The overall business grew steadily, and the core business pioneer AI platform performed well. The company announced business progress for the third quarter ending September 30, 2024. In the first three quarters of 2024, the company achieved total revenue of 3.218 billion yuan, up 26.1% year on year; gross profit of 1.35 billion yuan, up 12.5% year on year; gross profit ratio of 42.0%, which remained stable overall; R&D expenses were 1.373 billion yuan, and R&D expenses were 42.7%. In the first three quarters of 2024, thanks to the steady expansion of the industry's large-scale model business and the accelerated implementation of model products, the company's Prophet AI platform business achieved significant growth. The business revenue reached 2.171 billion yuan, an increase of 50.2% over the previous year, accounting for 67.5% of the company's total revenue, driving a steady increase in the company's performance. “Seer AI Platform 5.0” has achieved end-to-end capacity supply such as computing power layer, platform layer, model layer, and application layer. According to data reported by the International Data Corporation (IDC), an authoritative international research institution, the FOURTH PARADIGM has been ranked first in the market share of machine learning platforms in China for six consecutive years.
Continue to promote diversification of customer groups and steady growth of benchmark users. The company continues to diversify its customer base and continues to cultivate key industries such as transportation, energy and electricity, finance, operators, etc., and the depth of customer service and user stickiness continue to increase. As of September 30, 2024, the company's total number of users was 224 (185 as of 24Q2), 98 benchmark users (86 as of 24Q2), and the average revenue contribution of the benchmark user group was 14.87 million yuan. The company's core product, the “FOURTH PARADIGM Prophet AI Platform”, can support large-scale application of large-scale industry models. At the same time, the company has innovatively built an artificial intelligence product matrix through the “Prophet Inside Model”, and has efficiently built more than 40 artificial intelligence products, including smart mouse solutions, smart conference solutions, etc., to meet the needs of enterprises for applications, platforms and infrastructure during the construction of large industry models, and expand the application potential of AI technology to empower thousands of industries.
Commercialization is progressing steadily, and the trend of large-scale application is gradually emerging. In terms of commercialization, the company collaborated with benchmark users to create a number of exemplary industry models and accelerated the large-scale application of AI technology in thousands of industries. In the field of transportation, the company cooperated with the world's largest port group to build a large model of the port supply chain to provide intelligent capacity output to achieve comprehensive port supply chain service transformation and promote the sustainable application development of the large model in the port and shipping sector and surrounding business formats; in the field of academic translation, the company cooperated with modern publishing institutions to build the first major domestic publishing model, so that the translation level of the large model reached the level of academic publication. The mainstream evaluation standard for machine translation of the general model is BLEU (BLEU: machine translation), mainly by comparing machine translation results with manual reference translation (Evaluating translation quality) Increased by more than 30% to meet high-accuracy translation requirements in academic vertical fields.
Continuously optimizing key technical capabilities, AI Agent has penetrated the company's product capabilities. The company continues to optimize key technical capabilities such as AIAgent and GPU resource pooling. The AI Agent further improves the accuracy of solving complex reasoning and execution, and covers the entire process of large model Agent design, development, debugging, release, operation analysis, and iterative optimization, providing a stable and reliable guarantee for Agent production and application. In terms of product function iteration, the company added AI Agent capabilities to AIGS Builder, AIGS CodeX, and various intelligent hardware solutions, so that the product can obtain more complex reasoning and execution capabilities. Furthermore, AI agents based on reinforcement learning technology form a closed loop of self-learning. The more users use it, the stronger the model reasoning ability, and the more accurate the execution ability. GPU resource pooling capabilities help enterprises achieve platform-based management of hardware clusters, on-demand allocation and rapid scheduling of computing power resources, and help solve the problems of low utilization of heterogeneous computing power resources, large models and large models with high computing power adaptation costs.
The application scenarios of the SHIFT intelligent solution are constantly expanding, and the functions and effects are constantly being iterated. In the first three quarters of 2024, SHIFT's intelligent solutions business revenue was $0.688 billion, accounting for 21.4% of the Group's total revenue. Since the company's business development focus is mainly on the Prophet AI platform, the development and development of SHIFT intelligent solutions will all support the business growth of the Prophet AI platform. Affected by the business development strategy, the business's revenue decreased by 14.1% year over year. The SHIFT intelligent solution is a standardized solution created for different industry business scenarios based on the technology and capabilities of the Prophet AI platform. It can further deepen the company's technology and capabilities into a wider range of industries and scenarios, solve key business problems of traditional enterprises, and promote the digital intelligent transformation of the industry. In the first three quarters of 2024, the company launched a variety of solutions for different application scenarios, and continued to update and iterate the functions and effects of the original solutions to meet the requirements for enterprise characteristics such as accuracy, reliability, and real-time performance in different vertical scenarios. At present, the company's solutions are widely used in important industries such as finance, retail, energy and electricity, healthcare, operators, transportation, and academic translation. Take professional publishing-grade translation services as an example. The solution is based on a hybrid expert model (MoE) to train translation terminology databases for books in hundreds of specialized fields such as philosophy, history, law, and economics to ensure that translation accuracy reaches “publication level”, and solve core problems such as poor accuracy of traditional machine translation, insufficient understanding of the context of the original text, and heavy machine translation marks.
