China reports a domestic 100,000-card AI cluster as memory and AI infrastructure expand
China’s AI compute sector saw new domestic supercluster capacity, an Inner Mongolia AI base opening, and developments involving memory chips, accelerators and large-model integration.
• China’s first fully domestic 100,000-card AI supercluster has entered operation at the Zhengzhou core node of the National Supercomputing Internet, with multiple computing nodes also expanding capacity.
• Envision said its Galaxy base in Ulanqab, Inner Mongolia, has begun operations, with a 120,000-square-meter AI supercomputing center planned to reach more than 1 million PFLOPS and support up to 1 million AI accelerator chips when fully built.
• Apple is reportedly testing memory chips from China’s CXMT for iPhones and MacBooks, while preliminary supply discussions aim to support some devices sold in China.
• Apple removed a Simplified Chinese support document describing Apple Intelligence integration with Alibaba’s Qwen, and customer service said the feature has not launched in mainland China.
• Cambricon reported first-half revenue of about 6 billion yuan and more than doubled net profit, while a sharp increase in unfinished inventory leaves the conversion of materials into finished chips and sales as a second-half focus.
• The China Advertising Association held a second meeting to solicit feedback on group standards for generative engine optimization, with participants including large-model vendors, internet platforms, media and research institutions.