お知らせ • Sep 03

Xunce Technology Launches TokenCloud AI Model Training And Deployment Platform

Xunce Technology officially launched the TokenCloud all-in-one AI model training and deployment computing power platform. TokenCloud builds a four-tier, synergistic ecosystem based on heterogeneous computing devices, powered by mainstream algorithmic models, fueled by multi-source internal and external data, and serving vertical industry clients. The launch of TokenCloud marks Xunce Technology’s formal establishment of a full-chain product closed-loop encompassing “computing power—data—tokens—models—applications.” TokenCloud was created precisely to address these pain points. Positioned as the core “hardware and energy foundation” at the upstream of the token industry chain, it adheres to the principle of “ensuring every unit of computing power precisely serves every enterprise AI scenario.” It has built a complete closed-loop process spanning “scenario identification ? model selection ? hardware matching ? training optimization ? secure delivery.” Based on business scenarios and tasks, it intelligently recommends the optimal model and matches it with the most suitable hardware. It then uses distillation optimization to enable lightweight, “specialized” models to achieve accelerated inference with high accuracy, all the way through to secure delivery. With each step seamlessly linked, it provides enterprise customers with out-of-the-box AI infrastructure. In terms of product design, TokenCloud adopts a “full-chain visualization” philosophy, integrating the previously disparate processes of model selection, computing power configuration, security management, and cost assessment in AI deployment into a unified decision-making platform. Comprising five core modules—solution selection, model distillation, data security management, computing power acceleration, and computing resource management—this platform shifts enterprise AI deployment from “rule-of-thumb” approaches to “data-driven decision-making.” It ensures that the return on investment for every computing power allocation is clearly quantifiable, achieving a dynamic balance among performance, cost, and security. Through Hardware-Software Co-Optimization, TokenCloud enables the large-scale supply of high-quality tokens—ensuring that every piece of data unleashes its maximum token value through the optimal combination of models and hardware. TokenCloud has established industry-leading technological barriers through four core capabilities: a heterogeneous hardware platform, Cloud-Edge Collaboration Architecture, model evaluation and optimization, and secure isolation. Heterogeneous Hardware Platform—Integrates diverse computing resources such as GPUs, NPUs, FPGAs, and CPUs to enable unified management and intelligent scheduling across architectures. It supports full-scenario coverage ranging from hundreds to tens of thousands of cards, abstracting underlying, disparate hardware resources into unified computing units. Cloud-Edge Synergy Architecture—Establishes a hybrid deployment model combining “cloud computing power and on-premises computing power,” enabling elastic scaling to handle large-scale training and high-concurrency inference while ensuring that enterprises’ raw data remains within secure domains. Model Evaluation and Optimization System—Establishes a comprehensive evaluation framework covering foundational capabilities, scenario adaptation, engineering optimization, and ecosystem value. It supports task-specific distillation optimization, transferring the core capabilities of large models to lightweight scenario-specific models, thereby significantly reducing inference latency and hardware costs while maintaining high accuracy. Through end-to-end acceleration capabilities, TokenCloud significantly reduces first-token latency and enhances concurrent processing capacity, transforming acceleration cards from “waiting for computation” to “continuous computation.” Security Isolation Mechanism—We build a multi-layered security protection system where data visibility is tiered and data flow is fully traceable throughout its entire journey. Highly sensitive data is 100% locked within the enterprise’s local security domain, with a three-pronged approach combining multi-tenant isolation, encrypted transmission, and access control. TokenCloud achieves deep synergy between models and computing power—the platform centrally determines the degree of model distillation and the hardware for deployment, ensuring the optimal balance of accuracy, speed, cost, and efficiency. It provides enterprise customers with a comprehensive, one-stop service solution—from model selection and hardware matching to distillation optimization—covering the entire “computing power—data—Token—model—application” chain to accelerate the implementation of enterprise AI. The launch of TokenCloud has enabled Xunce Technology’s full-path product ecosystem to form a complete commercial loop. Previously, the company had introduced the world’s first TokenOS operating system; the two systems have clearly defined roles and operate in synergy. TokenOS focuses on data capabilities, transforming enterprises’ multi-source, heterogeneous data into standardized scenario-based Tokens in real time, thereby addressing the issue of “data usability.” TokenCloud focuses on model training and deployment, as well as “data + model” integration capabilities. It integrates computing power scheduling, model inference optimization, and the refinement, tuning, and orchestration of enterprise-level small models, bridging the entire chain from data ingestion to application deployment and providing hardware service support for Token generation. Through this synergy between software and hardware, combined with the underlying AIDP data resource processing service system and the top-level TokenRouters model outcomes service and scenario integration and exchange center, the two solutions together form a complete “computing power—data—Token—model—application” closed-loop ecosystem. Xunce Technology has partnered with multiple domestic GPU manufacturers to focus on the integration of “computing power and data.” As the localization of the full AI stack enters the stage of systematic engineering optimization, TokenCloud has become a key hub connecting domestic heterogeneous computing power with enterprise AI applications, offering significant strategic value within the ecosystem. The company has expanded into eleven high-value industries, including finance, telecommunications, power, energy, high-end manufacturing, biomedicine, robotics training, commercial space, urban operations, the low-altitude economy, and smart vehicles. In the first half of 2026, the Token business model was successfully implemented in high-value scenarios such as biomedicine, energy dispatch, and industrial quality inspection, laying the groundwork for cross-industry replication.

