Better than expected! Despite an 83% price increase, it still sells out rapidly! Zhipu's annual report causes a frenzy, with ARR skyrocketing 60 times, and the four major domestic computing power core tracks may enter a golden development period!

Ask AI · Zhipu API Price Adjustment Up 83% Why Demand Does Not Decrease but Increases?

According to “Oriental Fortune Research Center,” Zhipu released its first annual report after listing on the evening of March 31, with 2025 performance exceeding expectations. Zhipu’s MaaS API platform achieved an ARR of 1.7 billion yuan (about $250 million), a 60-fold year-over-year increase. Meanwhile, the gross profit margin of the MaaS API platform rose nearly 5 times year-over-year to 18.9%, significantly enhancing profitability. In 2025, the company achieved total revenue of 724 million yuan (RMB), a 131.9% increase; adjusted net loss for the year was 3.18 billion yuan, compared to 2.46 billion yuan in 2024; the full-year gross margin was 41%, far above industry standards.

Notably, Zhipu’s API call pricing was increased by 83% in Q1 2026. Even so, the market still shows a supply shortage, with call volume increasing by 400%. The annual report released by Zhipu demonstrates strong user activity and willingness to pay, indicating that its business model has gained initial market recognition. It also highlights the market competitiveness of domestic large models and the high prosperity of the entire domestic computing power chain. Affected by this, MINIMAX-W surged over 17% intraday, returning above 1,000 HKD; Zhipu once soared over 35%, reaching 938 HKD, setting a new record high, with market capitalization exceeding 400 billion HKD.

Market-Related Institutional Opinions

Guotou Securities: The domestic computing power industry has entered a new stage. Driven by external pressures and endogenous demand, the domestic computing power industry has moved from early single-point breakthroughs to a new stage characterized by “independent technological systems, full-stack ecological capabilities, and closed-loop business verification.” Guojin Securities pointed out that the prosperity of the entire domestic computing power industry chain is accelerating, with prospects for both volume and price increases. Under the strong logic of supply and demand, it is predicted that in 2026, the computing power industry chain will enter a “full-chain inflation” cycle.

Guojin Securities: The prosperity of the entire domestic computing power industry chain is accelerating, with prospects for both volume and price increases. Under the strong logic of supply and demand, it is predicted that in 2026, the computing power industry chain will enter a “full-chain inflation” cycle, with industry prosperity spilling over from core chips to AIDC, cloud and computing services, supporting power equipment, and servers.

China Post Securities: Token call volume is growing exponentially, indicating that data elements are achieving a closed loop from supply to value through a quantifiable model. Competition among large models is shifting from capability comparison to usage volume comparison. AI infrastructure, as the core support for expanding call scale, means underlying support systems such as computing power, networks, and data scheduling must expand synchronously or even proactively, which will benefit from the continuous rise in Token demand.

Bohai Securities: 2026 is expected to be the official start of the intelligent agent development year in the AI industry. Currently, AI applications are experiencing multiple catalysts, and the industry’s commercialization inflection point may arrive earlier. Driven by accelerated model iteration and the application promotion by major internet companies, the computing power industry chain offers relatively certain investment opportunities.

Market-Related Sector Overview

1. Chip Design and Manufacturing: Chips are the “heart” of the computing power industry and the most urgent and valuable segment for domestic substitution. This sector includes general-purpose and specialized chips such as CPUs, GPUs, NPU, as well as wafer manufacturing, packaging, and testing industries. As external technological restrictions become normalized, domestic government and enterprise procurement of chips in critical infrastructure fields is being forcibly increased, providing broad market space for companies like Hygon Information, Loongson Zhongke, Jingjia Micro, and Cambrian. These companies invest heavily in R&D (usually over 30% of revenue), continuously narrowing the gap with international advanced levels in instruction set architecture, interconnection technology, and energy efficiency. This sector features “high investment, high barriers, high elasticity.” Although short-term profits fluctuate due to R&D expense amortization, revenue growth generally exceeds industry averages. Market valuation logic has shifted from PE (price-to-earnings ratio) to PS (price-to-sales ratio) and R&D pipeline valuation, focusing on technological iteration speed and ecosystem adaptation. As domestic wafer foundry capacity is released, supply chain security is further enhanced, and the long-term growth logic remains solid.

2. Server and Complete Machine Manufacturing: Servers are the core equipment that convert chip computing power into actual productivity, representing the largest market size and fastest performance realization in the industry chain. Driven by the “East Data West Computing” project and AI large model training demands, the domestic server market, especially AI servers, continues to explode. Leading companies like Inspur, Sugon, and Unisound leverage strong supply chain integration, energy-saving technologies such as liquid cooling, and early advantages in adapting to domestic chips, leading to concentrated market share among top players. These companies typically have large revenue scales, relatively stable gross margins (10%-20%), and good cash flow. Investment logic mainly focuses on growth in “volume” (shipment volume and market share) and “quality” (higher proportion of high-margin AI servers, penetration of liquid cooling solutions). As domestic chip performance improves, the proportion of domestically produced servers winning bids in government, finance, and telecommunications sectors has increased significantly, with high performance certainty, making them key targets for institutional allocation.

3. Data Centers and Cloud Computing: Data centers are the “reservoirs” and “dispatch centers” of computing power, mainly supporting the “East Data West Computing” strategy. Policies explicitly require new data centers to meet PUE (energy efficiency) standards, promoting greener and more centralized development. Western regions, benefiting from energy costs and climate advantages, are becoming preferred locations for large and super-large data centers. Companies like GDS, DataPort, and Aofei Data, along with the three major telecom operators, are accelerating their layout at hub nodes. Cloud service providers, offering domestically powered cloud services, lower the threshold for small and medium-sized enterprises to access computing power. This sector involves heavy asset operation, with large upfront capital expenditure and high depreciation pressure. However, once capacity utilization reaches a balance point, it generates stable and continuous cash flow, resembling bonds. Investment focus includes “resource reserves” (land, energy consumption indicators), “customer structure” (internet giants, government and enterprise clients), and “operational efficiency” (PUE, capacity utilization). With optimized computing power network scheduling and improved cross-regional trading mechanisms, the asset value of data centers is expected to be further revalued.

4. Software and Application Ecosystem: Software is the “soul” of computing power, with operating systems, databases, middleware, and application software forming the ecological foundation of domestically produced computing power. After hardware localization, software adaptation is key to user experience. Companies like China Software (Kylin OS), Chengmai Technology (Tongxin Software), Dameng Data, and Renmin JinCang are accelerating deep integration with domestic chips and operating systems. Under the support of the “Xinchuang” policy, software localization in key industries such as government, finance, and energy is expanding from office systems to core business systems. Software companies typically feature “high margins, high stickiness, and light assets.” Once ecological barriers form, user switching costs are high, enabling long-term recurring revenue (subscriptions, service fees). Investment logic emphasizes “ecological compatibility” (hardware adaptation), “customer penetration” (key industry coverage), and “productization capability” (standardization). As domestic computing hardware performance improves, the “barrel effect” of the software ecosystem will gradually diminish, and the software sector may see a “Davis double play” in valuation and performance.

Risk Warning: The industry information and company dynamics mentioned in this article are for overview purposes only and do not constitute any investment advice; business operations and market fluctuations involve uncertainties, please be aware of related risks.