Today, the AI industry is moving from “model competition” to the second half of “land competition,” and data is becoming the key infrastructure for large-scale AI applications in enterprises.
In the first half of 2026, with the concentrated release of enterprise-level AI demand, the business of Xunze Technology (03317) ushered in a new stage of growth. According to the latest financial report, in the first half of the year, the company achieved operating income of 967 million yuan, an increase of 389% over the previous year; net profit to mother was 72.51 million yuan, compared with a loss of about 89.4 million yuan in the same period last year, turning a loss into a profit. At the same time, the company's Token business model has completed commercial verification, accounting for more than 10% of revenue, and has become a new revenue growth engine.
What is more noteworthy is that the company further launched the TokenOS operating system and TokenCloud platform to integrate computing power, data, tokens, models and applications, and upgrade to an AI to B integrated infrastructure service provider.
Looking at business evolution, Xunce Technology is completing a key transition: in the past, it relied on real-time data infrastructure capabilities to serve financial customers, but now this capability is being rapidly replicated in industries such as telecommunications, electricity, energy, high-end manufacturing, medical care, robotics, commercial aerospace, smart cars, and low-altitude economy. At the same time, the company is further extending “data processing capabilities” to “data supply, model calling, and AI application implementation” through TokenOS and TokenCloud.
This means that the company's future growth logic is no longer limited to traditional project revenue, but is beginning to evolve towards more frequent and more continuous token calls and platform-based service revenue.
Behind the four-fold increase in revenue: Tokenization opens a second growth curve
The Zhitong Finance App learned that according to the latest interim results data, Xunce Technology achieved a leapfrog performance breakthrough in the first half of 2026, with revenue reaching 967 million yuan, a sharp increase of 389% over the same period last year. This increase not only hit a record high for the company during the same period, but also marked that it has officially entered the fast track of “exponential growth”.
More importantly, the company successfully turned losses into profits, achieved net profit of 72.51 million yuan, and completely opened up the dark line of AI profit. The underlying driving force stemmed from the acceleration of enterprise-level AI real-time data infrastructure deployment and the continuous rise in the penetration rate of diversified industries.
As the company successfully replicates millisecond response and 100% accurate processing capabilities from financial asset management to 11 high-value industries such as telecommunications, electricity, energy, high-end manufacturing, biomedicine, robot training platforms, commercial aerospace, urban operations, low-altitude economy, and smart cars, the company's diversified (non-financial) industry's revenue share has also increased to 87.6%, fully confirming the maturity of its cross-industry replication capabilities.
At the business model level, the commercialization of the Token business has become a new engine for the company's revenue growth. The Zhitong Finance App learned that Xunce Technology pioneered the definition of a “scene type token” to refine private, heterogeneous, and non-standard raw data of enterprises into high-value data fuel.
This kind of scenario token has extremely high information density and business value. The price of each million scenario tokens is between $10 and $100, which is far higher than the “generic token” of the general model, and its price trend continues to rise as enterprises increase their dependence on high-quality pre-input data.
Among them, the data verified the strong vitality of this model: in the second quarter of 2026, the company's token model's annual recurring revenue (ARR) surged 410% from quarter to quarter, and in particular, achieved 300% and 320% month-on-month growth in April and May, respectively. This quantitative model upgrades the revenue stream from traditional project delivery to continuous revenue driven by “number of customers, token price, call frequency, and number of modules.”
Even if the number of customers remains stable, it can still drive the company's revenue to continue to rise as AI applications deepen, significantly enhancing the predictability of revenue and room for profit margin expansion.
Furthermore, the improvement in operating efficiency of Xunce Technology is also remarkable. Thanks to the scale effect released by platform-based and modular product architectures, the company has developed more than 400 standardized functional modules, which can be flexibly combined according to the application needs of different industries.
In the first half of the year, the company's average revenue per customer (ARPU) increased by leaps and bounds, from 1.64 million yuan to 5.56 million yuan, a sharp increase of 240% over the previous year, fully reflecting the strong locking effect brought about by the deep integration of products into customer core business processes.
