SenseTime delivers its strongest performance ever! Revenue exceeds 5 billion yuan, EBITDA turns positive in half a year, and generative AI surges.

Ask AI · How can ecosystem collaboration under the “1+X” strategy strengthen SenseTime’s business resilience?

As the AI industry gradually shifts from a technical race to realizing value, companies’ competitive logic is moving from breakthroughs in isolated areas to building system-level capabilities.

Against this backdrop, what SenseTime’s latest financial report shows is not only growth in revenue scale, but also a phased improvement in its loss structure—“the less you lose, the less you do”—and, more importantly, the systematic release of its full-stack capabilities centered on “compute power—models—applications.”

In fiscal year 2025, SenseTime achieved total revenue growth of 33%, exceeding 5 billion yuan, setting a new historical high and marking the fastest growth rate in nearly three years. Full-year net losses narrowed significantly by 58.6%. On an adjusted basis, the net loss achieved year-over-year accelerated loss reduction for four consecutive half-years. In the second half of the year, earnings before interest, taxes, depreciation, and amortization (EBITDA) was 380 million yuan; this was the first time it turned positive after going public, exceeding market expectations.

Half-year EBITDA turns positive—SenseTime steps into a milestone turning point

Looking back at 2025, SenseTime delivered a financial report with historic significance.

On the revenue side, growth momentum continued: full-year revenue exceeded 5 billion yuan, reaching a historical high, up 33% year over year, the fastest pace in nearly three years. While continuing to increase R&D investments, SenseTime also completed a critical transformation in operating quality: net losses narrowed sharply by 59% year over year. In the second half of 2025, the company’s EBITDA (earnings before interest, taxes, depreciation, and amortization) turned positive for the first time since listing, reaching 380 million yuan.

This combination of “revenue growth + narrowing losses” also means that, driven by generative AI, SenseTime is gradually finding a balance between technology investment and commercialization.

Even more noteworthy is the improvement in operating cash flow. In the second half of 2025, the group’s operating cash flow achieved a net positive inflow for the first time since listing. In terms of cash management, trade receivables collection reached 4.87 billion yuan, also a historical high. The cash conversion cycle (CCC) dropped dramatically from 228 days at the end of 2024 to 129 days at the end of 2025, significantly improving capital utilization efficiency. Visual AI, one of SenseTime’s cornerstone businesses, achieved net profit for CV 2.0 for the first time during the reporting period and maintained positive cash flow for two consecutive years.

Taken together, these changes in financial indicators fundamentally reflect the aggregation potential of the company’s business structure.

In 2025, all of SenseTime’s business segments showed improvements to varying degrees.

The generative AI business sustained its explosive growth trajectory: full-year revenue was RMB 3.6 billion, up 51% year over year, and it remains the group’s core business. After successfully breaking through the boundaries of traditional perception intelligence, the visual AI business is gradually shifting from the technology investment phase to the scaled earnings phase, becoming a true pillar supporting both revenue growth and cash flow improvement. During the reporting period, the visual AI business generated revenue of 1.08 billion yuan, up 3% year over year; and in the second half of 2025, the year-over-year growth rate increased further to 21%. This indicates that SenseTime has formed a more resilient operating system by reinvesting profits from mature businesses into frontier exploration.

X innovation business revenue exceeded 300 million yuan in 2025. During the reporting period, the group’s incubated ecosystem enterprises gained strong recognition from internet giants, top-tier venture capital firms, and industrial funds in the primary market. External financing progressed smoothly, showcasing strong growth resilience.

From a capital market perspective, valuation frameworks for technology companies are currently undergoing adjustment. For SenseTime, the rise in its revenue and profitability, combined with the long-term growth expectations for the AI industry, provides a foundation for it to attract attention in the next round of value reappraisal. In other words, the improvement in financial data is not only an outcome at the operating level, but also support for its long-term positioning in the capital market.

