Pulling a 70-degree bow, shooting the dream of Agent: the AI face in BaiRong's 25-year report

The capital market is a human cycle of old bottles holding new wine, with countless investors rushing in search of new lands within the blueprint of “technology,” which aims to enhance human efficiency. Previously it was the internet and new energy, now it is AI. Historical experiences have repeatedly told us that while many technology companies achieve operational success, very few investors can profit from investing in them. This is because the vast majority of investors think about sharing the joy with listed companies during performance surges, but few are willing to endure the hardships during peak research and development investments. This has led to generations of investors sighing, “Selling XX company on the eve of the Davis Double-Click.” But the harsh reality is that in a lifetime, one can only experience a maximum of three K-wave cycles. The current Agent track is also on the eve of an explosion, experiencing the trends of fortune-seeking and risk-avoidance in the capital market chip game.

Bairong Intelligent (formerly Bairong Cloud 6608.HK) today released its 2025 annual report, which, while showing a 9% increase in MaaS revenue, stable overall revenue, comprehensive expansion of Agent business, and the RaaS (Results as a Service) business model becoming a major narrative, also faces the unavoidable issue of a 70% year-on-year decline in net profit. Starting from the second trading day after Bairong Intelligent released its full-year performance forecast for 2025, the stock price performance also confirmed the true depiction of the global capital market’s choice between AI investment and output. Although those who have seen the future understand today’s AI Agent, just like NetEase in 2000 or Tesla in 2019, no one dares to promise a future that has been seen but not yet realized.

As the CEO of Bairong stated, he fully understands the urgency of investors wanting to see AI investments transform into financial returns, and he also knows that in the early development stage of the AI Agent business, there is a significant amount of “project-based” demand in the B-end market. If Bairong were to take on all these demands, the revenue would show very attractive short-term effects; however, compared to an AI company born out of adversity and warfare, one with sufficient background can accept the RaaS model and “silicon-based employees” with a long-term perspective. Even though investors may see this as living in luxury, the capital market will eventually redistribute wealth to those who are patient in the AI era, as historical cycles continually tread on the laws of human nature, repeating the same plot.

01 Daring to invest in Agents without concern for costs, because MaaS+BaaS has no worries

Many investors, upon seeing Bairong Intelligent’s 2025 performance forecast, worried that the decline in net profit was due to a revenue cliff, but the complete annual report dispelled this doubt: the MaaS business revenue achieved a 9% year-on-year growth (higher than the 5% in 2024); the BaaS business only saw a slight 5% decrease under policy pressure. Therefore, whether using API interfaces to allow commercial organizations to access various AI models from Bairong or using AI technology to help clients facilitate more transaction volume, both AI business models maintain a very stable foundation, meaning the downstream demand for AI is very robust. The business model needs to be connected to a product, and ultimately, B2B AI services must also serve the B2C market.

China’s business society is quite interesting; C-end users prefer human service rather than AI (feeling respected), while B-end users think about using AI to replace repetitive human tasks (improving efficiency). Hence, companies like Bairong aim to use AI products as “bridges” to meet the demands of both B-end and C-end, so that both feel “like humans.” For instance, Bairong’s VoiceAgent allows C-end users to hear what sounds like real human voice calls, while also helping B-end replace human efforts in smart marketing, product recommendations, and customer follow-ups. Various silicon-based employees built on Bairong’s AgentOS can also run workflows like humans, thus enabling the operation of both MaaS and BaaS models. Financial, retail, and internet clients require a large amount of analysis of user profiles and unstructured information daily, thus needing to invoke various “models” from AI companies for analysis and judgment. This is the “Model” within Bairong’s Model as a Service business. Since clients have daily user profiles to analyze, they need to invoke AI models every day, resulting in a 98% core client retention rate. As long as institutions are involved in increasing product transaction volumes daily, the BaaS business is required to provide assistance.

Every addictive demand corresponds to a wound that has not been completely healed, and this applies to B-end institutions as well. They want to use AI but do not want to develop AI themselves, and they cannot manage AI technology well, so their addictive demand has turned into Bairong’s continuous and super-stable revenue. Viewed from another angle, the choice of B-end clients, represented by financial institutions, to continue using Bairong’s AI products precisely shows that Bairong’s AI product strength holds up well under pressure, just like choosing to get drunk depends on whether the drink is good or bad. Because products like VoiceAgent and Bairong’s smart products are built on Bairong’s AgentOS, from the foundation to the product and then to application, Bairong aims to deliver a complete set to clients. So, does this mean that having a sustainable revenue source and AI products that can withstand the highest precision demands from financial institutions allows Bairong to invest 70% of its profits into its “silicon-based employees” and RaaS narrative without worries?

