Gasgoo Munich-SenseTime (0020.HK) unveiled its 2025 financial results on March 24, reporting annual revenue of 5.01 billion yuan. That represents a 33% year-on-year surge—the fastest clip in three years—and a record high. Buoyed by key operational breakthroughs, the company is shifting gears from a heavy investment phase to a period of sustainable, rapid expansion.
Image Source: SenseTime
SenseTime's net loss for 2025 narrowed by 58.6%, with adjusted losses shrinking at an accelerating pace for four consecutive half-year periods. Earnings before interest, taxes, depreciation, and amortization (EBITDA) turned positive in the second half, reaching 380 million yuan—the first profit at this level since its IPO and a figure that topped market expectations. The company's ability to generate cash took a qualitative leap: collection of trade receivables hit a record 4.87 billion yuan, while operating cash flow swung to a positive net inflow for the first time post-listing. That signals a marked improvement in both capital efficiency and cash flow stability.
Both R&D and commercialization delivered standout results. SenseTime's SenseNova multimodal large model continues to lead the pack, showing marked improvements in task completion quality and token cost optimization. The visual AI business reached a milestone, with CV 2.0 turning a net profit for the first time and maintaining positive cash flow for two straight years. Deployment of AI applications across various scenarios accelerated: the Xiaohuans family of products now serves thousands of enterprise users and more than 15 million individuals. Meanwhile, AI-native consumer apps like Kapi Camera and Kapi Accounting have climbed the App Store charts in multiple countries, further refining the ecosystem for AI agent applications.
The deep integration of language and vision is the most efficient path to breaking the limits of intelligence, according to Xu Li, SenseTime's Chairman and CEO. Built on the NEO native architecture, the company has unified understanding and generation to explore new "scaling laws" for multimodal AI. This combination of technical breakthroughs and deep integration with AI agents will unlock new application possibilities and empower a fresh range of vertical scenarios.
Separately, SenseTime announced plans to launch a new model based on the second-generation NEO architecture in the second quarter of 2026. The company expects this model to deliver a multiple-fold leap in both efficiency and cost-performance, further empowering AI agent applications across the board.
