Fujian maps AI infrastructure push as model deployments expand
China’s AI compute news centered on provincial infrastructure planning, domestic model deployment, and continued attention to AI hardware supply chains and server demand.
• Fujian’s 15th Five-Year Plan calls for stronger AI infrastructure, a provincewide integrated computing power network, public AI data centers in Fuzhou, Xiamen and Zhangzhou, and strategically deployed edge computing nodes.
• Fujian also plans to develop integrated circuits, compound semiconductors, emerging displays and optoelectronic components, including support for high-computing-power, low-power chips.
• Huawei said Ascend Atlas 800 A3 and Atlas 900 A3 SuperPoD supernodes have completed adaptation for Xiaohongshu’s open-source dots3-note preview model through the vLLM Ascend inference engine.
• The national supercomputing internet launched DeepSeek V4 Pro and the Harness agent framework, providing domestic computing power support for training through deployment.
• Qianwen Office added GLM-5.3 and DeepSeek V4 Pro to its “Frontier Models” tier for direct use by users.
• Tencent said it could make immediate profits by renting out a $53 billion hardware investment for AI workloads, but plans to build models because it sees longer-term returns from selling tokens.
• Centec Networks signed an Ethernet switch chip sales contract worth more than 100 million yuan.
• Market commentary said China’s AI computing hardware and server supply-chain names remained strong, while large-model stocks rose sharply before pulling back.