SenseTime Technology’s Chairman and CEO Xu Li stated: “We firmly believe that the deep integration of language and vision is an efficient path to breaking through the ceiling of intelligence. Based on the NEO native architecture, we efficiently unify understanding and generation, exploring a new ‘law of scale’ for multimodal data. This series of technological breakthroughs, when deeply combined with agent AI, will open up new application possibilities and empower entirely new vertical scenarios. While continuing to innovate at the underlying technology level, SenseTime has also achieved dual growth in both revenue and EBITDA, demonstrating outstanding growth resilience and operating efficiency in the industry, and steadily moving toward high-quality development.”

The value of the “1+X” strategy becomes clear—efficient collaboration between “mother ship + child ships” to compete in the AI finals

If financial improvement is the result, then the core logic driving that result comes from the deepening of the “1+X” strategy.

Among them, “1” represents the group’s core business, including generative AI and visual AI, which together form the company’s technical foundation and revenue base. “X,” on the other hand, refers to innovation businesses incubated by the group with independent growth potential. These include intelligent driving with frontier world-model and mass-production delivery capabilities; high-performance inference GPUs; edge AI chips that empower the intelligence of everything; Meta萝卜, which pioneers a new category of home AI robots; intelligent medical care and smart retail focused on industry digital transformation; as well as the Da Xiao robot representing frontier breakthroughs in embodied intelligence, among others.

Around the “1” business, SenseTime further builds a three-in-one full-stack capability closed loop: “compute infrastructure—large model R&D—large model applications.”

In 2025, SenseTime continued to invest in frontier technology R&D, achieving breakthrough progress in large model architecture innovation, training paradigms, inference efficiency, and spatial understanding, while continuously maintaining a leading position domestically. SenseTime’s 日日新 multimodal large model kept improving in performance, continuing to top leaderboards. It ranked first in comprehensive evaluations such as SuperCLUE and OpenCompass in January 2025, May 2025, and December 2025, respectively. Meanwhile, in 2025, SenseTime released and open-sourced the SenseNova-SI space intelligence model series. In evaluations across multiple internationally authoritative space intelligence benchmarks, SenseNova-SI ranked first globally among comparable models. Kairos-SenseNova, also open-sourced, became the first open-source embodied native world model to realize an end-to-end integration of “multimodal understanding—generation—prediction.” In December 2025, SenseTime released and open-sourced a new native multimodal model architecture, NEO. With only one-tenth of the training data and compute used by industry-equivalent models, it can achieve top-tier performance. These breakthrough advances not only redefine inference efficiency and training paradigms for models, but also signal that SenseTime has officially entered a new stage of development on the multimodal integration path.

Building on the leading performance of the 日日新 multimodal large model, SenseTime excavates the closed-loop value of agents in strategic tracks such as office work, finance, marketing, and content generation through the comprehensive dimensions of task complexity and fault tolerance. In addition, leveraging the core technical advantages of the 日日新 large model—such as full-modal interaction, low latency, and long-term memory—SenseTime is embedding multimodal large models deeply into various terminal hardware, including smartphones, AI glasses, in-vehicle interaction systems, and embodied intelligence devices. Customers include Xiaomi, BYD, Tecno, and others.

In 2025, the Big Device achieved a deep leap from a technical long-board to an industry closed-loop. Throughout the year, the Big Device supported nearly one million model R&D tasks, connecting the complete path from underlying hardware to top-level applications, from the software stack to model adaptation. Among them, the LightX2V world model inference system achieved a performance overtake of overseas chips on domestic hardware platforms.

The mode of “mother ship (group) provides the foundation, child ships (ecosystem) compete in the tracks” provides a broader space for development. By incubating enterprises in different fields, SenseTime extends its technical capabilities into multiple tracks, forming a diversified business structure. This ecosystem collaboration not only enhances the company’s overall ability to withstand risks, but also enables it to capture more growth opportunities in niche segments.

