While the market continues to debate when the application of large language models will translate into tangible financial results for companies, XUNCE (03317) has provided a resounding answer with a profit alert that far exceeded expectations. According to the latest announcement, XUNCE achieved revenue of RMB 1.283 billion in 2025, a year-on-year increase of 102.95%. The non-GAAP net loss narrowed significantly, with several core metrics surpassing market forecasts. More strikingly, second-half revenue surged by over 448% compared to the first half, demonstrating a typical non-linear growth curve. Behind this profit alert lies not just data growth but a critical signal validating the commercial closed-loop for AI data. It signifies that Chinese AI data infrastructure players, represented by XUNCE, have transitioned from the "proof-of-concept" stage to a new phase of high-quality growth driven by data elements.
The qualitative significance of RMB 1.283 billion is substantial. While "doubling growth" is common in capital markets, the base level drastically alters its meaning. A company growing from RMB 20 million to RMB 40 million, despite a 100% growth rate, possesses minimal risk resilience and industry influence. In contrast, XUNCE's 2025 revenue of RMB 1.283 billion represents a doubling from a high base of approximately RMB 600 million the previous year. The RMB 1 billion revenue mark is a "golden benchmark" for testing the maturity of tech firms. Surpassing this threshold indicates the company has moved beyond "single-point breakthroughs" and successfully built a replicable, scalable commercial flywheel. XUNCE not only crossed this barrier effortlessly but also established a leading industry position with its RMB 1.283 billion revenue, marking the formal transition of its platform capabilities from the "construction phase" to the "harvest phase."
The growth tempo is even more noteworthy. The over 448% sequential revenue growth in the second half of 2025 far exceeds normal seasonal fluctuations, pointing to deeper industrial logic. As large AI models shift from conceptual validation to practical deployment, enterprise-level data demand is experiencing a concentrated release. The latter half of the year coincided with numerous industry clients moving from "pilot phases" to "scale deployment." As a provider of underlying data infrastructure, XUNCE saw an acceleration in order fulfillment at this juncture. Concurrently, the company's non-GAAP net loss narrowed substantially, indicating its growth is not driven by sustained heavy investment. As the revenue scale expands, operational leverage effects have begun to appear, laying the groundwork for future profitability.
XUNCE's explosive growth is not accidental but aligns precisely with the industrial inflection point where large models transition from "parameter competition" to "application implementation." The significant revenue increase is primarily attributed to accelerating data demand fueled by the deployment of large AI models. According to the announcement, centered on its AI Data Agent, XUNCE has continuously enhanced millisecond-level real-time data processing capabilities, constructing a full-chain technical system covering data acquisition, cleansing, standardization, real-time computation, and large model fine-tuning. Leveraging this system, its AI infrastructure and solutions are being rapidly deployed across diverse industry scenarios, deepening collaboration with clients.
As industry focus gradually shifts from the compute layer to the application layer, the market is realizing that while general-purpose large models address "comprehension capability," solving specific business problems requires "data memory." Enterprises are increasingly entering phases of model fine-tuning and RAG application, leading to exponential growth in demand for processing high-quality, real-time, multi-source heterogeneous data. As a core conduit connecting data and models, XUNCE is a direct beneficiary of this trend.
This demand release exhibits distinct structural characteristics. From finance to smart manufacturing, and from smart cities to cutting-edge technology, the application boundaries of data intelligence are rapidly expanding. In frontier fields like autonomous driving training and robotic decision-making, data is no longer a "by-product" of business but a direct means of production. The company is actively extending its AI infrastructure development to emerging areas such as robotic data platforms and commercial aviation. These scenarios impose extremely high demands on the real-time capability, complexity, and reliability of data processing platforms, forming a business moat characterized by high average contract values for XUNCE.
A deeper change stems from a shift in client mindset. In the traditional software era, companies purchased "efficiency tools"; in the AI era, they are buying "core productive capacity." When data is directly linked to business growth, client budget scales and willingness to pay undergo a qualitative change. The doubling of XUNCE's average contract value in 2025 is a direct reflection of this trend.
