The AI industry is now shifting from the phase of "competing on models" to the next stage of "competing on implementation," with data emerging as the critical infrastructure for scaled enterprise AI adoption. In the first half of 2026, as enterprise-level AI demand was unleashed in a concentrated wave, XUNCE (03317) entered a new phase of business growth. According to the latest financial report, the company achieved revenue of RMB 967 million in the first half of the year, a year-on-year surge of 389%, while net profit attributable to shareholders reached RMB 72.51 million, turning around from a loss of approximately RMB 89.4 million in the same period last year. Simultaneously, the company's Token business model has completed commercial validation, with revenue contribution surpassing 10% and becoming a new revenue growth engine.

More notably, the company has further launched the TokenOS operating system and the TokenCloud platform, integrating computing power, data, Tokens, models, and applications to upgrade into an integrated AI To B infrastructure service provider. From a business evolution perspective, XUNCE is completing a pivotal leap: previously, it served financial clients through real-time data infrastructure capabilities, and now this capability is rapidly being replicated across industries such as telecommunications, power, energy, high-end manufacturing, healthcare, robotics, commercial aerospace, intelligent vehicles, and the low-altitude economy. At the same time, through TokenOS and TokenCloud, the company is extending its "data processing capability" further into "data supply, model invocation, and AI application deployment." This implies that the company's future growth logic is no longer confined to traditional project-based revenue, but is beginning to evolve toward higher-frequency, more sustainable Token invocation and platform-based service revenue.

Revenue Surges Fourfold: Tokenization Opens a Second Growth Curve

According to the latest interim results, XUNCE achieved a breakthrough performance in the first half of 2026, with revenue reaching RMB 967 million, a substantial 389% increase year-on-year. This growth rate not only set a new record for the company's interim period but also signaled its formal entry into the fast track of "exponential growth." More critically, the company successfully turned profitable, achieving a net profit attributable to shareholders of RMB 72.51 million, fully unlocking the underlying profitability of AI. The fundamental driver stems from the accelerated deployment of enterprise-level AI real-time data infrastructure and the continuous rise in penetration across diversified industries. As the company successfully replicated its millisecond-level response and 100% accurate processing capabilities from financial asset management to eleven high-value industries—including telecommunications, power, energy, high-end manufacturing, biomedical, robotics training platforms, commercial aerospace, urban operations, the low-altitude economy, and intelligent vehicles—the proportion of revenue from diversified (non-financial) industries has also risen to 87.6%, fully validating the maturity of its cross-industry replication capability.

At the business model level, the commercialization of the Token business has emerged as a new revenue growth engine. XUNCE has pioneered the definition of "scenario-specific vertical Tokens," refining enterprises' private, heterogeneous, and non-standard raw data into high-value data fuel. These scenario Tokens possess extremely high information density and business value, with pricing ranging from $10 to $100 per million scenario-specific Tokens, significantly higher than the "general Tokens" of large language models. Moreover, their pricing trajectory continues to rise as enterprises' reliance on high-quality pre-input data increases. The data validates the robust vitality of this model: in the second quarter of 2026, the company's Token model annual recurring revenue (ARR) grew 410% quarter-on-quarter, with particularly strong monthly growth of 300% and 320% in April and May, respectively. This quantitative model upgrades revenue sources from traditional project delivery to recurring revenue driven jointly by "customer count, Token price, invocation frequency, and module count." Even if customer numbers remain stable, the increase in Token invocation driven by deepening AI applications can still propel continuous revenue growth, significantly enhancing revenue predictability and margin expansion potential.

Furthermore, XUNCE's improvement in operational efficiency is equally remarkable. Benefiting from the scale effects unleashed by its platform-based and modular product architecture, the company has developed over 400 standardized functional modules that can be flexibly combined based on different industry application needs. In the first half of the year, the company's average revenue per user (ARPU) achieved a leap, rising from RMB 1.64 million to RMB 5.56 million, a 240% year-on-year increase, fully reflecting the strong lock-in effect of deep product integration into customers' core business processes. Meanwhile, the customer retention rate has remained above 90% for a sustained period, and workforce efficiency maintains a leading position among AI technology companies. With further optimization of the R&D, sales, and administrative expense structures, R&D expenditure as a percentage of revenue declined from 84.96% in the same period last year to 41.01%, demonstrating the inherent resilience of the business model and its strong self-sustaining cash generation capability.

