Fourth Paradigm Highlights Profitability Prospects in Earnings Report, Positions Itself as the Salesforce of the AI Era

In every era of the enterprise software sector, giant companies are born: Oracle and SAP in the information age, Salesforce in the internet age, and Palantir in the big data age. Now, with the boom in large models, artificial intelligence is sweeping in once again, and what Fourth Paradigm aims to do is become the new representative of the current era.

On March 20, the AI company Fourth Paradigm (06682.HK) announced its first annual results since listing. The company's revenue continued to grow steadily, reaching RMB 4.20 billion in 2023, up 36.4% year on year. Since 2019, the company's annual revenue compound growth rate has reached 73.9%. Gross profit was RMB 1.98 billion, up 33.2% year on year, with a gross margin of 47.1%.

On the earnings call, Dai Wenyuan, chairman of the board and CEO of Fourth Paradigm, said that Fourth Paradigm positions itself as an enterprise software company transformed by artificial intelligence, with the goal of empowering every industry with AI. An enterprise software company is precisely positioned to bring each generation of new technology to enterprises and empower them.

At present, the expansion of the company's negative operating cash flow is far smaller than the expansion of its revenue scale. Overall cash flow has improved, and the profit outlook is clear.

Three major segments support financial revenue

A major highlight of this earnings report is that it breaks the long-standing problem that artificial intelligence is difficult to commercialize and monetize, highlighting the profit outlook. The earnings report shows that Fourth Paradigm invested more than RMB 1.7 billion in R&D in 2023, with an R&D expense ratio of more than 40%. Adjusted net loss has narrowed quarter on quarter continuously since 2021; in 2023, the loss was reduced by RMB 88.88 million, a significant year-on-year narrowing of 17.6%. The adjusted net loss margin was only 9.9%.

On the one hand, Fourth Paradigm, founded in 2014, has accumulated years of customer resources and industry experience. To date, Fourth Paradigm serves important industries such as finance, energy and power, telecommunications operators, transportation, manufacturing, healthcare, and retail, using AI to empower the digital transformation of all industries. In 2023, the company provided AI services to 445 enterprise users, including 139 benchmark users (benchmark users: Global Fortune 500 companies or listed companies), up 33.7% year on year, with average revenue per benchmark user reaching RMB 18.38 million.

Behind its industry services are Fourth Paradigm's three major business pillars, specifically the Fourth Paradigm Sage AI Platform (4Paradigm Sage), SHIFT intelligent solutions (4Paradigm SHIFT), and Fourth Paradigm AIGS services (4Paradigm AIGS).

Among them, the Fourth Paradigm Sage AI Platform business has significant scale effects and is the core of all businesses. This business accounts for nearly 60% of revenue and is the underlying platform for developing AI models, supporting the efficient implementation of AI applications at all stages for enterprises. SHIFT intelligent solutions provide intelligent solution products for various industries, further expanding the company's commercial footprint. AIGS services are dedicated to empowering software development through generative AI technology, greatly improving development efficiency.

In terms of financial performance, the earnings report shows that the Fourth Paradigm Sage AI Platform generated RMB 2.51 billion in 2023, up 68% year on year, specifically accounting for 59.6% of the group's total business. The Fourth Paradigm Sage AI Platform achieves large-scale coverage of AI scenarios with low thresholds and low costs.

On the earnings call, Guo Qingyuan, board secretary of Fourth Paradigm, said that in the evolution of AI development and application, the design philosophy of Fourth Paradigm's core products has always insisted on solving pain points at every stage of implementation for users, so the Sage AI Platform has undergone an evolution from 1.0 to 4.0. The Sage 1.0 version was launched in 2015, greatly improving model accuracy through a high-dimensional real-time self-learning framework and solving the problem of poor early model performance. Version 2.0 greatly lowered the threshold for model development through the automated modeling tool HyperCycle, enabling tens of millions of programmers to directly develop AI models with Sage and solving the previously mentioned problems of high cost and relative scarcity of AI scientists. The AI 3.0 version solved the last-mile problem from building models to truly implementing applications. The AI 4.0 version introduced the North Star methodology, that is, truly landing on the core competitiveness issues of enterprises and maximizing the business value of AI application implementation.

In addition, the SHIFT intelligent solutions business made breakthroughs in two aspects in 2023: leading the direction of industry digital transformation development through the SHIFT Summit, and releasing the Fan ecosystem system in October 2023, providing end-to-end solutions with partners oriented toward North Star metrics.

The last business, AIGS services, currently has revenue of more than RMB 400 million, but in the long run, its efficiency gains for the entire business software development are very much worth anticipating. After going through the self-developed industry large model products ShiShuo 1.0 and 2.0 in February 2023 and the 3.0 version released in April 2022, the company also proposed the AIGS strategy, using a generative AI software development platform to boost our overall AI business capabilities.

At present, ShiShuo has been iterated to version 4.0, upgrading capabilities such as model capability, plug-in capability, reasoning capability, Chinese understanding capability, and data query. Related large model businesses have also been widely implemented in industries such as finance, manufacturing, healthcare, retail, and energy. According to estimates, the ShiShuo large model has improved overall software development efficiency by more than 30%.

