share_log

航锦科技(000818):抢占人工智能先机 构建英伟达算力全栈式服务

Hangjin Technology (000818): Seize AI pioneers to build Nvidia's computing power full-stack service

中航證券 ·  Jan 27

Report summary

Seize the opportunity of artificial intelligence and build a full-stack Nvidia computing power service. On July 25, 2023, the company established Hangjin (Wuhan) Artificial Intelligence Technology Co., Ltd. (hereinafter referred to as “Hangjin Artificial Intelligence”) to enhance the positioning of artificial intelligence in the company's strategic layout. On November 26, 2023, the shareholders of Hangjin Artificial Intelligence and Wuhan Chaoqing Digital Intelligence Technology Co., Ltd. (hereinafter referred to as “Superengine Digital Intelligence”) jointly signed the “Investment Agreement on Wuhan Chaoqing Digital Intelligence Technology Co., Ltd.”. Hangjin Technology will have 46.47% of the voting rights of Super Engine Digital Intelligence, become its largest shareholder, and merged.

SuperEngine Digital Intelligence mainly provides AI server+lossless network+optical connectivity solutions for application scenarios such as data centers, high-performance computing, edge computing, and artificial intelligence. It is the world's leading artificial intelligence company NVIDIA (NVIDIA) Networking (Network) ElitePartner (Elite Partner) and Compute (GPU) Preferred Partner.

On January 26, 2024, Hangjin Artificial Intelligence, SuperEngine Digital Intelligence, and Ziguang Xiaotong Technology Co., Ltd. established a strategic partnership on matters related to NVIDIA's latest generation GPU product cluster, NVIDIA Mellanox Infini-Band network, NVIDIA BlueField-3 DPU, and the provision of software and services, and signed a “Strategic Cooperation Framework Agreement” for a period of five years.

In the AI era, networks define data centers

Data centers conduct distributed AI model training and generative AI. This requires powerful network connections to a large number of GPU nodes for computation. Therefore, data center networks are leading the progress of the AI era. Nvidia provides two computing power networking solutions, InfiniBand and Ethernet. According to Nvidia statistics, of the top 500 supercomputers in the world, 357 (accounting for about 71% 6) use Nvidia network architectures, including 291 InfiniBand systems and 138 Ethernet systems. Among the top 100 supercomputers, InfiniBand accounts for 62%. Based on the existing business, Hangjin Technology makes full use of SuperEngine's technical accumulation and project experience in networks and computing power, as well as the design, R&D, and product advantages of SuperEngine Digital's complete computing power solutions to enter the computing power circuit, enhance the positioning of artificial intelligence in the company's strategic layout, and add new development momentum to the company.

Join hands with Golden Shovel+Well Digger to transform AI into a water seller

We believe that in order to maximize the return on AI investment, it is necessary to have high-performance computing power chips (gold shovels) and the ability to deploy large-scale computing power networks to achieve end-to-end solutions for AI network architectures (well diggers). In the current complex supply context, based on NVIDIA's latest generation GPU for the Chinese market in 2024, Ziguang Xiaotong acted as NVIDIA's general agent to provide strong hardware technical resources and software ecosystem support for Hangjin Technology's pioneers, enabling Hangjin Technology to take control of the “golden shovel”. As a dual elite partner of Nvidia's Compute (GPU) and Network (network), SuperEngine Digital Intelligence has the ability to deliver overall artificial intelligence solutions, which can enable Hangjin Technology to build a cost-effective “computing power well”. Through strong alliances with Shovel+Construction Team, we are optimistic about Hangjin Technology's transformation into an “AI computing power seller” and maintain the company's “buy” rating.

Risk warning

There is a risk that the company's business prices will fall due to internal and external factors such as the macro environment and trade frictions; the US limits on China's computing power are becoming more and more strict, and there is a risk that Nvidia GPU supply will be scarce domestically; large-scale profitable artificial intelligence applications have not yet appeared, and future computing power supply and demand gaps or fall short of expectations; the agreement signed by the company is a framework of cooperation, and there is uncertainty about the signing time, contract amount, and actual execution amount of specific project contracts; some Internet vendors are developing their own networking solutions.

The translation is provided by third-party software.


The above content is for informational or educational purposes only and does not constitute any investment advice related to Futu. Although we strive to ensure the truthfulness, accuracy, and originality of all such content, we cannot guarantee it.
    Write a comment