China’s AI Supernodes Take Center Stage at Shanghai Conference as US Congress Visit Gets Snubbed

At this year’s World Artificial Intelligence Conference (WAIC) in Shanghai, domestic “AI supernode” products became the main attraction [203868][200082]. These supernodes are giant supercomputers made by linking hundreds or thousands of individual chips to work as one unit, allowing companies to run massive AI models that a single chip cannot handle [203868]. Huawei Technologies and Biren Technology both unveiled new hardware designed to create supernodes, signaling a major shift in strategy for Chinese firms, which are now racing to master chip networking instead of competing solely on individual chip performance [203868]. Industry experts at the event noted that the products signal a growing shift toward self-reliance in China’s AI infrastructure, as they allow companies to process data faster without relying on foreign-made chips [200082].

Meanwhile, shares of Chinese artificial intelligence company Zhipu surged 37 percent in Hong Kong after the firm completed a massive 1-gigawatt data centre running entirely on Chinese-made chips [201619]. The facility is an AI computing centre used to train and run complex AI models, and the company will use it to develop its GLM series of AI models [201619].

The push for technological independence comes as China’s biggest technology companies refused to meet a visiting US congressional delegation, denying American officials a rare look inside one of the world’s fastest-moving AI industries [203829]. Analysts say the snub highlights Beijing’s increasingly selective approach to AI diplomacy, revealing a paradox: while the US and China continue to talk about AI safety, the channels that once helped them understand each other are closing [203829].