AI is making a comeback again, with copper high-speed connections, computing power leasing, optical modules and other sectors leading the gains.
Recently, the news of AI continues to ferment, and there may be new hype stories for AI, which defines computing power from the direction of quantum technology and clarifies quantum computing as a new type of computing power.
The main logic is also the American Dream and the over-falling rebound, driven by domestic substitution.
Four major news fermentation
First, on April 25, the Beijing Municipal Bureau of Economy and Information Technology and the Beijing Municipal Communications Administration jointly issued the "Implementation Plan for the Construction of Computing Power Infrastructure in Beijing (2024-2027)".
Concentrate on the construction of a number of large clusters of intelligent computing, and by 2025, the city's intelligent computing supply scale will reach 45EFLOPS;
By 2027, the full-stack independent and controllable software and hardware products of intelligent computing infrastructure will be realized, and the construction capacity of 100% independent and controllable intelligent computing center will be realized.
Second, the Beijing-Tianjin-Hebei Tianhe computing network platform was officially released a few days ago.
The platform aims to fully integrate the computing power and network resources of the Beijing-Tianjin-Hebei region, and strive to build three "exascale" supercomputing intelligent computing integrated computing power networks, namely exascale supercomputing capabilities, exascale intelligent computing capabilities, and exabytes of large-scale data management capabilities.
Third, overseas technology giants have raised AI capital expenditures.
Meta raised its capex guidance for this year in its Q1 earnings report, saying it had significantly raised its capex forecast for this year from $30 billion to $37 billion to $35 billion to $40 billion as it accelerated infrastructure investment to support its AI roadmap.
Microsoft's chief financial officer, Amy Hood, said the company's capital expenditure in the most recent quarter was $14 billion and is expected to continue to increase "substantially."
Fourth, China Mobile recently launched the procurement of new intelligent computing centers from 2024 to 2025.
According to the tender announcement, the procurement scale of this project reached 8,054 units.
The project is divided into two standard packages: Standard Package 1 is 7,994 artificial intelligence servers and ancillary products, and the number of winning bidders is 5 to 7, and Standard Package 2 is 60 white-box switches, and the number of winning bidders is 1. This project is the largest centralized procurement of artificial intelligence servers in the industry.
Computing power industry chain
According to the concept, computing power can be divided into three parts: AI chips, computing power servers, and computing power leasing.
AI chips mainly include Haiguang Information, Cambrian-U, Jingjiawei and other companies.
In terms of servers, A-share companies mainly include Inspur Information, ZTE, Industrial Fortune Union, etc., as well as HW server-related manufacturers, including Hengwei Technology, Digital China, Tuowei Information, Sichuan Changhong, Tianyuan Dike, High-tech Development, China G Great Wall, etc.
The theme of computing power leasing mainly tests the ability of listed companies to "take cards".
According to public information, Hengrun shares, Lotus Health, Zhongbei Communications, Jinji shares and other stocks plan to build a computing power scale of more than 10,000 P, and entrepreneurial dark horses, Winner Technology, Hongbo shares, Youkede, Runjian shares, Leon Technology and other companies plan to build a computing power scale of more than 1,000 P.
Among them, the upstream of the computing server industry chain is dominated by raw materials such as CPU, GPU, memory, hard disk, RAID controller, power supply, and software system, while the midstream is the server industry.
Downstream customer groups include Internet cloud service providers, telecom operators, third-party IDC service providers, government departments, and various types of enterprises.
According to the analysis of a brokerage research report, computing power is a system engineering, and at present, the mainland has a relatively complete computing power industry system, including chips, optical modules, switches, servers, connectors, PCBs and other links, and the promotion of computing power will accelerate the performance release of each link.
(Summary of related industrial chains, not recommended)
1) Computing infrastructure: ZTE, Unisplendour, Digital China, Tuowei Information, Industrial Fortune Union, Inspur Information, Feiling Kesi, Gongjin Co., Ltd.;
2) Computing power connection: Zhongji Innolight, Accelink Technology, Huagong Technology, Tianfu Communication, Yuanjie Technology, Borche Technology, Xinyisheng;
3) Computing power supply: Runze Technology, Capital Online, Dataport, Baoxin Software, Alpha Data;
4) Computing power cloud: Capital Online, UCD, Qingyun Technology, Yunsai Zhilian;
5) Computing power visualization and operation and maintenance: Hengwei Technology, Yakang Co., Ltd., Xinju Network;
6) Computing power liquid cooling: Invic, Jialitu, Shenling environment;
7) Computing power services: Runjian shares, Yakang shares.