Alibaba’s Breakthrough in Self-Developed AI Chips: A Booming Opportunity for Domestic Industry

Alibaba’s self-developed next-generation AI chip has achieved a significant breakthrough, compatible with the NVIDIA ecosystem and entirely produced by domestic enterprises, resulting in a market value increase of 250 billion yuan overnight.

Through its subsidiary, Pingtouge Semiconductor, Alibaba has developed and tested a new generation of chips aimed at AI inference tasks. This chip is primarily designed for AI inference, compatible with the NVIDIA software ecosystem (CUDA), and is completely manufactured by domestic companies. At the same time, Alibaba announced that it will invest 380 billion yuan over the next three years in cloud and AI hardware infrastructure, focusing on areas such as chips, computing power infrastructure, liquid cooling technology, and industry applications.

01 Alibaba’s Chip Strategy: Building an Autonomous and Controllable AI Ecosystem

Alibaba’s layout in the chip field is a core strategy for building the next generation of cloud computing and artificial intelligence infrastructure, primarily promoted through its subsidiary, Pingtouge Semiconductor Co., Ltd. Alibaba’s chip strategy revolves around three core areas: the AI chip strategy focuses on cloud-based AI training and inference chips, aiming to reduce dependence on NVIDIA GPUs; in terms of RISC-V ecosystem development, Pingtouge has launched the Xuantie series processors, becoming a significant promoter of the global RISC-V architecture; the full-stack technology layout encompasses everything from chip design to computing power infrastructure, creating a complete technology stack of “chip-server-data center-cloud service.” Alibaba’s self-developed AI chips are primarily intended to meet its own inference needs in the AI field, directly controlling chip design and production costs, and reducing reliance on external suppliers. In the context of the U.S. restrictions on NVIDIA’s export of advanced chips to China, Chinese tech companies are accelerating their shift towards domestic designs.

02 Chip Design and Manufacturing: Core Link for Domestic Substitution

In Alibaba’s AI chip ecosystem, companies involved in chip design and manufacturing are the core force for domestic substitution. Cambricon (688256), as the core supplier of Alibaba Cloud’s AI inference chips, has deeply adapted its Siyuan 590 chip to Alibaba Cloud servers. It received a 1 billion yuan order from Alibaba in the second quarter of 2025, becoming a significant force in replacing NVIDIA’s AI chips. Hygon Information (688041) has a clear demand for its deep computing DCU series GPUs, which are compatible with the CUDA ecosystem, in replacing NVIDIA’s A100. The company is jointly developing the next generation of training cards with Alibaba, expecting a 287% increase in AI chip revenue in 2024. Chipone Technology (688521) provides RISC-V GPU IP licensing for Pingtouge, covering smart home and automotive electronics. Its revenue grew by 69% in Q1 2025, securing multiple orders from Alibaba’s ecosystem projects in the AIoT field. In the manufacturing segment, SMIC (688981), as the largest domestic wafer foundry, undertakes the mass production of Alibaba’s AI chips using 6nm technology, achieving commercial yield standards; JCET (600584), as a global leader in the packaging and testing industry, is the exclusive packaging and testing provider for Alibaba’s chips, utilizing SiP/3D packaging technology with a monthly production capacity exceeding 5 million units.

03 Computing Power Infrastructure: The Foundation for AI Chips

The computing power infrastructure is the foundation for AI chips, including IDC service providers, server manufacturers, and optical module suppliers. DataPort (603881) is the core supplier of customized IDC for Alibaba Cloud, managing over 20 data centers, with Alibaba Cloud revenue accounting for more than 80%. Its liquid cooling technology reduces PUE to as low as 1.15, with an expected addition of over 5,000 new cabinets by 2025. Inspur Information (000977) is the main supplier of AI servers for Alibaba, with liquid cooling technology reducing PUE to below 1.1. It is expected to secure 50% of Alibaba’s new cabinet orders, with a projected 200% increase in AI server shipments in 2024. Inspur (300308) is the main supplier of 800G optical modules for Alibaba, with 1.6T products under validation, supporting high-speed interconnection in data centers. The market share of 800G optical modules is the highest globally, expected to increase to 35% by 2025. Hangang Co. (600126) provides 30% of the inference computing power for Alibaba’s Tongyi large model, securing a 15 billion yuan order in Q1 2025. The local IDC resource advantages in Zhejiang are significant, with a planned computing power scale reaching 10,000P.