The AIGS service directly targets industry pain points and closely matches core business needs. In the first three quarters of 2024, the FOURTH PARADIGM says that AIGS service business revenue was 0.359 billion yuan, up 17.0% year on year. This business revenue accounted for 11.1% of the Group's total revenue. The business provides efficient development tools and services based on generative AI for the Prophet AI platform business. The company continues to iterate and optimize the functions and application effects of products such as AIGS Builder, an enterprise-level software reconstruction assistant, and AIGS CodeX, an enterprise-level programming assistant, to create an enterprise-level “development assistant” that understands customers, understands business, understands knowledge, understands R&D, and is fast in coding. In order to lower the development threshold, the company's AIGSBuilder product has AI agents, codeless programming, and multi-tool calling capabilities, and can achieve full-stack capabilities such as intelligent software front-end and database development, transformation, and deployment. It replaces the complex pages of traditional software development, and even allows people without R&D or code foundation to develop software only through natural language descriptions, changing the efficiency of enterprise software development from “monthly level” to “daily level” or even “hour level”. The modified enterprise software incorporates a new type of multi-modal “dialog” interaction, which becomes an intelligent assistant for employees to operate software, changing the complicated and inefficient problem of enterprise software menu-style multi-click interaction, and improving the user experience of enterprise software. In order to improve code efficiency, in addition to covering basic functions such as code completion, business Q&A, code review, unit testing, and code interpretation, AIGS CodeX also differentiates the technical characteristics and coding style of enterprise customers. Combined with RAG search enhancements and private domain model fine-tuning, the AIGS CodeX products released by the company have achieved a dual increase in business code generation efficiency and business question and answer effects.
Through advanced development tools, we continue to reduce the difficulty of AI development and promote the rapid implementation of AI in practice. In October, the company released the “Big Model Tool Chain” product, incorporating the LLMOps concept, so that enterprise developers can flexibly and quickly build production-level generative AI applications within a few weeks, and the development cycle has generally been shortened by more than 95%.
The product has a built-in key technology stack required for generative AI applications, including hundreds of open model service supply and model lifecycle management, visual workflow process orchestration, automated analysis and vectorized storage of document knowledge, agent framework and batch evaluation, and support for connecting with business system APIs. At the same time, the product provides continuous closed-loop operation optimization capabilities, so that large-scale model technology can be implemented on a large scale within the enterprise in a full-process, standardized, and low-threshold manner, helping enterprises to intelligently transform and upgrade. In November, the company also released SageSuite, which has the most complete AI application suite in the industry. It can be used right out of the box to help enterprises deploy AI quickly and in one stop. The product has a comprehensive AI application suite, including AIGC, intelligent office, digital employees, AI local search, analysis and decision-making, large model development tools, model warehouses, intelligent agent platforms, DevOps, etc.; through leading GPU pooling (vGPU) and large model optimal inference frameworks, the product efficiently optimizes computing power resources to help customers enhance the utilization value of AI computing power in all aspects and greatly reduce AI application and development costs. We believe that the company continues to reduce the difficulty of AI development by continuously releasing advanced development tools, thereby directly helping potential customers implement AI, and the company's AI commercialization may continue to accelerate in the future.
Profit forecasting and investment advice. We believe that the company maintained a steady growth trend in the first three quarters. The outstanding performance of the core business Seer AI platform and the increase in the number of customers and benchmark customers further confirm the viability of the company's current business model. The SHIFT Intelligent Solution and AIGS service is an important support for the core business. Although it has increased and decreased under the influence of business development strategies, both have developed steadily, become an important complement to the Prophet AI platform, and closely cooperate with the development of the company's core business. In addition, the company also attaches great importance to the actual commercialization of AI. On the one hand, the company is building an AI product matrix through the “Prophet Inside Model”. On the other hand, the company is also actively expanding the customer base. In addition, the company also continues to reduce the difficulty of AI development by continuously releasing advanced development tools, thereby directly helping potential customers implement AI, and the company's AI commercialization will continue to accelerate in the future. As a platform-based AI leader, the company has now embarked on a steady and far-reaching path, and the company's long-term growth is worth looking forward to. Combining various factors, we gave the FOURTH PARADIGM 6-7 times PS in 2024, corresponding to the company's reasonable value range of HK$73.08-85.26 (1 HKD = RMB 0.9299; the exchange rate reference date is November 21, 2024), maintaining the “superior to the market” rating.
Risk warning. Risks that AI development falls short of expectations; risks that AI commercialization falls short of expectations, etc.