Image source @ Visual China
Wen | light cone intelligence, author | Sun Guangxin
The fourth paradigm, which submitted the prospectus a little earlier than SenseTime, failed to be listed first, and on February 14, the fourth paradigm became invalid because it did not pass the hearing within 6 months. 10 days later, The Fourth Paradigm submitted the prospectus again to continue the journey of Hong Kong stock IPO.
Like many AI companies, the fourth paradigm faces the reality of rapid revenue growth but still losing money. According to the prospectus, in the first three quarters of 2021, the fourth paradigm achieved revenue of 1.345 billion yuan, an increase of 134.30% year-on-year, and achieved an adjusted net loss of 393 million yuan, an increase of 30.1% year-on-year, and the increase in adjusted losses was less than the growth of revenue.
However, the fourth paradigm has experienced a decline in gross margin, and in the first three quarters of 2021, the gross profit margin of the fourth paradigm was 46.3%, down 1 percentage point year-on-year. For the decline in gross margin, the fourth paradigm says it is due to the increased demand for customization in new industries.
Revenue growth is also inseparable from the continuous increase in R & D and marketing investment. During the same period, the three major expenses of the fourth paradigm showed substantial growth, of which sales and marketing expenses were 367 million yuan, an increase of 108.02% year-on-year; general and administrative expenses were 486 million yuan, an increase of 154.21% year-on-year; and research and development expenses were 844 million yuan, an increase of 127.76% year-on-year, an increase of more than 100%.
According to the current trend, the fourth paradigm is still some time away from turning a profit, but it is not a long way off. What is the business model of the fourth paradigm compared to the visual algorithm company? After 6 months of resuming listing, can the fourth paradigm converge with SenseTime in Hong Kong stocks?
The three major costs are higher, and the profit problem is still there
The fourth paradigm's revenue sources are mainly "Prophet Platform and Products", "Application Development and Other Services", more than 50% of the current revenue comes from "Prophet Platform and Products", which mainly includes software license revenue and SageOne (AI computing platform).
The Prophet platform is the backbone of the fourth paradigm solution, helping enterprises build their own AI systems, which simply means that users develop artificial intelligence applications through software licensing and SageOne, with the help of the prophet platform of the fourth paradigm, which is mainly driven by automated machine learning algorithms to automate the processes of machine learning, application, decision-making and evaluation.
When users use the Prophet platform to develop more artificial intelligence applications or increase application scenarios for new application scenarios, and need more computing power, users also need to purchase additional licenses from the fourth paradigm to obtain more computing power. As a result, the fourth paradigm charges SageOne's customers based on the amount of hardware, computing power required, software license fees, and service fees for deployment, operations, and maintenance. In simple terms, it is to charge users through the combination of software licensing + computing power.
Its "app development and other services" also revolve around the Prophet platform. The fourth paradigm uses the Prophet platform to develop AI applications for customers according to the needs of users, and charges according to the manpower consumption of each project.
In simple terms, the fourth paradigm's revenue source is to use the Prophet platform to build artificial intelligence applications, after the commercialization of SageOne in 2019, although it will also develop applications for customers, but customers develop their own applications has become the most important source of income for the fourth paradigm.
The advantage of this is that it can concentrate manpower on serving large projects of large customers, reduce the cost of serving small projects of small customers, and attract more small and medium-sized enterprises to purchase their software licenses and specific products. Reflected in the prospectus, that is, the fourth paradigm of dependence on large customers weakened, in SageOne has not yet been commercialized in 2018, the revenue from the top five customers accounted for 60.1%, in 2019, 2020, 2021 the first three quarters, the fourth paradigm from the top five customers accounted for 40.9%, 17.4%, 13.3%.
However, the labor cost savings were not comparable to cost growth, which caused the gross profit margin of sageOne, the fourth paradigm, to decline. The prospectus shows that although it has improved from the end of 2020, in the first three quarters of 2021, SageOne's gross profit margin was only 25.9%, down 6 percentage points from the same period in 2020. The reason why the overall gross profit margin of the Prophet platform and products of the fourth paradigm did not decline was that the software license revenue with a gross profit margin of 95.3% increased by 154.9% year-on-year to 220 million yuan, and the revenue share rose to 16.3% year-on-year, offsetting the impact of sageOne's declining gross profit.
SageOne (AI computing power platform) is a software and hardware integration product, but the fourth paradigm itself does not have the corresponding hardware capabilities, which requires a large number of procurement, such as its 4Paradigm SageOne Advanced on the server is basically Intel's hardware, which leads to in the hardware, the fourth paradigm of the gross profit space is compressed; the computing power also needs to be solved by other cloud services, which also leads to a significant reduction in the profit margin of its computing power part.