Interpreting the logic of AI 2.0

Founded in 2014, Fourth Paradigm has now grown for exactly ten years. As an artificial intelligence technology and service provider, it has spanned the information age, the internet age, and the big data age, and has also moved from the so-called AI 1.0 era into 2.0, riding the wave of large model technology and welcoming new development opportunities.

Regarding the question of the so-called AI 2.0 era, Dai Wenyuan told Yicai reporters that he does not actually believe there is a very clear distinction between 1.0 and 2.0; a more rigorous way to put it is that artificial intelligence has moved from behind the scenes to the forefront. "In fact, whether in the entire 1.0 era or the 2.0 era, what we pursue is making models larger and larger and finding more and more patterns in a field. Take large language models as an example. In fact, they also need to discover more and more patterns of the next word. As the model becomes larger and larger, the ability to write articles becomes stronger and stronger," Dai Wenyuan said.

What AI 2.0 brings, by comparison, is greater hope: clearing away the fog-like profit prospects of the 1.0 era and overcoming the commercialization implementation difficulties that have persisted for many years. Dai Wenyuan told reporters that Fourth Paradigm is more concerned with the problems that all industries need to solve, and these problems awaiting solution are, from a fundamental logic, not essentially different from those of large language models or video models. Simply put, if a person reads articles every day, they will gradually be able to write articles; if they look at medical examination reports every day, they will gradually be able to interpret medical examination reports; if a person designs cars every day, they will gradually be able to design cars—that is, the so-called "you become good at whatever you study."

Through what channels are specific solutions delivered? At present, the industry temporarily divides large model solutions into general-purpose large models and vertical industry large models. In a general sense, outsiders roughly believe that the former has a larger parameter scale and stronger capabilities, which also brings higher cost pressure. The latter has a smaller parameter scale and lower computing power costs, offers stronger cost-effectiveness for startups, and is also a more suitable option.

In this regard, Dai Wenyuan told reporters that customers in all industries and the problems that need to be solved in various business scenarios are different, and therefore AI needs to learn different things. For AGI (artificial general intelligence), if the model could be infinitely enlarged and filled with infinitely more data, it might indeed ultimately reach a state of AGI, but in practice, enterprises need to balance their respective capabilities and costs in each vertical field and vertical application. It is like how enterprises hire people: it is impossible to find someone who knows everything for every task, and knowing everything also means higher costs.

Under this logic, Dai Wenyuan said that from a technical perspective, Fourth Paradigm is also pursuing AGI, and the company's vision is to work with customers across industries to hone capabilities in various fields. When this capability reaches a certain threshold, Fourth Paradigm's ShiShuo model will also achieve the effect of AGI. At the same time, for different customers' different scenarios, large models also need to be tailored to a certain extent.

Dai Wenyuan said that whether in AI 1.0 or 2.0, the industry is pursuing artificial general intelligence. As for implementation in specific applications and scenarios, certain tailoring is needed according to the actual application so that the cost-effectiveness of specific applications reaches the best state.

At present, large model companies in the industry place different emphasis on the consumer and business sides, and Fourth Paradigm mainly focuses its efforts in different directions based on its own thinking about market competitiveness. Specifically, first, from a technical perspective, what Fourth Paradigm has been doing in the past is continuously increasing technology investment while controlling the growth rate of technology investment to be lower than the growth rate of revenue. Among this, the year-by-year increase in R&D investment includes two aspects. First, it is oriented toward the future, including large models, automated machine learning, AI Agent robots, and other trends leading AI+. At the same time, it values ecosystem capabilities and promotes comprehensive service for the intelligent transformation of enterprises. It is not that an enterprise can solve all problems simply by possessing leading AI technology; rather, it needs to solve all problems in the process of enterprise AI+. This cannot be completed by one enterprise alone and requires more and more ecosystem partners to participate.

In 2023, Fourth Paradigm launched the Fan Ecosystem system, continuously building more and more ecosystem partners and stronger capabilities in integrating ecosystem resources. Only when enterprises and partners work together can they help enterprises complete the end-to-end AI+ transformation process.

The SHIFT Summit and the Intelligent Transformation Summit, which Fourth Paradigm began holding two years ago, are relevant representatives. Through the company's nearly ten years of accumulated industry influence, Fourth Paradigm attracts leading companies and relevant representative figures in the industry every year to participate in the wave of AI-enabled enterprise transformation. Dai Wenyuan believes that when this direction is defined by enterprises and the entire circle of friends, it can further enhance the barriers of relevant enterprises in market competition.

Clarifying its own positioning

Commercialization has been a challenge that the artificial intelligence field has been addressing in recent years. Correspondingly, enterprises need to clarify their own positioning and the business model that accompanies it.

