Shenzhen Xunce Technology Co., Ltd. has secured shareholder approval for an investment ceiling of up to RMB12 billion to build an AI inference and computing center based on computing power and data. The vote gives the Hong Kong-listed company authorization to pursue a large infrastructure project, but it does not mean the full sum has been spent, financed, or drawn.
The decision came at Xunce’s first extraordinary general meeting of 2026, held in Shenzhen on September 28. According to the company’s HKEX circular, the proposed project is designed around computing-power infrastructure, data processing, and related research and development. The filing describes a scope that includes computing clusters, standardized hardware systems, an internet data center, computing-power service facilities, equipment purchases, and an R&D center.
The resolution establishes an approved investment ceiling and project structure, not an operating facility. Earlier coverage of Alibaba’s 10,000-card AI computing cluster showed how compute projects can be framed around hardware deployment. Xunce’s resolution is earlier in the process: it establishes an approved investment ceiling and project structure, not an operating facility.
Shareholders approve the project and its financing framework
The shareholder vote covered three special resolutions, according to the September 28 poll-results notice reproduced by Sina Finance. Attending shareholders represented 252,517,111 shares, or approximately 76.33% of Xunce’s total shares. Each of the three resolutions received more than 99% approval in the results reports checked for this article.
The first resolution authorized group investment of up to RMB12 billion in the AI inference and computing center. The wording matters because the figure is a maximum approved amount. It should not be read as evidence that Xunce has already committed the entire sum, selected a site, installed servers, or fixed a completion date. The checked materials do not establish any of those milestones.
The second resolution authorized financing for Shenzhen Xunshen Technology Co., Ltd., a wholly owned subsidiary designated to obtain the proposed funding. The financing ceiling is RMB10 billion from independent parties, while group guarantees may reach RMB10 billion in aggregate. Those provisions create room for a financing structure, but they do not establish that a loan has closed or that the subsidiary has drawn the full amount.
The third resolution approved amendments to Xunce’s articles of association. The Sina results report said the amended articles became effective on the meeting date. Together, the three resolutions give the company a corporate and governance framework for advancing the project, subject to subsequent execution and financing steps.
The result also clarifies what the vote does not show. The available disclosures contain no independently checked analyst forecast, valuation target, bank endorsement, or evidence of institutional-holder support. Approval above 99% indicates the resolutions passed with a large majority of represented shares, but it is not a basis for describing the vote as unanimous or as proof of investor confidence.
The proposed center is broader than a server installation
Xunce’s filing presents the project as a combination of infrastructure and technical services rather than simply a building filled with accelerators. The listed elements include stable and efficient computing clusters, standardized hardware systems, internet data-center facilities, computing-power services, cloud data processing, high-performance computing services, and research into AI computing enablement.
That breadth is significant for how the RMB12 billion ceiling should be interpreted. The authorization can cover a range of equipment, facilities, and development work, while the filing does not provide an exact GPU count, power capacity, server model, land area, or expected compute capacity. Those omissions prevent a precise assessment of the center’s eventual physical scale.
The company has linked the proposal to TokenCloud, which it describes as an all-in-one platform for AI model training and inference. Xunce’s September 21 company-distributed release said the platform was launched in early September and would be built around customer orders. The same material described a target of near-100% cloud utilization. That is a company operating-model claim, not an independently audited utilization result.
The company also said customer retention exceeded 90% as of the end of June 2026. It reported that average revenue per user rose to RMB5.56 million from RMB1.64 million in the comparable period a year earlier, an increase of approximately 240% year over year. These figures should remain attributed to Xunce’s own release rather than treated as independently verified measures of future project demand.
The proposed platform sits within a wider enterprise-AI infrastructure story. EastFrontier previously examined how Tencent, Alibaba, and Baidu raised cloud rates as demand for AI compute grew. Xunce’s plan points to the supply-side question behind that market: whether additional computing facilities can be funded and operated in line with customer requirements without turning an authorization into an unsupported promise of utilization or returns.
Domestic GPU partnerships remain a stated strategic aim
Xunce’s company release said it has strategic partnerships with domestic GPU vendors MetaX, Biren, and Iluvatar CoreX to jointly develop training and inference chips and platform ecosystems. The statement provides context for the company’s intended technology direction, but the checked sources do not establish that any one of those vendors will supply all project hardware or that specific chip models have been selected.
This distinction is important because the project’s formal resolution is about investment, financing, guarantees, and governance. It does not document a hardware bill of materials. Nor does it identify named customer contracts tied to the proposed center. The company’s description of partnerships and customer demand therefore belongs to the rationale for the project, not to a confirmed construction or procurement schedule.
The financing authorization also creates a practical test for the plan. Xunshen may seek up to RMB10 billion from independent parties, with guarantees from the group capped at RMB10 billion in aggregate. The sources do not say that these facilities have been executed, what their pricing would be, or how much debt the subsidiary may ultimately use. Any assessment of leverage, repayment, cash flow, margins, or payback would go beyond the verified record.
Xunce’s vote nevertheless marks a concrete corporate step in China’s AI-compute build-out. The company now has shareholder authorization for a project with a RMB12 billion upper limit, a defined subsidiary financing route, and a project description spanning hardware, data-center services, and R&D. The next evidence will need to come from execution: financing completion, site and equipment disclosures, construction progress, customer documentation, and independently verifiable operating results.
That sequence will determine whether the resolution becomes an operating computing center or remains an approved plan. For now, the strongest conclusion is narrower and more defensible: Xunce shareholders backed the company’s proposed infrastructure and financing framework by more than 99%, while the capital deployment, facility build, and commercial performance remain future events.