New high! SenseTime Technology's 2025 revenue exceeds 5 billion yuan, with EBITDA turning positive in half a year

On March 24, SenseTime Technology (0020.HK) announced its annual performance results. In the fiscal year 2025, the group’s total revenue grew by 33%, exceeding 5 billion yuan, setting a historical record with the fastest growth rate in nearly three years; the annual net loss significantly narrowed by 58.6%, with adjusted net losses achieving a continuous four-semester year-on-year reduction, and the EBITDA in the second half of the year was 380 million yuan, turning positive for the first time since listing, surpassing market expectations. The company successfully transitioned from a technology investment phase to a sustainable explosive growth phase. During the reporting period, SenseTime achieved a “qualitative” leap in its self-sustaining capabilities, continuously improving capital efficiency, with healthy cash flow, and trade receivables reaching 4.87 billion yuan, a historical high; in the second half of the year, the group’s operating cash flow achieved positive net inflow for the first time since listing. SenseTime will launch a new model based on the second-generation NEO architecture in the second quarter of 2026, which is expected to achieve a multiplicative leap in efficiency and cost-effectiveness, widely empowering AI applications for intelligent agents.

SenseTime’s Chairman and CEO Xu Li stated: “We firmly believe that the deep integration of language and vision is an efficient path to breaking the limits of intelligence. Based on the NEO native architecture, we efficiently realize the unity of understanding and generation, exploring a new ‘law of scales’ for multimodality. This series of technological breakthroughs and the deep integration with intelligent agent AI will unlock new application possibilities, empowering new vertical scenarios. While maintaining innovation in underlying technology, SenseTime has also achieved dual growth in revenue and EBITDA, demonstrating exceptional growth resilience and operational efficiency in the industry, steadily moving towards high-quality development.”

Deep cultivation of native multimodal large models, driven by “efficiency enhancement + cost reduction” to accelerate business growth

In 2025, SenseTime continued investing in cutting-edge technology research and development, achieving breakthrough progress in large model architecture innovation, training paradigms, inference efficiency, spatial understanding, and maintaining a leading domestic level.

SenseTime’s daily updated multimodal large model performance has risen, consistently topping the charts, ranking first in comprehensive assessments such as SuperCLUE and OpenCompass in January, May, and December 2025. Meanwhile, in 2025, SenseTime successively released and open-sourced the SenseNova-SI series of spatial intelligence models, ranking first globally among similar models in several international authoritative spatial intelligence metrics; the open-sourced Kairos-SenseNova became the first to achieve integrated “multimodal understanding-generation-prediction” in an open-source embodied native world model. In December 2025, SenseTime released and open-sourced the new native multimodal model architecture NEO, which can achieve top performance with only one-tenth of the training data and computing power of comparable industry models. These breakthroughs not only redefine the inference efficiency and training paradigms of models but also mark SenseTime’s formal entry into a new generation of development in multimodal integration.

Based on the leading technology of the daily updated multimodal large model, SenseTime explores the closed-loop value of intelligent agents in strategic tracks such as office work, finance, marketing, and content generation through a comprehensive dimension of task complexity and fault tolerance, covering clients in various industries, including automotive, smart terminals, consumer, internet, embodied intelligence, finance, education, and healthcare.

In the general office sector, the Little Panda family has served 15 million individual users and thousands of enterprise clients, achieving a sevenfold increase in monthly active users over the year, significantly shortening business analysis cycles by 90%, and greatly enhancing office efficiency. In the intelligent marketing field, SenseTime comprehensively covers the entire e-commerce process and has reached a deep strategic cooperation with Worth Buying Technology, enhancing store operation efficiency by 20 times, live broadcast operation efficiency by 6 times, traffic investment efficiency by 5 times, and real-time monitoring efficiency by 3 times. In the content creation field, the creative intelligent agent Seko directly produced up to 3 minutes per episode, continuously generating 100 episodes of content, quickly attracting over 300,000 creators online, with monthly active users (MAU) reaching 100,000. The “Financial Intelligent Decision-Making Platform” has achieved rapid revenue expansion in the banking and insurance industries, while these practices in deeply cultivated vertical scenarios have contributed to steady growth for the group and fed back a large amount of high-value decision-making data to enhance the deep logical reasoning capabilities of the “daily updated” model. Relying on the core advantages of the “daily updated” large model in multimodal interaction, low latency, and long memory, SenseTime promotes the deep integration of multimodal large models into various terminal hardware, such as smartphones, AI glasses, in-vehicle interactions, and embodied intelligence, with clients including Xiaomi, BYD, and Transsion.

