First order released! Leading supercomputing company raises 8 billion yuan in additional financing

Last week, the largest domestic AI computing power pool was put into operation, and this week, Sugon (Inspur) has made new moves in computing power.

On the evening of February 9, Sugon announced that it plans to issue convertible bonds to unspecified targets to raise no more than 8 billion yuan, which will be used for projects including an advanced computing cluster system for artificial intelligence, a next-generation high-performance AI training and inference integrated machine, and a domestically produced advanced storage system. This is also the first refinancing demonstration case in the Shanghai Stock Exchange after the release of refinancing optimization measures.

Among them, the advanced computing cluster system project for artificial intelligence plans to invest 3.5 billion yuan, focusing on developing the next-generation advanced computing super-node hardware system and high-speed interconnection system, while also integrating various hardware such as CPUs, GPUs, and NPUs into a unified management system, to build an intelligent scheduling platform capable of handling large model training, inference, scientific computing, and cloud services across multiple tasks.

The next-generation high-performance AI training and inference integrated machine project plans to invest 2.5 billion yuan, divided into four main modules — hardware R&D for high-performance AI training and inference integrated machines, a one-stop deployment and management platform for large models, training and inference acceleration toolchains, and a large model integrated machine service platform, promoting the deep integration of AI technology with the real economy and accelerating industrial intelligent transformation and upgrading.

The domestically produced advanced storage system project, based on domestically produced processor chips and other core components, aims to develop domestically produced advanced storage systems. The project plans to invest 2 billion yuan.

“This refinancing will help the company grasp the latest trends in the AI industry, and build a competitive moat around intelligent computing power technology systems,” Sugon stated in the announcement. The company can leverage industry trends and demands to better utilize its technological and product advantages as a leading enterprise, accelerate the localization process in the computing infrastructure industry, achieve high-level technological independence and self-reliance, and build a domestic computing power moat.

Currently, AI is driving a reshaping of the global industrial landscape, and large-scale deployment requires breakthroughs in computing infrastructure to move from “technological feasibility” to “commercial usability.”

Previously, Sugon stated during institutional research that the company has significant growth potential in areas such as domestic computing power, AI computing expansion, and breakthroughs in computing technology architecture. Additionally, its global layout and ecological collaboration further open up long-term growth opportunities.

“Going forward, the company will continue to focus on core high-end computing business, and based on breakthroughs in frontier technologies such as super-node intelligent computing power, scientific large model development platforms, and super-cluster systems, it will continue to develop full-stack layouts in intelligent computing, computing power scheduling, data center solutions, and other fields, improving the full industrial chain system capabilities of ‘core—edge—cloud—computing,’” Sugon stated.

The reporter has noticed that recently, Sugon has been actively expanding its computing power business.

On February 5, three super-cluster systems provided by Sugon, totaling 30,000 cards, were simultaneously launched for trial operation at the national supercomputing internet core node, becoming the largest domestically produced AI computing power pool in China with 30,000 cards deployed and operational. This marks a major breakthrough in the scale and practicality of domestic AI infrastructure.

Prior to this, on November 6, 2025, during the World Internet Conference Wuzhen Summit, Sugon officially released the world’s first 640-card super-node scaleX640 in a single cabinet. It is built on a globally leading open system hardware architecture, driven by “open architecture + system innovation.”

According to reports, scaleX640 super-node adopts an “AI computing open architecture.” In September 2025, Sugon, in collaboration with more than 20 industry chain companies, jointly released this architecture and opened multiple key technological capabilities, aiming to lower the R&D threshold for AI clusters and avoid redundant investments. It is reported that the scaleX640 super-node supports multi-brand accelerators at the hardware level, is compatible with mainstream computing ecosystems at the software level, and supports frontier scenarios such as trillion-parameter MoE large model training, high-throughput inference, and scientific intelligence (AI4S).

Sugon stated that the scaleX640 super-node is accelerating the shift of AI computing power from “expensive resource” to “inclusive infrastructure,” helping China build a new ecosystem of computing power that is both autonomous and controllable and open and mutually beneficial in the global intelligent competition.

(Source: Shanghai Securities News)