Merely looking at revenue growth is insufficient to understand XUNCE's uniqueness. Its business model, which differs from traditional software companies, is more noteworthy. Among various metrics, XUNCE's most impressive indicator is its revenue per employee. In 2025, XUNCE's revenue per employee reached a remarkable RMB 2.9 million, whereas the figure for mainstream domestic traditional software companies mostly falls within the RMB 500,000 to 800,000 range, making XUNCE's number more than double. Behind this data lies the complete颠覆 (disruption) of the traditional labor-intensive model by "platformization + AI intelligence."
The dual doubling of average contract value and revenue per employee, coupled with a high compound growth rate sustained over the long term, corroborates the characteristics of this model from different angles: expansion from single modules to multiple modules, and from localized trials to full-line deployment, all based on strong client retention. This continuous accumulation of platform capability is a manifestation of its business model. In the future, as the revenue scale expands, the pace of XUNCE's profit release is expected to far outstrip its revenue growth rate, which is the core logic behind its rapidly narrowing losses and path to profitability.
If the current performance proves the viability of XUNCE's business model, then the upcoming business model upgrade will unlock new potential. According to the company's plan, XUNCE will gradually evolve from its current solutions model in three directions: billing based on the volume of data processing tokens, linking costs to actual client usage; charging per training session and usage volume for model fine-tuning and training scenarios; and engaging in deep partnerships with clients to share in the benefits of their business growth.
With the proliferation of enterprise AI Agents, data querying and analysis have shifted from low-frequency operations to high-frequency, real-time interactions. Token-based billing directly links XUNCE's income to client business activity, providing strong revenue elasticity. For fine-tuning and training vertical industry large models, XUNCE's model of charging based on training consumption accurately captures the core value of "data fuel."
More disruptive is the emergence of the "pay-for-performance and revenue-sharing" model. In critical scenarios like financial trading and supply chain optimization, XUNCE is no longer merely a tool provider but evolves into a "partner" for its clients. By sharing in the excess returns clients gain from optimized data-driven decisions, XUNCE completely removes the ceiling on its revenue. This transformation enables XUNCE to metamorphose from a traditional "software company" into a "data asset operation company." In the enterprise AI era, as clients' willingness to pay for "deterministic decision-making" surges dramatically, XUNCE precisely becomes the provider of that certainty.
In the latter half of the large model competition, an industry consensus is becoming clear: the core of competition is no longer the models themselves, but the AI data infrastructure. Models can be open-sourced, and compute power can be purchased;唯独 (only) high-quality, real-time, structured private data and the capability to process it constitute an enterprise's irreplicable core barrier. At this historical juncture, the market is redefining XUNCE as the "Chinese version of Palantir." The core logic behind Palantir's performance explosion in the AI era is that its platform became the "operating system" for governments and large enterprises to handle complex data. Similarly, XUNCE serves massive data integration, handles尖端 (cutting-edge) complex scenarios, and is experiencing a performance inflection point in the AI era.
Within the Hong Kong stock market and across the entire Chinese capital market, XUNCE is an absolute稀缺 (scarce) asset, being an AI data company that simultaneously possesses "large-scale revenue realization," a "high-efficiency model," and "diversified industry penetration" capabilities. From its deep cultivation in the asset management industry to explosive growth in sectors like energy, telecommunications, and retail, XUNCE's diversified penetration rate is accelerating. This is not only a source of revenue but also strategic proof of its role as the foundational "base" for China's AI data infrastructure.
In summary, the 2025 profit alert is both a critical milestone in XUNCE's development and a practical footnote to the Chinese AI industry's transition from conceptual exploration to value realization. The RMB 1.283 billion revenue scale signifies it has crossed the threshold of scaled development. The construction of a full-chain technical system and the deepening of diverse scenarios have paved the way for its entry into the trillion-yuan data要素 (element) market. The RMB 2.9 million revenue per employee refreshes the market's perception of software companies from an efficiency dimension.
For investors, XUNCE no longer represents a conceptual stock reliant on future expectations, but a value sample with a solid performance foundation, a clear growth path, and global对标 (benchmarks). Its strategic positioning in the AI data infrastructure赛道 (track), the business model leap from project-based to "token-based pricing," and the profit improvement expectations driven by the continuous release of operational leverage collectively form the core support for its investment value. Having surpassed the RMB 10 billion门槛 (threshold), XUNCE is poised to enjoy the long-term红利 (dividends) brought by the deepening of AI applications, leveraging its scarcity and first-mover advantage to become a definitive beneficiary in this wave of AI advancement.
Comments