行业报告 13小时前 98 阅读 0 评论

SenseTime Reports Strong H1 2025 Results, Generative AI Drives Growth Amid Strategic Restructuring

作者头像
AI中国

AI技术专栏作家 | 发布了 246 篇文章

TMTPOST – Hong Kong-listed artificial intelligence leader SenseTime Group posted interim results for the first half of 2025 that exceeded market expectations, highlighting robust growth in generative AI and the impact of its “1+X” strategic framework.

For the six months ending June 30, SenseTime reported total revenue of RMB 2.358 billion (approximately US$330 million), up 35.6% year-on-year. Adjusted net loss narrowed sharply to RMB 1.162 billion (around US$162 million), a 50% reduction from the same period last year, while adjusted EBITDA losses fell 72.5% to RMB 521 million (US$73 million).

Trade receivable collections surged 96% to RMB 3.2 billion (US$447 million), setting a new record, and cash reserves reached RMB 13.2 billion (US$1.84 billion) following a recent HKD 2.5 billion placement aimed at expanding AI infrastructure and research and development.

SenseTime’s “1+X” strategy positions Generative AI and Computer Vision as core growth engines (“1”), while the “X” segment focuses on innovative verticals including Smart Auto, Smart Healthcare, Home Robotics, and Smart Retail. Chairman and CEO Dr. Xu Li emphasized that this framework is accelerating adoption across industries while improving operational efficiency and financial transparency.

“Amid China’s State Council ‘AI Plus’ initiative, SenseTime is implementing real-world applications at scale,” Xu said. “Our dual engines of Generative AI and Computer Vision enable sustained leadership and create long-term value for employees, customers, and shareholders.”

The generative AI segment accounted for 77% of total revenue, reaching RMB 1.815 billion (US$253 million), a 73% year-over-year increase. SenseTime’s SenseNova large model leads in global multimodal capabilities, deployed in productivity tools and interaction platforms.

The Raccoon series, with finance, education, and public service variants, surpassed 3 million users, while SenseNova V6.5 integrates text, audio, and video to enable multimodal real-time interactions, with engagement hours up 510% year-on-year. These advancements have strengthened enterprise and consumer adoption, reinforcing loyalty and industrial scalability.

The Computer Vision segment served over 660 domestic and international clients by mid-2025, with a 57% repeat purchase ratio. Leveraging high-quality client engagement and market leadership, the division is expected to further enhance profitability and cash flow in the second half of the year.

SenseTime’s X-Innovation segment, including smart healthcare (SenseCare/Shancui), home robotics (Yuanluobo), smart retail (SenseRetail/Shanhui), and intelligent driving (Jueying Automotive), operates under autonomous founding teams through the company’s “Re-CoFound” organizational transition. While revenue totaled RMB 106 million (US$15 million), down 39.5% year-on-year, CFO Wang Zheng noted these ventures are designed to operate independently, secure external financing, and gradually deconsolidate from group accounts, supporting overall profitability.

Collectively, several X-Innovation businesses have raised RMB 1 billion (US$139 million) in external financing, while additional investments in projects such as Unisound, GalBOT, PaXiniTech, Rokid, BrainCo, and Terminus Technologies remain part of the X portfolio.

SenseTime revised its reporting structure from “Generative AI, Visual AI, and Intelligent Vehicles” to “Generative AI, Visual AI, and X-Innovation Businesses.” Generative AI remains defined as Large Models + Large Devices, Visual AI now includes smart cockpit-related businesses, and the X-Innovation segment focuses on four vertical tracks, with adjustments made according to financing and governance developments.

SenseTime’s large device platform reached 25,000 PetaFLOPS of computing power as of August 2025 and completed an upgrade to SenseCore 2.0, improving inference performance by roughly 15%. Its heterogeneous cluster of 5,000 domestic chips operates at 80% utilization, with training efficiency achieving 95% of homogeneous clusters.

The company also partnered with CATL on lithium-ion battery energy storage systems, reducing internal electricity costs while contributing to regional energy balance. Despite competition from major cloud providers, SenseTime remains a top-three leader in technology and market performance, leveraging its “large model + large device + AI application”integration and customer insights.

The company strengthened cost management, with sales and administrative expenses at RMB 272 million (US$38 million) and R&D spending increasing to RMB 2.119 billion (US$295 million), improving the ratio of R&D to revenue. Trade receivables totaled RMB 6.597 billion (US$917 million), slightly down from the end of 2024.

The HKD 2.5 billion placement will support expansion of large-scale devices, generative AI R&D, and innovation in AI verticals including embodied intelligence, blockchain, and stablecoins.

Executive VP and Chief Scientist Lin Dahua highlighted the importance of multimodal AI and AI Agents for expanding commercial applications. The RiRiXin large model roadmap focuses on evolving multimodality, optimizing model architectures, leveraging hardware, and enhancing intelligent agent interactions.

Dr. Xu outlined three industry trends: standardization of computing and data, large-scale industrial adoption of AI, and the need for governance and international collaboration as applications expand.

Following the State Council’s “AI Plus” initiative, SenseTime shares surged over 10% on Wednesday, closing up 8.9%, and have risen roughly 40% since early July, with a 48.2% year-to-date increase and a market capitalization of HKD 80.829 billion (US$10.9 billion).

With Generative AI and Computer Vision driving growth, a robust “1+X” strategy, and advancements in AI infrastructure and multimodal capabilities, SenseTime is positioned to capitalize on China’s accelerating AI landscape. By balancing high-growth innovation with disciplined financial management, the company is laying the foundation for sustained leadership and a near-term path to profitability.

作者头像

AI前线

专注人工智能前沿技术报道,深入解析AI发展趋势与应用场景

246篇文章 1.2M阅读 56.3k粉丝

评论 (128)

用户头像

AI爱好者

2小时前

这个更新太令人期待了!视频分析功能将极大扩展AI的应用场景,特别是在教育和内容创作领域。

用户头像

开发者小明

昨天

有没有人测试过新的API响应速度?我们正在开发一个实时视频分析应用,非常关注性能表现。

作者头像

AI前线 作者

12小时前

我们测试的平均响应时间在300ms左右,比上一代快了很多,适合实时应用场景。

用户头像

科技观察家

3天前

GPT-4的视频处理能力已经接近专业级水平,这可能会对内容审核、视频编辑等行业产生颠覆性影响。期待看到更多创新应用!