What Happened: Zhipu AI's First Earnings Report Sends Shockwaves Through Hong Kong Markets
Shares of Zhipu AI surged as much as 35% on Wednesday, hitting a record high of HK$938, before trimming gains to around 30% by afternoon trading in Hong Kong. The rally followed the company's first earnings release since its landmark stock market debut and delivered the kind of headline-grabbing numbers that investors in China's artificial intelligence sector have been waiting for.
Revenue for the full year ended December 2025 climbed 131.9% to 724.33 million yuan, equivalent to roughly $104.8 million, according to the Beijing-based company's filing released on Tuesday. While the figure represents a more than doubling of top-line income in a single year, it came in slightly below the mean analyst estimate of 760 million yuan. That minor shortfall did little to dampen the mood of investors, who appeared far more focused on the trajectory than the gap.
The stock's move tracked a broader rally in Chinese technology equities, with the Hang Seng index climbing approximately 2% on the same day. Zhipu was not the only beneficiary. Shares of MiniMax, a competing Chinese AI firm that also listed in Hong Kong in January, climbed roughly 16% in Wednesday trading.
Who Is Zhipu AI and Why Does It Matter to Global Investors
Founded in 2019 by researchers from Tsinghua University, Zhipu is considered one of China's so-called AI tigers, a cohort of startups building large language models to rival the likes of OpenAI and Anthropic, making it a strong barometer for China's overall AI landscape.
The company went public in Hong Kong in January, raising approximately $558 million in one of the first listings by a pure-play AI model developer anywhere in the world. That IPO drew significant institutional and retail attention, positioning Zhipu as a flagship bet on China's ambition to become a self-sufficient artificial intelligence superpower.
Zhipu is the first foundational AI model startup in the world to complete an IPO, a distinction that carries real weight at a moment when global investors are searching for direct exposure to the AI value chain beyond the established giants.
How Zhipu's Revenue Growth Breaks Down Across Business Lines
The headline growth figure of 132% is striking on its own, but the composition of that revenue tells a more nuanced story about where Zhipu's commercial momentum is concentrated.
On-premise deployments, where AI models run on the company's own servers, accounted for 74% of total revenue. Meanwhile, cloud-based deployments, where the company does not directly host its AI services, saw revenue jump nearly 300%.
Revenue from the open platform and API, which covers cloud services, surged 292.6% year-on-year to 190 million yuan, while revenue from the enterprise-grade intelligent agent business grew 248.8% year-on-year to 166 million yuan.
CEO Zhang Peng said the company is undergoing a deliberate strategic shift from its traditional on-premises model deployment toward cloud-based services. That pivot is not without short-term cost. The gross margin on enterprise-level general large models declined from 69.6% to 47.0%, a significant compression that the company attributed to the higher delivery resources required to support local deployment customers and the growing share of cloud business.
What the Losses Reveal About Zhipu's Long-Term Strategy
Revenue growth at this pace rarely comes cheap, and Zhipu's financials make that reality transparent.
Total losses soared 59.5% to 4.72 billion yuan in 2025, driven by mounting research and development spending, which jumped 44.9% to 3.18 billion yuan. Adjusted net losses totalled 3.18 billion yuan, a 29.1% increase from the previous year.
R&D expenditure reached the equivalent of 439% of total revenue, a ratio that underscores just how aggressively Zhipu is investing in foundational model development relative to what it currently earns. That is a deliberate posture, not an accounting anomaly.
CEO Zhang Peng said he expects exponential growth for the company's cloud-based models business, and revealed during a post-earnings call that Zhipu is benchmarking its commercialization strategy against Anthropic. For an AI company that only recently crossed the threshold of public markets, that level of competitive ambition signals clear intent.
Where Zhipu Stands Amid U.S. Export Restrictions and Chip Constraints
No analysis of a Chinese AI company's results would be complete without examining the geopolitical pressure bearing down on the sector.
Zhipu has faced constraints due to U.S. export restrictions that have limited its access to advanced chips and related technologies. The company was also added to the U.S. Commerce Department's Entity List last year, restricting its ability to source certain components.
Rather than treating those restrictions as a ceiling, Zhipu appears to be treating them as a forcing function. During the earnings call, CEO Zhang Peng said the company is accelerating its use of domestic Chinese chips to meet a significant rise in computing demand since February, aligning directly with Beijing's push to develop its homegrown semiconductor industry.
Through technical optimizations including Muon Split and MLA-256 improvements, Zhipu reduced GPU memory usage during model training while maintaining performance, and its dynamic sparse attention mechanism reportedly cut deployment costs by 50%. These are not workarounds driven by desperation. They represent constraint-driven innovation that could give Zhipu structural cost advantages as the technology matures.
What Zhipu's GLM Models and AI Agent Business Signal for Enterprise Adoption
Zhipu recently released its GLM-5 model, which it claims matches U.S. rivals on several key performance benchmarks. The company has also been expanding its AI agent offerings substantially over the past year.
Nine of China's top ten internet companies are described as deeply integrated with Zhipu's GLM platform, and the platform had crossed 4 million registered users as of March 2026. That level of enterprise penetration within a single market is not trivial. It reflects a product that has moved well beyond pilot programs and into operational infrastructure for major organizations.
The company noted that over 4 million small and medium-sized enterprises and developers have collaborated with its products, which are now available across 218 countries and regions.
While China remains its primary market, Zhipu has begun expanding overseas, particularly into Southeast Asia. That international push will be one of the clearest indicators of whether its commercial model can scale beyond the structural tailwinds of domestic government support.