Huawei Technologies has unveiled its high-performance AI computing system, CloudMatrix 384, in a major public debut at the World Artificial Intelligence Conference in Shanghai, signaling a bold challenge to Nvidia’s dominance in advanced AI hardware.

The system, which integrates 384 of Huawei’s proprietary 910C chips, has been touted by analysts as a serious competitor to Nvidia’s GB200 NVL72 platform—widely regarded as the United States chipmaker’s most advanced AI system to date.

SemiAnalysis founder Dylan Patel said in an April report that Huawei’s system capabilities now rival—and in some aspects, surpass—those of Nvidia.

CloudMatrix 384 drew significant attention at WAIC, one of China’s premier AI events, although Huawei staff declined to provide details at the booth.

The company has also yet to officially comment on media inquiries about the system.

The platform’s performance, according to SemiAnalysis, stems not from the superiority of individual chips but from Huawei’s system-level design.

The CloudMatrix leverages a “supernode” architecture that enables ultra-fast interconnection among its chips, offsetting any performance gap with Nvidia’s B200 chips used in the GB200