Financial Association News, November 27 (Reporter Fu Jing) Recently, many manufacturers have launched autonomous intelligent products, and discussions regarding AI Agent and AI applications have resurfaced. Richinfo Technology (300634.SZ) is targeting both AI intelligent computing and applications and has recently gained significant market attention due to its collaboration with Zhipu AI.

Financial Association News reporter investigated Richinfo Technology and learned that the company is focusing on AI applications this year, with AI Agent also viewed as one of its core directions. The AI music generation application, Cai Ling Tu Ge Le, has shown significant growth in consumer users. Meanwhile, the company is aligning with AI development trends to advance its computing power layout, the AI-native cloud computing solution, Rich AICloud, has already been implemented in several large model training, inference, and AI tool acceleration at the level of hundreds of billions of parameters.

"Since our establishment, especially after our listing in 2018, our goal has been quite clear, which is to focus on the digital transformation of enterprises. In the past two years, our goal has become even clearer, that is, to achieve significant results in the AI field," said Wang Xin, secretary of the board of directors and financial director of Richinfo Technology, during an interview with a Financial Association reporter.

This year focuses on exploring AI applications.

Wang Xin told Financial Association reporters, "This year we focused on exploring AI applications. Although it is relatively early at the moment, once it erupts, the potential is considerable."

Financial Association reporters learned that the AI music generation application Cai Ling Tu Ge Le is currently fully market-oriented, and Richinfo Technology is vigorously promoting it commercially. The Cai Ling Tu Ge Le product iterates every two weeks according to the promotional rhythm and currently has hundreds of thousands of seed users in the consumer market. The company is also collaborating with multiple clients on the business side, and partnerships with celebrities and offline music events are all part of the current product planning. "Cai Ling Tu Ge Le allows you to generate a customized personal song using just one photo, which is very fun, and users enjoy and recognize it very much," Wang Xin stated.

According to Wang Xin's observation of the industry, very few AI application products can form a commercial closed loop. He emphasized during the interview that Cai Ling Tu Ge Le is relatively rare in the market and can achieve a commercial closed loop, expecting it will have a broad development space. The company previously disclosed that after comprehensively considering the technical feasibility, computing power costs, applicability, and other aspects, it found that the field of image-generated music is still relatively blank, thus focusing on R&D for image-generated music scene applications.

Recently, baidu's founder, Li Yanhong, expressed the view that intelligent agents are the most mainstream form of AI applications. According to journalists from the Financial News Agency, AI agents are also considered one of the important core directions for richinfo technology, which will focus on digital employees in its collaborative office product line this year, centered around email and unified office platforms. Wang Xin told Financial News Agency reporters that digital employees represent AI agents in the office field and that the penetration rate will be very high in the future.

Not long ago, zhizhu AI showcased its autonomous intelligent agent AutoGLM, triggering sustained activity in the concept of AI agents in the secondary market.

Richinfo technology previously responded on its interaction platform that zhizhu is its strategic partner, and both sides have cooperation in the implementation of large model industries and AI application levels. Wang Xin informed reporters that the company is currently also exploring collaborative efforts in AI agents with partners such as telecommunication operators.

In fact, several domestic and foreign manufacturers recently launched autonomous intelligent agent products, such as microsoft integrating autonomous AI assistants into Dynamics 365 and apple officially launching Apple Intelligence.

In Wang Xin's view, the overall development of AI agents will go through three stages: assistants, agents, and interactive networks. In the assistant stage, AI agents primarily respond to human instructions, assisting humans in completing specific tasks; in the agent stage, AI agents can not only receive human instructions and execute tasks but also actively perceive the environment, understand human needs, and autonomously make decisions and take actions based on those needs; in the interactive network stage, multiple different AI agents will form an intelligent network where each agent acts as a node in the network, capable of communicating, collaborating, and sharing resources to collectively complete complex tasks. "Currently, the overall development of AI agents is still in its early stages; it might even be said that the first stage has not yet been reached. Whether it takes a year or two or three to five years to pass through a stage actually depends on the maturity and innovation of technology."

Wang Xin believes that it's too early to talk about the maturity of AI taking over mobile phones at this stage, as numerous issues such as software and hardware compatibility, security, and privacy still exist. However, mere software iteration and updating are far from sufficient; AI hardware powered by AI applications may be what to expect next.

Huaxin Securities, in its research report, predicts that the market size of AI agents in china is expected to grow from 55.4 billion yuan in 2023 to 852 billion yuan in 2028, with an annual compound growth rate as high as 72.7%. "I don't think this forecast is overly optimistic; everyone is just waiting for an explosion of AI applications and needs to explore the commercial models," said Wang Xin.

Aligning with the operational layout of telecommunication operators to develop computing power.

According to a reporter from financial alliance, richinfo technology has laid out its AI strategy in three layers: on the AI solution and application level, it has launched products such as AI email, digital human, 5G new call, and colorful spirit music; in the middle platform layer, the Rich AIbox one-stop AI application development platform assists customers in quickly customizing AI applications; at the bottom layer, it focuses on building the Rich AICloud AI native cloud computing solution to provide the intellectual computing infrastructure foundation for AI business development.

Richinfo technology is a core supplier of mobile cloud services. During the interview, Wang Xin repeatedly mentioned deep cooperation with operators, stating that the next step will also involve exploring collaboration on computing power with operators.

"Operators are the national team representatives in the construction of intelligent computing centers in china and are also one of the block orders in the all market. Developing computing power aligns very well with the business layout trends of operators," Wang Xin stated. "The company has a leading technology team that has continuously deepened and accumulated profound technology and rich experience in the intelligent computing field, which provides a solid guarantee for us to develop our computing power business."

Public information shows that richinfo technology has launched the AI native cloud computing solution Rich AICloud, which can build highly available and high-performance AI computing power clusters at low cost and high efficiency. Currently, AICloud has been implemented in multiple large model trainings, inference, and AI tool acceleration at the trillion parameter level.

"The prices of computing power cards may be relatively transparent, but in reality, in addition to the cards, computing power clusters must provide services externally. Network scheduling is the core. Our networking actually saves more than 25% of costs compared to the currently available solutions in the market," Wang Xin told the financial alliance reporter.

When discussing demand, he stated that there is new computing power demand in core scenarios across fields such as self-driving cars, medical, finance, and manufacturing. However, he also admitted that there are issues within the industry, such as low utilization rates of intelligent computing and difficulties in achieving targeted optimization of computing power demand.

It is worth mentioning that the "2024 Restricted Stock Incentive Plan (Draft)" of richinfo technology has completed the granting process. One of the company's performance assessment indicators is achieving a year-on-year growth of 16% or more in non-net profit after deducting the stock payment expenses related to all stock-based incentive plans and employee stock ownership plans that are being implemented within the validity period from 2024 to 2026.

Wang Xin told the financial alliance reporter that to achieve the above goals, richinfo technology will simultaneously focus on three product lines. "In the past few years, we have upgraded our business and innovated technology in the fields of collaborative office, smart channels, cloud, and big data, achieving high growth."

According to reports from the Financial Association, richinfo technology officially signed a cooperation agreement with Huawei in the first half of this year to work on HarmonyOS. A comprehensive and complete development and integration framework for HarmonyOS native apps has been established, and several HarmonyOS native apps have already been launched, such as a HarmonyOS app for a telecom operator's mobile business hall, a HarmonyOS version of a music car machine app, and a HarmonyOS version of a personal email app, among others. In addition, the mobile version of the HarmonyOS email app, Rich M@il, released by richinfo technology this year, was officially listed on the HarmonyOS NEXT application market in August.