04 Liquid Cooling Technology: Key Support for High-Density Computing Power

As the power consumption of AI data centers rapidly escalates from the traditional 20kW-30kW per rack (the fully equipped NVL AI server cabinet power may exceed 100kW), liquid cooling technology has become a key support for high-density computing power. Liquid cooling technology can significantly reduce cooling energy consumption, easily lowering PUE to an ultra-low level of 1.1-1.2, far superior to the 1.5 or higher of air cooling. Inspur (002837) provides a 50kW+ high-power liquid cooling solution, leading the industry in market share. Alibaba’s procurement accounts for over 60%, with power density per cabinet increasing threefold and energy consumption reduced by 40%. Gaolan Co. (300499) focuses on liquid cooling solutions for data centers, benefiting from the energy efficiency upgrade demand driven by Alibaba’s chips. Its immersion liquid cooling solution has been certified by Alibaba, with a projected 150% increase in orders in 2024. Yimikang (300249) is the exclusive supplier of precision air conditioning for Alibaba Cloud, with its liquid cooling solution receiving green certification. In Q1 2025, net profit increased by 34.77%, winning multiple projects for Alibaba’s data centers.

05 Applications and Ecosystem Cooperation: Pathways to Realizing Chip Value

The application and ecosystem cooperation segment transforms the value of AI chips into practical application scenarios, covering areas such as edge computing, fintech, and smart cities. Runhe Software (300339) develops edge computing modules based on Xuantie processors, with Cainiao Logistics holding over 30% market share. The HiHope chip platform supports multiple Xuantie processors, with related revenue expected to grow by 80% in 2024. Hang Seng Electronics (600570) collaborates with Alibaba Cloud to build financial AI solutions, with Ant Group holding 20.79% stake. Quantitative trading solutions based on Alibaba chips have been implemented by multiple brokerage firms. Qianfang Technology (002373) collaborates with Alibaba Cloud on the “City Brain 2.0” project, winning over 800 million yuan in Alibaba ecosystem projects in H1 2025. Smart traffic chips are developed based on Xuantie processors, deployed in over 100 cities.

06 Bohui Co.: A Rising Star in Liquid Cooling Technology Transformation

Bohui Co. (300839) indirectly participates in Alibaba’s AI chip ecosystem through its subsidiary Wuxi Extreme Liquid Cooling Technology, but does not directly engage in chip design/manufacturing, focusing mainly on liquid cooling support. The company leverages its parent company’s refining foundation to self-develop fluorinated liquid formulations that reduce system costs by 8%-12%, and collaborates with Alibaba Cloud to develop customized cooling solutions. Bohui Co. invested 390 million yuan to procure GPU servers from DingTalk service provider Xinfengwei, building a computing power pool of 200-300P for AI intelligent computing cloud and stablecoin on-chain verification. It also participates in Alibaba Cloud’s UAE data center project, exploring the co-construction of a Middle East data center, benefiting from tax incentives between China and the UAE. However, Bohui Co. faces challenges such as the need for technology validation, strong order dependency, and pressures from cross-industry transformation. Its traditional chemical business is experiencing losses (net loss of 59.28 million yuan in the mid-2025 report), and liquid cooling revenue has yet to scale, facing competition from established companies like Inspur.

07 Investment Perspective and Risk Warning

From an investment perspective, Alibaba’s chip concept stocks cover the entire industry chain from design, manufacturing to application. Cambricon and Hygon Information have first-mover advantages in the AI chip substitution field; DataPort and Inspur Information are most closely tied to Alibaba in the computing power infrastructure sector; Inspur and Gaolan Co. benefit from the explosive demand for high-density computing power in the liquid cooling technology field. It is important to note that this field carries risks such as rapid technological iteration, strong policy dependence, and uncertainty in performance realization. Some concept stocks have seen significant price increases, which may pose a risk of overvaluation. Investors should focus on companies that have secured clear orders, possess strong technical capabilities, and have deep collaborations with Alibaba. It is recommended to closely monitor Alibaba Cloud’s quarterly financial reports for capital expenditure guidance, Pingtouge’s chip mass production progress, and the specific order fulfillment situation of each company. Alibaba’s new AI chip is primarily aimed at inference tasks, designed to handle a broader range of AI inference applications rather than training tasks. Notably, it is compatible with the NVIDIA architecture, meaning engineers can reuse programs written for NVIDIA chips, significantly lowering the switching costs for users. Due to sanctions, the new chip is no longer manufactured by TSMC but is instead produced by a domestic company. This shift marks a significant step for China in building a self-sufficient AI supply chain. Alibaba Cloud’s revenue grew by 26% year-on-year, reaching 33.398 billion yuan, the highest growth rate in three years. More importantly, revenue from AI-related products has seen triple-digit growth for eight consecutive quarters, accounting for over 20% of external commercialization revenue. The rapid application of artificial intelligence in the Chinese economy has generated enormous demand for inference, which typically does not require the most advanced chips, providing a vast market space for domestic chips.