Source/Fourth Paradigm official website
Another business of the fourth paradigm, "application development and other services", also faces the problem of declining gross profits. In the first three quarters of 2021, the gross profit margin of the business was 44.6%, down 2 percentage points year-on-year. Although the fourth paradigm states that the revenue of a single customized project can reach more than 30 million yuan, and the average project income is one million, it has shown a downward trend in recent years. From 2018 to 2020, the average revenue of customized projects was 2.6 million yuan, 2.1 million yuan and 1.4 million yuan, respectively, and increased by 2.2 million yuan in the first three quarters of 2021.
However, the gross profit margin of application development and other services did not increase due to the increase in average revenue, and the gross profit margin of the fourth paradigm was 41.4%, 48.8%, 46.3% and 44.6% in the first three quarters of 2018-2020 and the first three quarters of 2021, respectively. This is because each industry has different needs for AI applications, and the initial customization of AI applications for new industry customers will have a certain impact on gross profit margins.
However, the main reasons for the huge losses are still high expenses, and the fourth paradigm summarizes the main reasons as: "R&D expenses for developing and enhancing solutions and technology stacks", sales and marketing expenses "for increasing brand awareness and expanding user base", and general and administrative expenses for "share-based compensation". From 2018 to 2020, although the absolute value of the three major expenses is rising, the proportion of total revenue is declining; in the first three quarters of 2021, due to the obvious growth of the three major expenses, the proportion of total revenue has increased slightly.
The three major costs of the fourth paradigm are not without the possibility of reduction. According to the prospectus, employee benefits expenses accounted for the majority of sales and marketing expenses, general and administrative expenses, and were the main part of research and development expenses for 2018-2020, with share-based compensation being the main part of employee benefit expenses. In the first three quarters of 2021, share-based compensation grew rapidly to $604 million, up 331.1% year-on-year. However, because share-based compensation is not a loss caused by actual business, but comes from the stock options given to employees, it is a non-cash nature that does not lead to cash outflows, and if successfully listed, this part of the loss can disappear quickly.
Source/Prospectus
Kuaishou, Xiaomi, etc. all had changes in the fair value of redeemable convertible preferred shares of 106.9 billion yuan and 54.1 billion yuan respectively when they were listed, and this part of the loss also did not come from the business itself, but from the loss calculated according to accounting standards. After the listing, this part of the loss was also reduced by the conversion of preferred shares into common stock.
But in 2021, the fourth paradigm's increase in R&D spending is not largely due to higher employee benefit spending. According to the prospectus of the fourth paradigm, 60.8% of the R&D expenditure in the first three quarters of 2021 came from technical service fees, which refer to the outsourcing costs of a number of non-core and less complex R&D projects.
The fourth paradigm states that the non-core is mainly for the development of the prophet platform, such as the development of applications and miscellaneous product modules in its artificial intelligence app store, and the design of the user interface, less complex refers to the kind of work such as product testing. Combined with the expansion of application development and other services into new industries in the first three quarters of 2021, the increase in outsourcing project expenses may be related to this.
Customer retention affects profitability
According to the prospectus, in 2018-2020 and the first three quarters of 2021, the number of users of the fourth paradigm was 38, 79, 156 and 186, respectively, of which the number of benchmark users was 18, 32, 47 and 55, respectively.
The benchmark users referred to in the fourth paradigm are the end users of the Prophet platform of the Fortune Global 500 or public listed companies, and the number of benchmark users in a certain period refers to the number of users of the fourth paradigm that generate revenue directly or indirectly through solution partners during that period. From 2018 to 2020 and the first three quarters of 2021, the average revenue from benchmark users was 3.9 million yuan, 8.3 million yuan, 12.3 million yuan and 12.4 million yuan, respectively.
The fourth paradigm uses the net revenue expansion rate of benchmark users to illustrate the long-term value of business relationships with users, customer retention, and the ability to expand revenue. The so-called net income expansion rate of benchmark users refers to the ratio of the same group of specific users, that is, the ratio of benchmark users' income in a certain year to the previous year. In the first three quarters of 2019, 2020 and 2021, the net income expansion rate of the fourth paradigm benchmark users was 250%, 167%, and 148%, respectively.
However, the fourth paradigm has some pressure on customer retention. According to the prospectus, the Fourth Paradigm Prophet platform and product business is primarily sold through software licensing and SageOne, rather than on a subscription basis. This means that the fourth paradigm cannot automatically renew customers like SaaS companies, and the fourth paradigm is difficult to quantify customer retention through customer retention rates like SaaS enterprises.