IT 業界の成長	6.82%
収益成長率	275.28%
株主資本利益率	1.83%
ネット・マージン	3.30%
前回の決算情報	30 Jun 2026

日付	収益	収益	G+A経費	研究開発費
30 Jun 26	2,054	68	448	845
31 Mar 26	1,669	-13	403	731
31 Dec 25	1,285	-94	359	616
30 Sep 25	916	-83	216	493
30 Jun 25	547	-72	124	370
31 Mar 25	590	-78	137	410
31 Dec 24	632	-84	150	450
31 Dec 23	530	-55	133	379
31 Dec 22	288	-75	92	259

データ	最終更新日（UTC時間）
企業分析	2026/09/21 00:28
終値	2026/09/21 00:00
収益	2026/06/30
年間収益	2025/12/31

パッケージ	データ	タイムフレーム	米国ソース例
会社財務	10年	損益計算書キャッシュ・フロー計算書貸借対照表	SECフォーム10-K SECフォーム10-Q
アナリストのコンセンサス予想	+プラス3年	予想財務アナリストの目標株価	アナリストリサーチレポート Blue Matrix
市場価格	30年	株価配当、分割、措置	ICEマーケットデータ SECフォームS-1
所有権	10年	トップ株主インサイダー取引	SECフォーム4 SECフォーム13D
マネジメント	10年	リーダーシップ・チーム取締役会	SECフォーム10-K SECフォームDEF 14A
主な進展	10年	会社からのお知らせ	SECフォーム8-K

アナリスト	機関
Yixi Lian	Citic Securities Co., Ltd.
Leo Chiang	Deutsche Bank
Tu Wu	Guotai Haitong/Guotai Junan International Holdings Limited

Shenzhen Xunce Technology 過去の業績

過去基準チェック /26

主要情報

最近の業績更新

First half 2026 earnings released: EPS: CN¥0.23 (vs CN¥0.30 loss in 1H 2025)

Shenzhen Xunce Technology Co., Ltd. to Report First Half, 2026 Results on Aug 21, 2026

Shenzhen Xunce Technology Co., Ltd. to Report Fiscal Year 2025 Results on Mar 27, 2026

Recent updates

Xunce Technology Launches TokenCloud AI Model Training And Deployment Platform

First half 2026 earnings released: EPS: CN¥0.23 (vs CN¥0.30 loss in 1H 2025)

Shenzhen Xunce Technology Co., Ltd. to Report First Half, 2026 Results on Aug 21, 2026

Forecast breakeven date pushed back to 2027

Forecast breakeven date moved forward to 2026

Shenzhen Xunce Technology Co., Ltd. Provides Consolidated Earnings Guidance for the Six Months Ended June 30, 2026

Shenzhen Xunce Technology Co., Ltd. has filed a Follow-on Equity Offering in the amount of HKD 784.3791 million.

Shenzhen Xunce Technology Co., Ltd., Annual General Meeting, Jun 26, 2026

Less than half of directors are independent

Shenzhen Xunce Technology Co., Ltd. to Report Fiscal Year 2025 Results on Mar 27, 2026

収支内訳

収益と収入の歴史

フリー・キャッシュフローと収益の比較

過去の収益成長分析

株主資本利益率

総資産利益率

使用総資本利益率

過去の好業績企業の発掘

企業分析と財務データの現状

データソース

分析モデルとスノーフレーク

業界およびセクターの指標

アナリスト筋

Shenzhen Xunce Technology 過去の業績

過去 基準チェック /26

主要情報

最近の業績更新

First half 2026 earnings released: EPS: CN¥0.23 (vs CN¥0.30 loss in 1H 2025)

Shenzhen Xunce Technology Co., Ltd. to Report First Half, 2026 Results on Aug 21, 2026

Shenzhen Xunce Technology Co., Ltd. to Report Fiscal Year 2025 Results on Mar 27, 2026

Recent updates

Xunce Technology Launches TokenCloud AI Model Training And Deployment Platform

First half 2026 earnings released: EPS: CN¥0.23 (vs CN¥0.30 loss in 1H 2025)

Shenzhen Xunce Technology Co., Ltd. to Report First Half, 2026 Results on Aug 21, 2026

Forecast breakeven date pushed back to 2027

Forecast breakeven date moved forward to 2026

Shenzhen Xunce Technology Co., Ltd. Provides Consolidated Earnings Guidance for the Six Months Ended June 30, 2026

Shenzhen Xunce Technology Co., Ltd. has filed a Follow-on Equity Offering in the amount of HKD 784.3791 million.

Shenzhen Xunce Technology Co., Ltd., Annual General Meeting, Jun 26, 2026

Less than half of directors are independent

Shenzhen Xunce Technology Co., Ltd. to Report Fiscal Year 2025 Results on Mar 27, 2026

収支内訳

収益と収入の歴史

フリー・キャッシュフローと収益の比較

過去の収益成長分析

株主資本利益率

総資産利益率

使用総資本利益率

過去の好業績企業の発掘

企業分析と財務データの現状

データソース

分析モデルとスノーフレーク

業界およびセクターの指標

アナリスト筋

過去基準チェック /26