Huawei escalated its push into AI infrastructure, unveiling an Atlas 950 “SuperCluster” built on its Ascend chips and claiming performance leadership over forthcoming Nvidia systems and Elon Musk’s xAI Colossus. The Chinese firm outlined a roadmap through 2028 to double compute annually and scale from hundreds of thousands to more than a million chips in future clusters, a strategy designed to offset U.S. curbs on advanced semiconductors by networking vast numbers of domestic processors. Nvidia acknowledged the competitive pressure as Beijing intensifies scrutiny of the U.S. chipmaker, including an extended antitrust probe and reported limits on certain workstation chips. Analysts cautioned Huawei’s performance claims warrant validation but said the company’s ambitions align with China’s bid for tech self-reliance. Huawei said it has already deployed hundreds of Atlas supernodes across industries, while Nvidia shares slipped amid China headlines.
Related articles:
— Blackwell (microarchitecture)
— Huawei
— Nvidia
— Semiconductor industry in China