From “1” to “X,” and then from “X” back to “1,” SenseTime is forming a self-growing, self-evolving business economic entity, laying a solid foundation for long-term sustainable development.

With the “1+X” strategy entering the value validation stage, SenseTime is responding to the macro proposition of the “intelligent economy” with an all-stack layout. It is deeply coupling financial performance, business structure, and industry trends, gradually building certainty for the next growth cycle.

Building a blueprint for the intelligent economy

In 2026, the “intelligent economy” is formally proposed as a new economic form. Against this macro backdrop, SenseTime’s business layout provides a representative practical case.

The report proposes implementing new infrastructure projects such as ultra-large-scale intelligent compute clusters and compute-power-and-electricity coordination, strengthening nationwide integrated compute monitoring and dispatch to support the development of public cloud. This wording elevates compute power from a single technical factor to a national strategic infrastructure base. SenseTime’s “Big Device” is a forward-looking implementation of this strategy.

The Big Device continues to accelerate domestic ecosystem collaboration. Together with more than ten chip manufacturers—including Huawei Ascend, Hygon, and Cambricon—it jointly released “SenseTime Big Device Compute Mall,” forming a complete domesticized solution from underlying hardware to top-level applications. It has now become a core partner for leading research institutions, internet giants, pan-entertainment enterprises, embodied intelligence companies, and large-model unicorns, and it has also been implemented in Saudi Arabia as the first domestic overseas compute cluster from China.

At the same time, the “compute-power-and-electricity coordination” AI intelligent system built with support from Contemporary Amperex Technology Co., Limited (CATL) achieves end-to-end integration of compute management and energy dispatch. It is estimated to bring 7% savings in electricity costs and carbon emissions reduction exceeding 4,000 tons per year per PetaFLOPS. This practice, with the “compute-model-application” closed loop at its core and domestic collaboration and green development as its extension, enables SenseTime’s Big Device to be not only the company’s own core competitive advantage, but also a public infrastructure that empowers industries and can be shared across the entire ecosystem.

The report also emphasizes building a new form of the intelligent economy. It calls for further expanding “artificial intelligence +,” promoting faster adoption of the next generation of intelligent terminals and intelligent agents, driving large-scale commercialized applications of AI in key industry sectors, and cultivating new business formats and operating models for intelligent natives.

This is exactly the goal that SenseTime’s “1+X” and “three-in-one” strategies are committed to achieving. In the application layer, SenseTime is deeply embedding AI capabilities into specific industry scenarios such as general office automation, finance, intelligent marketing, and content generation. Across a rich and diverse set of scenarios, AI is gradually becoming a key tool for improving efficiency and creating value.

SenseTime breaks down its future strategic priorities in its financial report. On one hand, it remains on the native multimodal large model route to build an entry point for the Agent-native era. At the same time, it optimizes compute costs and advances domestic substitution. In the visual AI domain, it further consolidates its profitability, turning the “SenseTime Solution” into a global benchmark for smart industries. In addition, it continues to release ecosystem value and兑现 the “1+X” strategy dividends.

As the industry gradually moves into deeper waters, a single technical advantage can no longer support long-term growth. True competition will take place between system-level capabilities and commercial efficiency. From this perspective, SenseTime’s current demonstrated financial improvements, strategy execution, and ecosystem building are forming a new growth paradigm. The core of this paradigm is not whether there are short-term earnings fluctuations, but whether it can continuously establish a positive feedback loop between technology, business, and the industry.

For the entire AI industry, this may also be an important signal: as the intelligent economy moves from concept to reality, companies’ value will no longer be determined solely by technological leadership. Instead, it will depend on whether they can achieve coordinated evolution of technology, capital, and industry within complex systems. / Written by Jin Song

(This article does not constitute any investment advice. The information disclosure shall be based on company announcements. Investors act at their own risk.)

Daily Economic News