Four-Stage Leap in AI Infrastructure: TokenOS + TokenCloud Build an Integrated Service Barrier

Looking ahead, XUNCE is constructing a clear evolution path of "four-stage leaps in AI data infrastructure," aiming to achieve the ultimate evolution from data governance to a platform ecosystem. In the "Stage 1.0," the company honed its millisecond-level response and 100% accuracy data processing capabilities through deep cultivation in the financial asset management industry, completing initial cross-industry replication. The "Stage 2.0" is marked by the release of the world's first TokenOS operating system in the first half of 2026, achieving the "Tokenization leap" of enterprise private data and solving the core bottleneck of large models lacking scenario-specific vertical data. In the subsequent "Stage 3.0," the company plans to launch a global Token exchange platform in the second half of 2026, breaking down cross-enterprise and cross-industry circulation barriers, upgrading the business model from "Token invocation fees" to "platform-based ecosystem fees." Finally, in the "Stage 4.0," the company will focus on training and deploying smaller-scale, lower-cost enterprise-level private small models, driving AI to transform from general capabilities into exclusive productivity for enterprises.

To support this grand vision and strengthen its competitive advantages, XUNCE has focused on launching the core platforms for integrated "AI to B" services—TokenOS and TokenCloud. TokenOS and TokenCloud work in synergy to reduce deployment costs and improve matching efficiency between computing power, data, and models for enterprises, aiming to become a computing power infrastructure partner for enterprise AI transformation and continuously create commercial value for customers. In terms of product synergy, TokenOS focuses on data capabilities, refining heterogeneous raw data into high-quality scenario Token fuel to solve the "data usability" problem. TokenCloud, as a one-stop AI infrastructure, focuses on model capabilities and the deep integration of "data + models," covering the full lifecycle from underlying computing power scheduling and data Tokenization to model inference optimization, enterprise small model refinement, tuning, and deployment. This full-stack integration capability builds an extremely high technical barrier. Combined with internal FDE (Front-end Deployment Engineers) directly stationed at customer sites for modular "assembly" delivery, the Token Factory can quickly adapt to different industries, significantly improving delivery efficiency and enhancing customer stickiness.

In the future, at the specific strategic execution level, XUNCE will implement initiatives around five key directions. First, deepening business model evolution to drive sustained high growth in annual recurring revenue, tightly binding revenue with customer value creation. Second, accelerating cross-industry replication, expanding into national strategic industries such as intelligent vehicles, the low-altitude economy, and commercial aerospace. For example, the company has collaborated with PATEO and Cemu Technology to co-develop physical AI and world models, entering the intelligent connected vehicle sector. Third, exploring frontier applications, using the Token exchange platform and enterprise small models as strategic depth to seize the high ground of future industries. Fourth, building a strategic cooperation ecosystem, deeply binding with leading domestic GPU manufacturers such as Moore Threads, Iluvatar CoreX, and Biren Technology, as well as leading algorithm companies, to construct a closed-loop system of "computing power-data-Token-model-application." Fifth, steadily expanding overseas business, leveraging TokenOS's compliance capabilities to explore high-value incremental space overseas and build a prototype for global layout. Currently, the company has completed the strategic positioning of the "North-South dual poles" for national data element infrastructure, with deep cooperation with Shenzhen Data Exchange and Beijing International Data Exchange laying a solid foundation for compliant data circulation. Additionally, the company completed approximately HKD 2.35 billion in placement and convertible bond financing in early July this year. These funds will significantly improve the company's cash flow position, providing ample ammunition for future computing power deployment, new industry expansion, and global layout.

In summary, XUNCE, with its millisecond-level real-time processing, deep industry know-how, and extremely high customer stickiness, is transforming from a traditional software company into an enterprise-level AI infrastructure platform. Driven by both performance delivery and valuation recovery, it is leading the data infrastructure track of the AI-native era with an unprecedented posture.