National Supercomputing Internet Launches Core Node, Offering 30,000+ Domestic AI Chips in Trial Run

A core node of China's National Supercomputing Internet officially commenced trial operations on February 5th in Zhengzhou. Marking a significant leap in the country's integrated computing infrastructure, this node provides access to a massive, single-site pool of over 30,000 domestic AI accelerators.

The node is powered by the Sugon scaleX ultra-large-scale cluster system, designed to meet the explosive demand for computing power from large language model training, high-throughput inference, and AI for Science. It represents the largest single domestic AI computing resource integrated into the national platform since its launch.

Built on an open AI computing architecture, the scaleX cluster boasts compatibility with mainstream software ecosystems like CUDA and supports hybrid deployment of multi-brand domestic accelerator cards. It has been optimized for over 400 mainstream large-scale and scientific models. Through the National Supercomputing Internet platform (www.scnet.cn), this computing power can be efficiently scheduled alongside other resources, providing accessible and user-friendly AI computing services.

The platform's operational success is evident, having served over 1 million users by the end of 2025, with more than 7,300 application solutions available. It has cumulatively supported 196 million computing jobs. A trial invitation program was also announced, offering both lightweight testing and large-scale application adaptation tests for professional users, as well as free resources for startups and students.

The launch of this core node is a strategic move to address the dual challenge of computing power shortages and underutilization of existing resources. It aims to elevate the allocation efficiency of national computing infrastructure, empower regional industrial upgrades in Central China, and solidify the foundation for the "AI+" initiative across various sectors.