02 Drawing back a 70-degree bow to shoot out the dream of Agents

“Last year, the company already set market expectations that it would significantly enhance R&D and business investments throughout the year, even at the cost of unrestrainedly capturing market share. If we add last year’s investment back, we can see that the company’s mature business still has robust profitability.” This was the original statement from Bairong’s CEO during the earnings conference. Breaking down the details from the company’s annual report, it indeed reflects the unavoidable cost investments for many players in the Agent track: (1) As mentioned above, good AI products need to resonate with human nature; after all, AI products are developed by humans. Compared to data and algorithms, AI engineers are the most important asset. Global AI giants are frantically pursuing AI talent, and Bairong is no exception; (2) By choosing a small parameter but relatively vertical AI model path, it increases the costs of data and algorithms to fully extract model capabilities; (3) The “BaiJi” AI Infra layer in the Results Cloud platform, while deploying model training on cloud vendors, also sees model inference landing on self-built clusters, leading to unavoidable cost increases related to IDC investments. At the same time, the BaiJi AI Infra is open to the developer ecosystem, representing the AI PaaS model.

High investment but delayed output has already become the harshest reality in the crowded AI track for the 2025 capital market. Broadcom, Micron, and Tencent have all experienced this situation, but the capital market is even more stringent towards Bairong, an AI company with stable revenue but investing a massive proportion of 70% of its profits.

The year 2026 is a critical period for this fully drawn 70-degree bow to deliver results. Bairong’s AI strategy needs to transition from a capability-building phase to a value-realization phase, with “mature scenarios scaling profitability + innovative scenarios’ forward-looking layout” being the information gleaned from Bairong’s 2025 annual report. Gaining an advantage in the Agentic AI industry hinges on the know-how of over 9,000 clients and more than a decade of exclusive data accumulation; after all, while large models are common, specific vertical experience is rare. In terms of industry layout, the risk clearance in the financial and insurance sectors has already entered its final phase, with silicon-based credit marketing, wealth management, and non-performing asset disposal witnessing counter-cyclical growth (and currently representing blue ocean markets); in non-financial sectors, there’s a contact center scale of 30 million seats; professional services in law, business, finance, and tax, along with consulting for small and medium-sized enterprises going global, are now market gaps. In addition to organic growth, Bairong is also leveraging capital empowerment, partnering with top-tier capital firms like Hillhouse, Sequoia, and GGV to establish AI industry funds, investing in and nurturing AI-native companies. Internal growth combined with external expansion has been a breeding ground for past capital market blue-chip stocks.

As for whether Bairong’s investment of 70% of its profits can lead to future performance growth and whether it can expand the AI Agent products (“silicon-based employees”) across various industries, it’s simple. Investors just need to ask themselves two questions: If an NBA star wears Nike shoes to compete in the playoffs or even the finals, will fans follow and buy AJ? If Tesla knows its FSD can become a subscription project, will Musk still be entangled with the initial application only in commuting scenarios? The first question answers Bairong’s AI product strength, while the second indicates the revenue growth rhythm brought by Bairong’s future silicon-based employees and RaaS model. It’s really that simple, right?

03 Every stock price rising after a deep squat is an attempt to break free from the valuation cage

We all know that conductivity and resistance are inversely proportional, but the true determinant of conductivity is not resistance, but the material itself; similarly, what determines a company’s valuation level in the capital market is not liquidity and risk preference, but the business model itself. In 2025, Bairong is reshaping its business model using RaaS and a silicon-based employee ecosystem, while the capital market is expressing its expectations for Bairong’s valuation reconstruction through repeated upward breakthroughs in K-line boxes. Bairong itself is also investing the 70% of profits that can be preserved into its future business model and valuation.

A platform + two types of silicon-based employees + three pricing models + four flagship silicon-based positions is the product structure that Bairong has prepared for its RaaS business model. This means using a “Results Cloud” platform integrated with computing power and Agent development systems to develop EX and CX types of silicon-based employees, providing clients with three pricing models based on task, position salary systems, and value creation sharing. The operation results from four flagship silicon-based positions serve as references for enterprises.

If this business model can be achieved with 70% of short-term profits, will the capital market reshape its valuation system for Bairong? From the current layout path, Bairong’s long-term goal may be to become China’s Anthropic, stepping onto the core logic of production supply-side leaps that belong to this AI revolution—just as Anthropic surpassed others through enterprise-level Agents, Bairong’s layout of “delivering results rather than tools” fundamentally differs from traditional tool-oriented software. It is evident that Bairong is currently targeting the trillion-level new market for silicon-based labor.

Let’s boldly hypothesize that the era of intelligent agents has already led to an epic convergence of product services and pricing models for 2B tech companies in China and the U.S., representing a rare opportunity for Chinese 2B tech companies to overturn historical disadvantages. Bairong currently holds the largest scale of real landing cases for enterprise-level Agents in China. If it can seize this crucial two-year leap, akin to how ByteDance seized the mobile internet wave, it will establish itself as a leader in the Agent era, and the current R&D investment will fully transform into long-term value returns for shareholders.