Institutional shareholders cast a “vote of confidence”! Xunce Technology (03317)'s 12 billion AI computing power bill was unanimously passed, driven by high-quality orders, and was favored
The Zhitong Finance App learned that on September 28, Xunce Technology (03317) announced that the company held its first extraordinary shareholders' meeting in 2026, and the three special resolutions were passed unanimously, with an approval ratio of over 99%. This includes investing no more than 12 billion yuan to build an AI reasoning and calculation center based on computing power and data, applying for syndicated loans of up to 10 billion yuan and providing guarantees accordingly, and proposals to amend the company's articles of association to expand the scope of business.
The three bills run through the main line of “computing power.” The bill passed successfully, marking that this company, known by the market as the “first token share”, has officially incorporated the AI computing power business into its strategic layout.
According to the market, this is not only a shareholders' meeting vote, but also a collective choice made by institutional shareholders using their right to vote. Institutional shareholders and long-term shareholders of Xunce Technology used their votes to cast a vote of confidence for a computing power layout based on high-quality order construction and 100% resource utilization. This vote not only recognizes the company's AI computing power path, but also lays a solid foundation for further deepening the main business and deepening the closed loop of the AI to B link.
Driven by customer demand, providing “software and hardware integration” services
According to a previous circular, the total investment of Xunce Technology's AI reasoning and calculation center project is no more than 12 billion yuan, of which about 90% is for fixed asset investment. The construction period is 5 years, and it is progressing in stages. In order to ensure the smooth progress of the project, Xunce Technology plans to apply for a syndicated loan of up to 10 billion yuan from financial institutions through its wholly-owned subsidiary Xunshen Technology, for a period of 5 years. The interest rate will be agreed between the syndicate and the borrower, and guarantees will be provided. This project focuses on computing power infrastructure, building stable and efficient computing power clusters, providing customers with computing power services, and obtaining continuous benefits.
Why did a company that tokenize data go down and build a computing power center? According to institutional investors, the answer lies in real orders on the demand side: the project is based on comprehensive considerations of the latest market developments, customer demand trends, and its own business development. It can accurately match the AI industry's immediate computing power needs, effectively solve the problem of mismatch between industry computing power supply and demand, and improve the regional computing power infrastructure system.
Currently, Xunze Technology has served 11 major industries, including finance, telecommunications, electricity, energy, and high-end manufacturing. The customer retention rate has remained stable at over 90%. The ARPU value jumped to 5.56 million yuan in the first half of this year. This also means that Xunce naturally holds a “high quality customer list” in its hands. With the deepening application of enterprise AI, more and more existing customers are demanding privatization, compliance, and billable computing power, hoping to achieve “integrated software and hardware” delivery.
Everything about Xunce Technology revolves around customer needs. This layout logic agency also recognizes that if the company does not use TokenCloud, customers will be forced to use inference cloud services provided by other companies in the future, which will have an impact on TokenOS's stickiness and service. Just as Alibaba doesn't do Alibaba Cloud, Taobao and Tmall merchants will migrate, which will have an impact on the e-commerce ecosystem.
Based on this understanding, TokenCloud is probably first and foremost a defensive product for Xunce, which is necessary to protect the basic market of TokenOS customers; on this basis, it also plays an offensive role — through an “integrated software and hardware” delivery method, customers can purchase TokenOS software capabilities while also purchasing TokenCloud's computational power training services, thereby greatly improving customer stickiness and bargaining power of the original product.
Holding high-quality orders, the “business experience” behind the reverse batch construction model
The investment scale of 12 billion yuan attracted market attention, but the computing power infrastructure logic of Xunce Technology is completely different from the industry's common practice of “build first, then rent”.
The quick approach model is to build entirely according to high-quality orders — reverse batch construction with high-quality long-term leases. At the time of actual implementation, every construction project is guaranteed to be “only done if you make money”, and there is no prior investment. This model enables the TokenCloud server to be seamlessly connected to the grid once deployed, and the utilization rate is close to 100%, which basically eliminates idle computing power resources for rent. In a window period where supply and demand for computing power are tight, this kind of accurate matching of “production based on sales” directly translates into shorter recovery cycles, more stable cash flow, and lower marginal costs.
Xunce's computing power orders naturally have the characteristics of “high cost and long cycle time”, which is superior to short-term flexible pricing commonly found in the ordinary computing power rental market. It solves the two biggest headaches for pure hashrate players: weak income certainty and high renewal risk.
Holding high-quality orders is also one of the core reasons banks grant credit to Xuntse. From the perspective of commercial bank credit granting business, the promotion of the 10 billion credit line itself is a reflection of Xunce Technology's financing ability recognized by the banking system, reflecting the financial institutions' full recognition of the company's business model, underlying asset quality, and future cash flow. At the same time, Xunce can flexibly withdraw funds within the next five years according to actual progress to avoid idle funds.
Forward circulation flywheel: TokenOS and TokenCloud collaborate to achieve “1+1>2”
The strategic value of a computing power center also lies in its synergy with its main business. In Xunce's product system, TokenOS focuses on data refining, transforming enterprise private data into high-density scenario tokens; TokenCloud coordinates computing power scheduling, model inference optimization, and enterprise small model refining and tuning to provide hardware support for token generation. The two work together to form a closed loop of “computing power-data-token-model-application”.
This closed loop is driving a positive cycle flywheel: the deployment of TokenCloud brings increased TokenOS revenue and improved gross profit, while TokenOS optimization in turn brings better customers and orders to TokenCloud. The company's comprehensive gross margin remained high at 60% in the first half of 2026. As the share of the “software and hardware integration” order structure of the computing power business increases, the overall gross margin is expected to increase further. Referring to the performance of overseas pure computing power platforms such as Nebius, the company's gross margin is expected to remain at a high level for a long time. Combined with the low-interest leverage advantage of syndicated loans, the pace of net profit release will further accelerate.
On the computing power supply side, Xunce has signed strategic cooperation agreements with three leading domestic GPU companies, Mu Xi, Tianshu Zhixin, and Bijiao, to jointly develop training and inference chips and platform ecosystems for vertical industries and enterprise AI. This deep binding to the domestic computing power ecosystem not only guarantees the security and stability of the supply chain, but also enables Xunce to occupy a key position in the domestic substitution wave of “computing power+data” integration.
Looking at industry trends, in 2026, China's computing power demand is shifting from “cloud training” to “training+inference” two-wheel drive, and the share of AI inference computing power continues to rise. Some agencies predict that the global AI infrastructure spending will reach 3 trillion to 4 trillion US dollars per year by 2030. Xunce Technology is focusing on the growth track of card slots.
The three bills of the current shareholders' meeting were successfully passed, clearing the barriers to the implementation of Xunce Technology's AI computing power business. When customer demand becomes a strategic driving force, when every investment has an order endorsement, and when software and hardware form a flywheel effect where they coexist, Xunce Technology's computing power layout is not just a capital expenditure decision, but a systematic upgrade around the closed-loop capability of the entire AI to B link.