On September 9, XUNCE (03317) issued a circular announcing an upcoming extraordinary general meeting to review three special resolutions that have captured market attention.
The company intends to channel up to 12 billion yuan through its wholly-owned subsidiary Xunshen Technology into establishing an AI inference and computing center anchored on computational power and data. Additionally, it seeks to secure a syndicated loan facility of up to 10 billion yuan from financial institutions, carrying a five-year tenure.
Industry insiders suggest that the advancement of this billion-yuan credit facility underscores XUNCE's proven ability to secure bank financing, reflecting institutional confidence in its business model, underlying asset quality, and projected cash flows. The project represents a strategic extension of the company's core operations, driven by market opportunities and robust client demand, which is expected to reinforce its full-chain closed-loop capability spanning "compute power—data—tokens—models—applications." This, in turn, should amplify customer retention and pricing leverage.
Regarding the project's specifics, the planned AI inference and computing center, with total costs capped at 12 billion yuan, allocates approximately 90% of the budget to fixed-asset investments. Construction will phase in over five years, aligning closely with the company's primary business operations. The initiative centers on building a stable, high-efficiency computing cluster to deliver compute-as-a-service to clients, generating recurring revenue streams.
The AI computing hub is tightly aligned with the positioning of the TokenCloud platform, functioning as the hardware and energy backbone within the closed-loop system, directly enabling large-scale industrial output of scenario-specific tokens. Strategically, this foray into AI inference and computing infrastructure is both a deliberate progression from XUNCE's existing operations and a response to broader industry trends, further solidifying its end-to-end AI-to-Business ecosystem.
XUNCE is precisely addressing the surging demand for AI compute capacity, resolving the market's supply-demand imbalance by offering an integrated "hardware-plus-software" solution. Clients purchasing software capabilities like TokenOS can concurrently acquire training and inference compute services through TokenCloud, significantly enhancing product stickiness and negotiation power. This positioning has led the market to view XUNCE as a Chinese counterpart to a combination of Palantir and Nebius/CoreWeave.
From a financial standpoint, the new compute business is set to comprehensively empower existing operations, propelling XUNCE's revenue scale to new heights. Moreover, the low-interest leverage from the syndicated loan is expected to accelerate net profit growth, while gross margins could remain elevated, drawing parallels to overseas pure-play compute platforms like Nebius. Looking further ahead, once the inaugural project is successfully operationalized, replication costs are projected to decrease substantially, thereby consistently boosting profitability over time.