At the same time, the company's customer retention rate has remained above 90% for a long time, and the level of human efficiency remains leading among AI technology companies. With the further optimization of R&D expenses, sales expenses, and administrative expenses, the share of R&D expenses in revenue fell from 84.96% in the same period last year to 41.01%, showing the inherent resilience of the business model and strong self-hematopoietic ability.
Fourth-step transformation of AI infrastructure, TokenOS + TokenCloud build an integrated service barrier
Looking forward to the future, Xunce Technology is constructing a clear “four-step transition to AI data infrastructure” evolution path, which aims to complete the ultimate evolution from data governance to platform ecology.
Among them, in the “1.0 phase”, the company refined data processing capabilities with millisecond response and 100% accuracy through deep cultivation in the financial asset management industry, and completed initial replication across industries; “Phase 2.0” was marked by the release of the world's first TokenOS operating system in the first half of 2026, which realized the “tokenization transition” of enterprise private data and solved the core bottleneck of large models lacking scenario-based vertical data;
Following the “3.0 phase”, the company plans to launch a global token exchange platform in the second half of 2026 to break up circulation barriers across enterprises and industries, and upgrade the business model from “token transfer charges” to “platform-based ecological charges”; eventually, in the “4.0 phase”, the company will focus on training and deploying small enterprise-level privatization models with smaller parameters and lower costs to completely transform AI from general capabilities into enterprise productivity.
Full link product matrix of “computing power-data-token-model-application”
To support this grand vision and strengthen its competitive advantage, Xunce Technology has focused on launching the core platforms for “AI to B” integrated services — TokenOS and TokenCloud. TokenOS and TokenCloud work together to enable enterprises to reduce deployment costs and improve matching efficiency between computing power, data and models. They are committed to becoming a computing power infrastructure partner for enterprise AI transformation, and continue to create commercial value for customers.
According to the Zhitong Finance App, in terms of product collaboration, TokenOS focuses on data capabilities and is responsible for refining heterogeneous raw data into high-quality scenario tokens to solve the “data availability” problem; while TokenCloud, as a one-stop AI infrastructure, focuses on the deep integration of model capabilities and “data+model”, covering underlying computing power scheduling, data tokenization, and the entire life cycle of model inference optimization and enterprise small model refinement, tuning, and deployment.
This full-stack integration capability has built an extremely high technical barrier, and cooperates with internal FDE (front-end deployment engineers) to directly enter the customer site for modular “assembly” delivery, enabling the Token factory to quickly adapt to different industries, significantly improving delivery efficiency and enhancing customer stickiness.
In the future, according to the Zhitong Finance App, at the level of specific strategy implementation, Xunce Technology will implement measures in five major directions:
First, deepening the evolution of business models and promoting continuous high annual recurring revenue growth, so that revenue is deeply tied to customer value creation; second, accelerate cross-industry replication and expand in depth to key national industries such as smart cars, low-altitude economy, and commercial aerospace. For example, the company has collaborated with Botai Auto Network and Saimu Technology to develop physical AI and world models to enter the intelligent connected car circuit.
Third, develop cutting-edge applications, use token exchange platforms and small enterprise models as strategic depth to seize future industrial dominance; fourth, build a strategic cooperation ecosystem, and deeply bind with leading domestic GPU manufacturers (such as Mu Xi, Tianshu Zhixin, Bizao Technology) and leading algorithm companies to build a closed loop system of “computing power-data-token-model-application”.
Fifth, steadily develop overseas business, use TokenOS's compliance capabilities to explore high-value incremental space overseas, and construct a prototype global layout. At present, the company has completed the strategic card position of the “North and South Poles” of the national data element infrastructure. In-depth cooperation with Shenzhen Digital Institute and North Digital Institute will lay a solid foundation for data compliance circulation.
Furthermore, the company completed about HK$2.35 billion in placement and convertible bond financing at the beginning of July this year. These funds will significantly improve the company's cash flow situation and provide sufficient ammunition for computing power deployment, new industry expansion and global layout in the future.
Taken together, Xunce Technology is transforming from a traditional software company to an enterprise-level AI infrastructure platform with its millisecond real-time processing, deep industry know-how accumulation, and extremely high customer stickiness. Driven by the dual drive of performance implementation and valuation repair, it is leading the data infrastructure circuit in the AI native era with an unprecedented attitude.