China Trials Largest Homegrown AI Computing Cluster with 15 Eflops Power
February 7, 2026
China’s state-backed SuperComputing Network (SCNet) began beta testing a new AI computing node on Thursday in Zhengzhou, Henan province. This node is powered by the country’s largest domestically built AI computing cluster. It uses three scaleX platforms made by Chinese supercomputer maker Sugon. The platforms support more than 30,000 AI acceleration cards. SCNet and Sugon announced this deployment but did not reveal the total computing power. However, based on a single scaleX's capabilities, the three platforms combined could deliver 15 eflops. An eflop means exa floating-point operations per second, or a quintillion calculations every second. For comparison, the world’s top supercomputer, El Capitan by Hewlett Packard Enterprise, reaches 1.8 eflops. This move shows China’s major push to build large-scale AI computing infrastructure using local technology.
Read More at Scmp →
Tags:
China
Ai Computing
Supercomputing Network
Sugon
Scalex Platform
Eflops
Comments