And, even if customers are very satisfied with the Prophet Platform of The Fourth Paradigm, the Fourth Paradigm can only currently increase revenue by acquiring new customers or selling additional licenses to existing customers.
In addition, it is not difficult to see from the products of the fourth paradigm that the prophet platform and products with a high proportion of revenue are mainly to allow customers to build their own AI applications, neither directly contacting the user's data, nor providing users with a ready-made AI application, under the premise that the technical conditions of the prophet platform itself can be realized, the use effect depends more on the customer itself.
Therefore, although the fourth paradigm is likely to make users purchase licenses from the fourth paradigm again due to the increasing demand for AI applications in business due to the increasing demand for AI applications in the operation, it is more conducive to the revenue growth of the fourth paradigm to expand the user scale in the short term.
In its prospectus, The Fourth Paradigm states that a portion of the proceeds raised by an IPO may be used to invest in strategic technology enablers, proprietary solution providers focused on industries or sub-sectors, and one of the main reasons is also to expand the size of users. The prospectus shows that about 80% of the revenue of the fourth paradigm comes from end customers who sign up through solution partners.
The fourth paradigm states that some end users of their solutions tend to purchase from system integrators when choosing product vendors or service providers, in order to save them the trouble of direct negotiations with a large number of different product suppliers or service providers. End users engage third-party system integrators who embed fourth-normal form solutions into their products to provide a variety of services to end users. Citing insight consulting, The Fourth Paradigm says it is industry practice to engage with end customers through solution partner customers.
Break through among the giants
Although there are not many AI companies that have successfully landed in the secondary market in China, there are still several companies that have successfully landed. Although the specific areas involved are different, through the comparison of prospectuses and financial reports, it is still possible to see some of the overall situation of the current AI industry.
"Increasing revenue without increasing profits" is the most headache for current AI companies, but in fact, these companies that have been listed or are in the process of IPO are gradually narrowing their losses.
According to SenseTime's prospectus, in 200 and the first half of 2021, SenseTime achieved revenue of RMB3.446 billion and RMB1.652 billion respectively, an increase of 13.9% and 91.8% year-on-year, respectively, and adjusted net losses of RMB708 million and RMB578 million, down 38.7% and 18.3% year-on-year, respectively. However, SenseTime's adjusted net profit has not been able to continue to decline, according to its prospectus, due to increased investment in research and development, SenseTime expects to expand its adjusted net loss for the whole year of 2021.
At present, the gross profit margin of AI companies is also relatively stable, and the gross profit margin of enterprises focusing on software is relatively high, reaching 70%, and the enterprises that integrate software and hardware are less than 50%. In 2020, the gross profit margins of SenseTime and Yitu Technology, which are mainly software products, were 70.6% and 70.99%, respectively, and Megvii Technology and Yuncong Technology, which are integrated with software and hardware, were 33.11% and 43.21%, respectively.
SenseTime also has the heart to step into the hardware field, in the prospectus, SenseTime for the first time disclosed the progress of its self-developed AI computing chip STPU: mass production of 11,000 lessons; Itu Technology has already released its cloud AI chip QuestCore. It can be seen that although the current software-only model is more conducive to achieving high gross margins, hardware can achieve more complete functions, integrated products, and achieve higher revenue figures.
However, compared with the visual AI companies such as SenseTime and Yitu Technology, the current focus of the fourth paradigm to help enterprises improve the ability of AI applications with the platform is still on the software, and at a press conference in June 2021, Zheng Cao, vice president of the fourth paradigm technology, announced the Nott "China Knot" plan to achieve more than 95% of the core technologies open source within 1 year.
The fourth paradigm does this in relation to the segmentation it is in. According to the research report of Guosen Securities, the real opponent of the fourth paradigm of decision-making AI solution providers is not these AI unicorns, but old technology giants with rich business lines such as Baidu, Alibaba, Huawei, and Tencent.
Source/Guosen Securities Research Report
The To B market is segmented and fragmented, giving a new generation of AI companies like The Fourth Paradigm the opportunity to emerge.
Although on the surface, compared with these technology giants, the shortcomings of the young fourth paradigm in the decision-making AI market based on enterprise services are very obvious, that is, there is neither hardware and computing power advantages, nor ecological advantages. However, in terms of results, the fourth paradigm of the segmentation field can still get a relatively good market share, which may stem from the deep cultivation and focus on the subdivision industry.