In addition to forging a solid, scalable, and replicable business model in the To B direction, SenseTime has made progress in exploring next-generation AI-native consumer-grade applications. The Kapi series products have successfully accumulated a user base of millions, validating the explosive potential of AI-native applications and demonstrating SenseTime’s capability to integrate AI technology into daily life.

Large Devices: Deep coordination of underlying computing power and model research and development, witnessing the commercial potential of going global

As a solid technical foundation in the “three-in-one” strategy, the large device achieved a deep transition from technical advantages to industrial closure in 2025. Throughout the year, the large device supported nearly one million model research and development tasks, streamlining the complete path from underlying hardware to top-level applications, from software stack to model adaptation. Among them, the LightX2V world model inference system surpassed the performance of overseas chips on domestic hardware platforms.

The large device continues to accelerate the collaboration of domestic ecosystems, positioned as the “chain master” of the domestic ecosystem. It jointly released the “SenseTime Large Device Computing Mall” with more than a dozen chip manufacturers, including Huawei Ascend, Haiguang, and Cambrian. The large device has become a core partner for top research institutions, internet giants, pan-entertainment companies, embodied intelligence firms, and large model unicorns, establishing China’s first overseas domestic computing cluster in Saudi Arabia.

With the support of CATL, the AI intelligent system created by the large device can accurately predict electricity loads through computing task data using large models, dynamically generating optimal energy scheduling strategies, estimating that this system can bring about a 7% reduction in electricity costs and over 4,000 tons of carbon emissions reduction. As of the date of this performance announcement, the total operational computing power of the large device has increased to 40.4 thousand PetaFLOPS (FP16).

CV2.0 achieves net profit and positive cash flow for the first time! “X” innovative business receives high external capital recognition

Visual AI is transitioning from a technology investment phase to a period of large-scale harvest. During the reporting period, CV2.0 achieved profitability for the first time and maintained two consecutive years of positive cash flow, becoming a solid pillar driving the group’s revenue growth and cash flow improvement. SenseTime has maintained the number one market share in the CV market for nine consecutive years. Clients in Southeast Asia, Northeast Asia, and the Middle East continue to repurchase visual AI products and services, and in the past year, more clients from South America and Europe have shown increased interest, forming a replicable and scalable “Chinese-style going overseas” plan.

In 2025, SenseTime firmly implemented the “1+X” strategy, building an efficient collaborative system where the “mother ship (group) provides the foundation and the sub-ships (ecosystems) compete in their respective tracks,” achieving a leap from single business growth to ecological value multiplication. The ecological enterprises incubated by SenseTime Group have made smooth progress in primary market financing, gaining high recognition from external capital, including internet giants, top venture capital, and industrial funds. The edge-side chip and intelligent driving business sectors have successfully completed financing and achieved independent operations off the balance sheet.

Looking ahead to 2026, the group is ready to embrace the true “finals” of the global AI industry, continuing to delve into native multimodal architecture, establishing SenseTime’s global leadership in the integration of native multimodal large models and spatial intelligence; simultaneously seizing the first-mover advantage in the native AI intelligent agent market, achieving a dual explosion of traffic scale and commercial value. The group will also fully promote the adaptation of domestic chips, continuously reducing the inference costs of large models to solidify product competitiveness with extreme cost-effectiveness. In the visual AI business, it will continue to activate dual engines at home and abroad, promoting large-scale expansion and creating a global benchmark for the intelligent industry. The group will continue to implement the “1+X” strategy, capturing the value-added dividends brought by the explosion of intelligence in vertical industries.

(Editors: Wang Zhiqiang HF013)

[Disclaimer] This article represents only the views of a third party and does not represent the position of Hexun.com. Investors should act at their own risk based on this.

	 [Advertisement] This article represents the author's personal views and is unrelated to Hexun.com. Hexun maintains neutrality regarding the statements and opinions in the text and does not provide any express or implied guarantees regarding the accuracy, reliability, or completeness of the content. Readers should refer to this and assume full responsibility. Email: news_center@staff.hexun.com

Report