Chinese AI chips, memory and model labs see fresh activity
China’s AI compute sector saw new memory-chip progress, IPO plans, model releases and policy support, while several AI companies expanded funding and product deployments.
• ChangXin Memory Technologies reportedly began risk trial production of HBM3E, and its DRAM capacity is expected to reach 350,000 wafers per month by year-end.
• Enflame Technology priced its STAR Market IPO at 142.18 yuan per share, plans to issue 43.035173 million shares, and expects to raise 6.119 billion yuan.
• Memory chipmaker Longsys filed for a Hong Kong listing seeking up to HK$6.28 billion and expects to start trading on September 8.
• DeepSeek released DeepSeek-V4-Flash-Vision-Exp on Hugging Face under the MIT license, while Zhipu outlined a next-generation large-model roadmap centered on a large base model and day-one business usability.
• Alibaba Qwen launched Agent Teams in Qwen Creative with Wan 3.0, and Tencent Hunyuan said Hy4 preview traffic surged after WorkBuddy integration, forcing it to expand its inference cluster.
• The MIIT launched a campaign to cultivate AI application service providers and said it will increase procurement of large models, agents and token-based services.
• Kuaishou said Beijing Kling separately signed accession agreements with the National Artificial Intelligence Fund and CP Robot, which will inject RMB 1.4 billion and about US$19.29 million, respectively.