Western: Accelerated landing of AI applications, dual prosperity cycles of electric power and computing infrastructure initiated.

Western released a research report, stating that transformers, SVG, power quality equipment, distribution equipment, and other power equipment core in data centers are in the high-volume growth stage. Breakthroughs in AI technology and the application of AI lead to an increase in computing power demand, which in turn leads to the expansion or construction of data centers, resulting in increased procurement of power equipment, forming a chain of demand transmission. The continuous increase in demand for copper connections brought about by the construction of data centers should be noted. Combined with the advancement of electricity market reform, the emergence of AI + power trading platforms as a new landing scene, the market will gradually realize the layout from 0 to 1, with existing service providers in the power generation end being the first to complete customer volume and product layout. The main points of Western are as follows: Alibaba's Thousand Questions releases QwQ-32B model, performance comparable to DeepSeek-R1 On March 6, 2025, Alibaba's TTY Thousand Questions team officially launched the open source inference model QwQ-32B, which achieved performance comparable to the 671 billion parameter DeepSeek-R1 with 32 billion parameters. QwQ-32B uses large-scale reinforcement learning technology to excel in mathematical reasoning (AIME24 evaluation set), programming ability (LiveCodeBench), and general tasks such as Meta's LiveBench and Google's IFEval, surpassing DeepSeek-R1 in some indicators. Its innovation lies in combining reinforcement learning with basic models, significantly reducing deployment costs, supporting consumer-level GPU local operation, and being open source under the Apache 2.0 license to promote the popularization of AI technology. Chinese team globally launches Manus AI Agent, leading the wave of intelligent agent applications The world's first general-purpose AI Agent product "Manus" developed by the Chinese team Monica.im has attracted market attention on March 6, 2025. Through a multi-agent architecture (planning, execution, verification modules coordination), the product can autonomously complete complex tasks such as resume screening, travel planning, etc., and refreshes SOTA scores in the GAIA benchmark test. AI scenario applications are entering an accelerated deployment phase. Multiple listed companies sign intensive computing power service contracts, electricity infrastructure welcomes prosperity cycle On March 4, 2025, a wholly-owned subsidiary of Zhejiang Haikong Nanke Huatie Digital Intelligence & Technology signed a "Computing Power Service Agreement" with X company in Hangzhou, with a total contract value of 3.69 billion yuan (including tax) for a service period of 5 years. Glory View Technology signed a 3-year 486 million yuan computing power service contract with Shenzhen X company, and on March 5, 2025, a wholly-owned subsidiary of Lanyun Technology, Aluoyu Technology, preliminarily signed a "Cloud Computing Power Service Agreement" with X company, with a total contract value of 3.707 billion yuan (including tax) and a service period of 5 years. The total of these three orders amounts to nearly 7.9 billion yuan, reflecting a sharp increase in demand for data center construction. High-density computing power support is required for AI large model training and inference, and intelligent agent task execution According to Sina.com, for example, QwQ-32B requires hundreds of TOPS of computing power for single inference, while the 24/7 task load of Manus drives the transformation of data centers from "machine rooms" to "computing power factories." In addition, the popularization of AI intelligent agent applications may also lead to a subscription-based service model based on computing power leasing, provide stable long-term demand for power equipment suppliers. The increased requirements for the stability and efficiency of the power system brought about by high-performance AI models may drive the research and application of high-end products such as SVG and power quality equipment. Risk warning: AI technology development falls short of expectations, data center construction progress slows down, and market competition intensifies.