Chinese AI chip and model firms push capacity, cooling and monetization plans
China’s AI compute news centered on chip supply constraints, server cooling investment, edge inference silicon and paid AI-model services.
• Alibaba Cloud plans to more than triple its global data center capacity, according to a TMTPost morning brief.
• Qiangrui Technology plans a private placement of up to 1.05 billion yuan for AI server precision liquid-cooling components and other projects.
• Nations Technologies will raise prices on some high-performance MCUs used in industrial applications, AI power supplies and digital power by 10% to 20% from Oct. 1, citing longer international vendor lead times, domestic substitution demand and AI-related upstream capacity pressure.
• Axera said its next-generation high-computing-power AI chip has completed tape-out and supports two-chip or four-chip cascading for full-capability large-model inference at the edge.
• Axera also said it raised inventory from more than 300 million yuan to more than 700 million yuan in the first half to lock in advanced process node capacity amid tight industry supply.
• SK Hynix reportedly restarted construction of its second NAND flash plant in Dalian, with plans to install equipment before year-end and begin formal production in the first half of next year, lifting local capacity by about 50%.
• Ant Bailing open-sourced the Ling-3.0-tiny model on Hugging Face, offering BF16, FP8 and INT4 weights and supporting deployment on NVIDIA DGX Spark and Apple Mac Mini.
• Alibaba’s Qwen app launched paid Office Assistant plans for professional users, with annual pricing up to 1,499 yuan while basic features remain free.