In Dai Wenyuan's view, the company's underlying logic is to use artificial intelligence technology to help all industries discover more and more patterns and form larger and larger models. To use a currently popular expression, this means industry large models, through which the efficiency of enterprise production and operations is improved. In his view, the opportunity that artificial intelligence technology brings to all industries is the emergence of various business forms transformed by artificial intelligence. For example, an autonomous vehicle company can be regarded as a logistics company in the AI era, and a robotics company can be regarded as a housekeeping company in the AI era. In the future, the AI era will give birth to more and more companies in industries transformed by artificial intelligence.

From the perspective of the industry's development trajectory, the enterprise software sector has gone through several stages in the past. The first stage was the informationization era, when software replaced paper to assist enterprise management; this era gave birth to enterprise software giants such as Oracle and SAP. Then came the internet era. After internet technology emerged, a group of SaaS companies represented by Salesforce was born. Next came the big data era, which produced a group of big data software companies, with Palantir as a typical representative; they believed that the data retained during software operation could help people perform better analysis. In summary, the past upgrades of enterprise software were basically replacements based on new technologies, and whenever a new generation of technology emerged, new software technology giants would appear. The opportunity that artificial intelligence gives the industry is that enterprises need new technologies to empower AI technology.

The enterprise software sector has two characteristics. First, when new technologies emerge, new forces can often gain an advantage in competition with the old forces. For example, in the competition between Salesforce and Oracle, the former actually gained the advantage; in the competition between Palantir and Salesforce, with big data as the representative, the new force also actually gained the advantage. Although the old forces may represent a more reliable brand, the new forces represent a newer technology brand. Therefore, whenever there is technological innovation, opportunities for new giants emerge. What Fourth Paradigm hopes is that after AI technology emerges, it can represent a new technology brand and correspondingly gain new opportunities, thereby achieving relatively large growth.

From the conventional path, the growth path of a new company usually starts with a breakthrough at a single point. For example, to this day, many people still mainly associate Salesforce with CRM, but in fact it is no longer just that. Salesforce achieved a breakthrough through the single point of CRM, grew into a unicorn company, and expanded horizontally into every aspect of enterprise operations. Now basically all the software an enterprise needs can be obtained from Salesforce.

This is also the path Fourth Paradigm hopes to take—achieving a single-point breakthrough through ten years of accumulation. Dai Wenyuan said that since 2018, according to IDC data, Fourth Paradigm has ranked first in China's machine learning platform market share. In the next ten years, what the company needs to do is not only achieve a breakthrough at one point and become a unicorn at that point, but also develop from a point to a surface and gradually grow into a comprehensive software company that empowers intelligent transformation across all industries.

At present, Fourth Paradigm has passed the earliest filing for national large models and has been selected for the first batch of outstanding large-model application cases by the China Academy of Information and Communications Technology, as well as advanced cases of Chinese large models by MIT. In 2023, Fourth Paradigm announced the first batch of more than 40 partners joining the Fan Ecosystem, landing dozens of industry scenarios. The Fan Ecosystem and SHIFT continue to expand the company's business market coverage, thereby achieving the goal of expanding its commercial territory.

Among the specific industries where it has landed, Dai Wenyuan said that the financial industry has always been one of Fourth Paradigm's largest industries. Over the past year, based on large-model technology, the company has helped itself and its customers further improve their business levels in risk control, marketing, and compliance. Compared with the past, the company is steadily increasing the value it contributes to the financial industry.

The positioning of enterprise services needs to be supported by a clear business model and profit prospects. Financial reports show that Fourth Paradigm's cash and cash equivalents as of 2020, 2021, 2022, and 2023H1 were 1.052 billion yuan, 1.293 billion yuan, 1.327 billion yuan, and 1.563 billion yuan, respectively. It continues to focus on business investment and has been recognized by investors.

Specifically, a research report by China Securities analyzes that, in terms of operating cash flow, the company adopted an expansionary strategy in the short term to lay the foundation for subsequent development. During the period, R&D expenses, sales and marketing expenses, and general and administrative expenses were relatively high, causing short-term operating cash flow to be negative. However, the degree of expansion of negative operating cash flow was far smaller than the degree of expansion of the company's revenue scale, indicating that the company's overall cash flow situation has improved somewhat.

In terms of investing cash flow, the company has invested heavily in R&D and innovation. The improvement in net cash from investing activities was mainly due to the sale of financial assets. During 2020-2023H1, the company gradually increased its R&D investment. Since 2022, the company's total cash from investing activities began to improve, with the marginal contribution mainly coming from the fair value change gains and losses item. In terms of financing cash flow, the company's development strategy has been recognized by capital. During 2020-2023H1, the company's net cash from financing activities was generally positive, with the largest net cash from financing activities in 2021, because at the beginning of that year the company completed a $700 million Series D financing led by Boyu Capital, Primavera Capital, and Hopu Investment.

"At this stage, the pace of the company's loss reduction is a phased phenomenon, and the profit prospect is very clear. At the same time, the customers and business territory we already have have laid a certain foundation for future growth. With the growing demands of various enterprises, the number of customers and average revenue have also shown a trend of rising volume and price." Guo Qingyuan said.