Xunce (03317.HK) Launches TokenCloud: Enabling One-stop End-to-End AI Implementation

On 3 September, Xunce (03317.HK, “Xunce”) unveiled TokenCloud, an all‑in‑one AI model training, inference and computing platform. Positioned as hardware infrastructure that converts data resources into Tokens, the platform builds a four‑layer collaborative ecosystem underpinned by heterogeneous computing devices, powered by mainstream algorithmic models, fed by multi‑source internal and external data, and tailored for vertical industry clients. It unlocks the full value chain from data to Tokens, delivering out‑of‑the‑box AI infrastructure for enterprises.

AI implementation is now shifting from technical feasibility to competition over engineering efficiency, while computing power has entered a new era marked by rising volume and prices. Statistics show China’s daily Token call volume has surged more than 1,000 times within two years, with a shortage exceeding 35% in high‑end intelligent computing capacity. IDC projects the global computing power rental market to top USD 80 billion this year, while China’s market will surpass RMB 2.6 trillion.

Driven by exploding Token consumption, tight supply of high‑end computing resources and rapid expansion of the computing power rental market, there is a strong demand for an integrated platform that seamlessly connects computing resources, data and models. Xunce targets this structural supply gap. TokenOS focuses on data refinement, while TokenCloud centrally orchestrates heterogeneous computing resources, model inference optimization and fine‑tuning of enterprise small models, enabling deep synergy.

Covering the entire enterprise AI implementation lifecycle, TokenCloud features a 5‑capability matrix spanning solution selection, model training & inference, computing resources and security. Its Selection & Matching Center leverages 5‑tier linked configuration and 6‑dimensional dynamic scoring to shift solution selection from experience‑based judgement to data‑driven decision‑making. Model training and distillation condenses capabilities of large models into lightweight alternatives with nearly no loss in accuracy, faster inference and simpler deployment. Computing acceleration prioritizes optimization before capacity expansion to fully tap the potential of existing computing resources. The computing resource management module uses a unified dashboard to oversee on‑premise and cloud resources in a single view, delivering full visibility and flexible scheduling. Tiered domain locking is deployed for data security governance, ensuring 100% containment of highly sensitive data within local secure domains.

For enterprises, TokenCloud cuts computing investment and operating costs substantially via heterogeneous computing optimization and solution selection. Through model inference optimization and refinement, it strikes an optimal balance across accuracy, speed and cost. Its one‑stop services drastically shorten AI deployment cycles. More importantly, TokenCloud transforms enterprises’ years of domain expertise into proprietary data assets and AI capabilities, enabling Tokens to generate tangible business value.

For Xunce, TokenCloud fills a critical gap in its full‑value‑chain loop covering computing power, data, Tokens, models and applications. It marks Xunce’s transition from a digital infrastructure provider to an AI productivity platform player. By systematizing and productizing scenario‑specific capabilities, TokenCloud extends Xunce’s reach from data governance to Token generation and circulation. Riding the industry shift from hardware sales to Token‑as‑a‑service, Xunce is poised to capture strategic advantages amid the Token economy and cement its position as a key gateway for local AI infrastructure.

As more industry clients and scenarios adopt Token services, a virtuous cycle will form across Token generation, circulation and monetization, where high‑quality Tokens continuously amplify commercial value across diverse use cases. Going forward, Xunce will continue to iterate its full-stack product ecosystem, enabling precise computing allocation for diverse enterprise AI scenarios and empowering businesses to transform raw data resources into scalable, real-world AI productivity.