but the expenses during the period soared from 338 million yuan to101.
During the reporting period, the company's sales expenses increased from 65.438+0.29 billion yuan to 274 million yuan, and the sales expense ratio increased from 26.63% to 36.28%, which is a very high level. In addition, the company's R&D investment continued to increase, from 20 18 years1480,000 yuan to 578 million yuan, and the proportion of revenue increased from 30.6 1% to 76.59%, which was enough to make the company lose money:
During the reporting period, equity incentives were implemented from Cloud Technology and corresponding expenses were incurred. However, such expenses will put pressure on the company's profits in the short term, and the impact will be eliminated in time. However, the growth of sales expenses and R&D expenses is continuous. After all, this is closely related to the operation of the company. For example, the largest proportion of cloud technology sales expenses is personnel salary, mainly due to the expansion of the company's business, the increase of sales staff and average salary.
Artificial intelligence is still a technology-intensive enterprise. In order to ensure its continuous competitiveness, the company has invested a lot of money in research and development. At present, the solution iteration speed of artificial intelligence related technologies and application scenarios is relatively fast. Taking the cloud as an example, the product iteration cycle is generally 2-6 months, so the research and development of artificial intelligence industry is a long-lasting and high-input process. For example, the R&D expense ratio from cloud technology will exceed 75% in 2020. The company has eight research projects based on man-machine cooperative operating system, including basic platform, algorithm factory, AI fusion data lake, knowledge computing and man-machine natural interaction.
The same is true of the technology with the biggest loss. During the period from 20 17 to 2020Q3, the company's expenses increased from 402 million yuan to1349 million yuan, which exceeded the company's income in scale. Among them, the sales expense ratio, management fee rate and R&D expense ratio increased from 24. 14%, 33.45% and 66.50% to 4 1.6 respectively.
In addition, in order to improve the enthusiasm of personnel and managers, or create a situation of lack of money, AI company will also implement equity incentives, which will generate huge share-based payment fees and erode the company's profit space. For example, in 20 19, equity incentive was implemented from cloud technology, resulting in share payment fee of13.03 million yuan; 20 19-2020Q3 Tian Yun Li Fei paid 208 million yuan and 7190,000 yuan in share-based payment fees to motivate the core team and ensure the stability of the team.
At present, aside from other things, the huge cost of R&D investment and equity incentives has caused most AI companies to lose money.
Great customer change, high customer concentration and high dependence on a single customer are still the same problems faced by AI companies, which is related to whether the company's operation is sustainable or not and is also one of the obstacles for such companies to go public. Whether it is the science and technology innovation board under the registration system, the Growth Enterprise Market or the main board under the audit system, from the audit committee to the Shanghai Municipal Committee, they are all staring at this issue.
In March this year, the Shanghai Stock Exchange asked the company to inquire about "the sales content, sales revenue and reasons for the changes of the top five customers of different types of products, and whether the changes of the top five customers are in line with industry practices".
In 20 18, the largest customer of Congyun Technology was Beijing Wulian Xinbo Technology Co., Ltd., with revenue accounting for 30.11%; From 20 19 to 2020, Beijing Huizhi Lingyun Data Technology Co., Ltd. is the largest business of the company, with revenue accounting for 30.49% and 10.98%, and the sales amount has changed greatly. In addition, Jiangsu Qiu Yun Information Technology Co., Ltd. and Jiangxi Ma Jun Technology Co., Ltd. became the top five customers of the company shortly after their establishment. The Shanghai Stock Exchange also inquired about the rationality and fairness of the transaction price, as well as the existence of interest transfer or other special interest arrangements.
The situation of cloud technology also exists in other AI companies. For example, on 20 17-2020Q3, the top five customers of Defiance Technology successively experienced four companies, namely, Hangzhou Lianhui Technology Co., Ltd., China Mobile, Beijing Yihualu Information Technology Co., Ltd. and Donghua Software Co., Ltd., with sales amounts ranging from more than 25 million to more than 85 million. Many customers experienced a round of tourism and disappeared in the second year:
From the feedback of cloud technology, AI company is facing the problem of fragmentation, not only the fragmentation of the scene, but also the fragmentation of the order. According to the customer distribution of man-machine collaborative operating system in 2020, the order scale of most customers of cloud technology is below 6.5438+0 million yuan, and the order scale of more than 6.5438+0 million yuan is very low. In the application scenario, the company's products cover smart governance, smart finance, smart travel, smart business and other fields, and the artificial intelligence solution with the largest revenue share also shows similar characteristics:
From the perspective of customer concentration, the sales proportion of the top five customers of Cloud Technology decreased from 62.23% to 27.92%. On the contrary, the sales proportion of the top five customers of Yi Tu Technology increased from 35. 12% to 62.02%, while contempt technology hovered around 20%-30% all the year round.
The uncertainty of customers shows that the reusability of artificial intelligence technology in the client is very low, and the fragmentation of orders shows that the commercialization level of artificial intelligence technology is still at a low level, and it is difficult to achieve large-scale application. If AI company wants to seek development, it must constantly develop new users and extend new application scenarios, which will inevitably increase the company's extra expenses. As mentioned above, the sales expense ratio of Congyun Technology and Defiance Technology is very high, especially the salary of employees accounts for the main proportion, mainly to recruit people to expand business areas and explore customers, and the corresponding expenses are increasing.
The industrial chain of artificial intelligence is divided into three parts: basic layer, technical layer and application layer. Among them, companies represented by Defiance Technology and Tian Yun Li Fei are mostly technology-level companies, which mainly empower smart cities, smart finance and other application scenarios by developing relevant algorithms. At present, the gap between China's AI industry and the United States lies in the following aspects: First, the strength of the basic layer is weak, especially there are too few enterprises with global competitiveness in the fields of chips and sensors, and some enterprises such as Huawei encounter difficulties in operation because of the influence of the entity list:
Cloud has successively laid out the technical layers of computer vision, speech recognition and natural language processing, but more enterprises are in the application layer. Referring to Internet companies, the competition in the application layer will be more intense, and companies with poor technical strength and insufficient comprehensive ability will gradually fall behind. It is also worth noting that, similar to Google, Amazon and Microsoft in the United States, the addition of giants such as Huawei, Tencent and Alibaba has made the competition in the artificial intelligence industry more intense. Huawei, Tencent and other companies have the ability to open up the basic layer, technology layer and application layer, and have unshakable advantages in technology, research and development, customers and markets, which will inevitably bring great pressure to these companies.
Judging from the current industrial development status and the development curve of artificial intelligence technology, it has reached the key node from technology to large-scale application. At present, how to scale has become a pain point in the industry. However, for companies that have been transfused all the way from cloud technology and defiance technology, it is imperative now to let former investors withdraw while enriching blood through listing. After all, they can't wait that many years.
If the invested company goes bankrupt, everything will go down the drain.